{"blog-article.author":"Author","blog-article.category":"Category","blog-article.details":"Article details","blog-article.published":"Published","blog-article.readTime":"Read time","blog-article.story":"Article","blog-article.tags":"Tags","blog-article.toc":"In this article","blog-article.7599438ac9":"Every release, every change, every upgrade path.","blog-article.cd5af7344f":"Release notes","blog-article.164b3d4809":"Changelog","blog-article.5ef37eb675":"Active projects across six research tracks.","blog-article.ab4c72dd0d":"Research programme","blog-article.be601df25e":"Research","blog-article.77f3603653":"Peer-reviewed papers and sector briefings. All open-access.","blog-article.889ba28374":"Technical papers","blog-article.1e93263c97":"Papers","blog-article.946796817e":"Engineering signal, research threads, and honest postmortems.","blog-article.91570fb542":"All articles","blog-article.0b9d2b2362":"Blog","blog-article.4f4840a823":"Continue reading","blog-article.d4a046c345":"What next","blog-article.894876f6d6":"Back to blog","blog-article.2bf93931c7":"min read","blog-article.canvas.footerInstruction":"hover or use tab/enter","blog-article.canvas.hoverInstruction":"hover parts for detail","blog-article.87df60de33":"Resources","blog-article.cc5cefbb9e":"diagramScroll","blog-article.b848452bc9":"Dweve","blog-visuals.agent-reliability-gap.copy.checked":"checked agent","blog-visuals.agent-reliability-gap.copy.checked.detail":"Tools, policy gates, receipts and recovery logic close the reliability gap.","blog-visuals.agent-reliability-gap.copy.control":"control","blog-visuals.agent-reliability-gap.copy.control.detail":"Planning, action and recovery fail in different ways, so the controls differ too.","blog-visuals.agent-reliability-gap.copy.demo":"demo","blog-visuals.agent-reliability-gap.copy.demo.detail":"A demo can survive one happy path and a forgiving operator.","blog-visuals.agent-reliability-gap.copy.operations":"operations","blog-visuals.agent-reliability-gap.copy.operations.detail":"Real operations punish retries, missing permissions and partial state changes.","blog-visuals.agent-reliability-gap.copy.steps":"steps","blog-visuals.agent-reliability-gap.copy.steps.detail":"The slider raises how many actions the agent may take before a human checks it.","blog-visuals.agent-reliability-gap.copy.verdict":"the gap","blog-visuals.agent-reliability-gap.copy.verdict.detail":"Agents become useful when the organisation can trust the boring failures, not only the impressive success path.","blog-visuals.agent-reliability-gap.eyebrow":"Agent reliability","blog-visuals.agent-reliability-gap.mode.act":"act","blog-visuals.agent-reliability-gap.mode.plan":"plan","blog-visuals.agent-reliability-gap.mode.recover":"recover","blog-visuals.agent-reliability-gap.range":"autonomy pressure","blog-visuals.agent-reliability-gap.title":"Actions are where demos become systems","blog-visuals.ai-agent-loop.copy.act":"act","blog-visuals.ai-agent-loop.copy.act.detail":"Tool calls, messages and transactions change external state.","blog-visuals.ai-agent-loop.copy.autonomy":"autonomy","blog-visuals.ai-agent-loop.copy.autonomy.detail":"The slider increases how many steps the agent may take before asking a human.","blog-visuals.ai-agent-loop.copy.check":"check","blog-visuals.ai-agent-loop.copy.check.detail":"Policy, budget and evidence gates decide whether the action is allowed.","blog-visuals.ai-agent-loop.copy.plan":"plan","blog-visuals.ai-agent-loop.copy.plan.detail":"A goal becomes a short sequence of candidate actions.","blog-visuals.ai-agent-loop.copy.sense":"sense","blog-visuals.ai-agent-loop.copy.sense.detail":"The agent reads the world before it decides anything.","blog-visuals.ai-agent-loop.copy.stage":"stage","blog-visuals.ai-agent-loop.copy.stage.detail":"The mode selects which part of the loop is being inspected.","blog-visuals.ai-agent-loop.copy.state":"state","blog-visuals.ai-agent-loop.copy.state.detail":"The result updates the agent state for the next turn.","blog-visuals.ai-agent-loop.eyebrow":"Agent control loop","blog-visuals.ai-agent-loop.mode.act":"act","blog-visuals.ai-agent-loop.mode.check":"check","blog-visuals.ai-agent-loop.mode.plan":"plan","blog-visuals.ai-agent-loop.mode.sense":"sense","blog-visuals.ai-agent-loop.mode.state":"state","blog-visuals.ai-agent-loop.range":"autonomy budget","blog-visuals.ai-agent-loop.title":"An agent is a loop with consequences","blog-visuals.ai-attention-map.copy.context":"context vector","blog-visuals.ai-attention-map.copy.context.detail":"Attention blends weighted token information into the next representation.","blog-visuals.ai-attention-map.copy.focus":"focus","blog-visuals.ai-attention-map.copy.focus.detail":"The slider changes how concentrated the attention weights become.","blog-visuals.ai-attention-map.copy.token.detail":"A token receives weight depending on the active query and head.","blog-visuals.ai-attention-map.copy.view":"view","blog-visuals.ai-attention-map.copy.view.detail":"The mode changes which attention component is highlighted.","blog-visuals.ai-attention-map.eyebrow":"Attention map","blog-visuals.ai-attention-map.mode.head":"head","blog-visuals.ai-attention-map.mode.key":"key","blog-visuals.ai-attention-map.mode.query":"query","blog-visuals.ai-attention-map.mode.value":"value","blog-visuals.ai-attention-map.range":"focus sharpness","blog-visuals.ai-attention-map.title":"Attention is weighted routing, not magic focus","blog-visuals.ai-data-quality-kitchen.copy.bias":"bias","blog-visuals.ai-data-quality-kitchen.copy.bias.detail":"Biased data teaches the system a biased version of the world.","blog-visuals.ai-data-quality-kitchen.copy.coverage":"coverage","blog-visuals.ai-data-quality-kitchen.copy.coverage.detail":"Rare cases count, especially the boring ones customers actually hit.","blog-visuals.ai-data-quality-kitchen.copy.freshness":"freshness","blog-visuals.ai-data-quality-kitchen.copy.freshness.detail":"Stale examples make yesterday look like a law of physics.","blog-visuals.ai-data-quality-kitchen.copy.ingredient":"ingredient","blog-visuals.ai-data-quality-kitchen.copy.ingredient.detail":"The mode selects the quality ingredient under inspection.","blog-visuals.ai-data-quality-kitchen.copy.labels":"labels","blog-visuals.ai-data-quality-kitchen.copy.labels.detail":"The answer key has to mean the same thing across examples.","blog-visuals.ai-data-quality-kitchen.copy.mess":"mess","blog-visuals.ai-data-quality-kitchen.copy.mess.detail":"The slider increases missing labels, inconsistent units and unknown provenance.","blog-visuals.ai-data-quality-kitchen.copy.needsPrep":"needs prep","blog-visuals.ai-data-quality-kitchen.copy.trainingSet":"training set","blog-visuals.ai-data-quality-kitchen.copy.trainingSet.detail":"The model learns the dataset it receives, not the neat story on the slide.","blog-visuals.ai-data-quality-kitchen.copy.usable":"usable","blog-visuals.ai-data-quality-kitchen.eyebrow":"Data quality kitchen","blog-visuals.ai-data-quality-kitchen.mode.bias":"bias","blog-visuals.ai-data-quality-kitchen.mode.coverage":"coverage","blog-visuals.ai-data-quality-kitchen.mode.freshness":"freshness","blog-visuals.ai-data-quality-kitchen.mode.labels":"labels","blog-visuals.ai-data-quality-kitchen.range":"dataset mess","blog-visuals.ai-data-quality-kitchen.title":"The model eats the dataset you actually prepared","blog-visuals.ai-embedding-space.copy.code":"code","blog-visuals.ai-embedding-space.copy.docs":"docs","blog-visuals.ai-embedding-space.copy.images":"images","blog-visuals.ai-embedding-space.copy.operation":"operation","blog-visuals.ai-embedding-space.copy.operation.detail":"The mode selects clustering, distance, search or binary comparison.","blog-visuals.ai-embedding-space.copy.point.detail":"Nearby points are treated as semantically related by retrieval.","blog-visuals.ai-embedding-space.copy.points":"points","blog-visuals.ai-embedding-space.copy.points.detail":"The slider changes how crowded the semantic space becomes.","blog-visuals.ai-embedding-space.eyebrow":"Embedding space","blog-visuals.ai-embedding-space.mode.binary":"binary","blog-visuals.ai-embedding-space.mode.cluster":"cluster","blog-visuals.ai-embedding-space.mode.distance":"distance","blog-visuals.ai-embedding-space.mode.search":"search","blog-visuals.ai-embedding-space.range":"space density","blog-visuals.ai-embedding-space.title":"Meaning becomes distance you can search","blog-visuals.ai-explainability-lens.copy.blackbox":"model","blog-visuals.ai-explainability-lens.copy.blackbox.detail":"A black-box score is not yet an explanation someone can act on.","blog-visuals.ai-explainability-lens.copy.causal":"causal","blog-visuals.ai-explainability-lens.copy.causal.detail":"Which factors caused movement instead of merely correlating?","blog-visuals.ai-explainability-lens.copy.counter":"counterfactual","blog-visuals.ai-explainability-lens.copy.counter.detail":"What smallest change would flip the outcome?","blog-visuals.ai-explainability-lens.copy.global":"global","blog-visuals.ai-explainability-lens.copy.global.detail":"What patterns does the model use overall?","blog-visuals.ai-explainability-lens.copy.local":"local","blog-visuals.ai-explainability-lens.copy.local.detail":"Why did this exact input receive this decision?","blog-visuals.ai-explainability-lens.copy.opacity":"opacity","blog-visuals.ai-explainability-lens.copy.opacity.detail":"More inspection should reduce opacity, not decorate it.","blog-visuals.ai-explainability-lens.eyebrow":"Explainability lens","blog-visuals.ai-explainability-lens.mode.causal":"causal","blog-visuals.ai-explainability-lens.mode.counter":"counterfactual","blog-visuals.ai-explainability-lens.mode.global":"global","blog-visuals.ai-explainability-lens.mode.local":"local","blog-visuals.ai-explainability-lens.range":"inspection depth","blog-visuals.ai-explainability-lens.title":"A score needs a reason someone can inspect","blog-visuals.ai-hallucination-pressure.copy.ambiguity":"ambiguity","blog-visuals.ai-hallucination-pressure.copy.ambiguity.detail":"The slider raises the pressure to invent plausible filler.","blog-visuals.ai-hallucination-pressure.copy.constraint":"constraint","blog-visuals.ai-hallucination-pressure.copy.constraint.detail":"Schemas and rules make unsupported facts harder to smuggle through.","blog-visuals.ai-hallucination-pressure.copy.context":"context","blog-visuals.ai-hallucination-pressure.copy.context.detail":"Thin context leaves the model to fill gaps from pattern memory.","blog-visuals.ai-hallucination-pressure.copy.control":"control","blog-visuals.ai-hallucination-pressure.copy.control.detail":"The selected control changes where the answer is constrained.","blog-visuals.ai-hallucination-pressure.copy.falseClaims":"false claims","blog-visuals.ai-hallucination-pressure.copy.falseClaims.detail":"Grounding and checks reduce the claims that survive.","blog-visuals.ai-hallucination-pressure.copy.review":"review","blog-visuals.ai-hallucination-pressure.copy.review.detail":"Human or verifier review catches confident nonsense before deployment.","blog-visuals.ai-hallucination-pressure.copy.source":"source","blog-visuals.ai-hallucination-pressure.copy.source.detail":"Grounded source material narrows what the answer is allowed to claim.","blog-visuals.ai-hallucination-pressure.eyebrow":"Hallucination pressure","blog-visuals.ai-hallucination-pressure.mode.constraint":"constraint","blog-visuals.ai-hallucination-pressure.mode.context":"context","blog-visuals.ai-hallucination-pressure.mode.review":"review","blog-visuals.ai-hallucination-pressure.mode.source":"source","blog-visuals.ai-hallucination-pressure.range":"ambiguity","blog-visuals.ai-hallucination-pressure.title":"Confident nonsense grows in the gaps","blog-visuals.ai-inference-path.copy.cache":"KV cache","blog-visuals.ai-inference-path.copy.cache.detail":"Cached keys and values prevent repeated work during generation.","blog-visuals.ai-inference-path.copy.cacheHit":"cache hit","blog-visuals.ai-inference-path.copy.cacheHit.detail":"Inference economics improve when repeated work is avoided.","blog-visuals.ai-inference-path.copy.latency":"latency","blog-visuals.ai-inference-path.copy.latency.detail":"Deployment optimises response time, cost and reliability.","blog-visuals.ai-inference-path.copy.model":"model","blog-visuals.ai-inference-path.copy.model.detail":"The model computes the next-token distribution.","blog-visuals.ai-inference-path.copy.output":"output","blog-visuals.ai-inference-path.copy.output.detail":"Tokens stream back and trigger any post-processing.","blog-visuals.ai-inference-path.copy.prompt":"prompt","blog-visuals.ai-inference-path.copy.prompt.detail":"The request is tokenised and shaped for the model.","blog-visuals.ai-inference-path.copy.requests":"requests","blog-visuals.ai-inference-path.copy.requests.detail":"The slider increases concurrent request pressure.","blog-visuals.ai-inference-path.eyebrow":"Inference path","blog-visuals.ai-inference-path.mode.cache":"cache","blog-visuals.ai-inference-path.mode.model":"model","blog-visuals.ai-inference-path.mode.output":"output","blog-visuals.ai-inference-path.mode.prompt":"prompt","blog-visuals.ai-inference-path.range":"request pressure","blog-visuals.ai-inference-path.title":"Serving AI is a latency and cost problem","blog-visuals.ai-invisibility-shift.copy.ambient":"ambient","blog-visuals.ai-invisibility-shift.copy.ambient.detail":"The interface fades because the right action appears where the work already happens.","blog-visuals.ai-invisibility-shift.copy.chat":"chat","blog-visuals.ai-invisibility-shift.copy.chat.detail":"The user still has to visit a separate box and ask explicitly.","blog-visuals.ai-invisibility-shift.copy.context":"context coverage","blog-visuals.ai-invisibility-shift.copy.context.detail":"The slider shows how much work context the AI can see before it has to ask.","blog-visuals.ai-invisibility-shift.copy.copilot":"copilot","blog-visuals.ai-invisibility-shift.copy.copilot.detail":"AI sits beside the work, useful but still visibly separate.","blog-visuals.ai-invisibility-shift.copy.phase":"phase","blog-visuals.ai-invisibility-shift.copy.phase.detail":"The prediction is less about smarter chat and more about less visible interface.","blog-visuals.ai-invisibility-shift.copy.workflow":"workflow","blog-visuals.ai-invisibility-shift.copy.workflow.detail":"The model starts inside the process and sees enough context to help without ceremony.","blog-visuals.ai-invisibility-shift.eyebrow":"Interface shift","blog-visuals.ai-invisibility-shift.mode.ambient":"ambient","blog-visuals.ai-invisibility-shift.mode.context":"context","blog-visuals.ai-invisibility-shift.mode.interface":"interface","blog-visuals.ai-invisibility-shift.range":"context maturity","blog-visuals.ai-invisibility-shift.title":"The best AI interface may disappear","blog-visuals.ai-memory-window.copy.faded.detail":"This token is outside direct attention unless retrieved.","blog-visuals.ai-memory-window.copy.store":"retrieval store","blog-visuals.ai-memory-window.copy.store.detail":"External memory works only when the right record is retrieved back into context.","blog-visuals.ai-memory-window.copy.tokens":"tokens","blog-visuals.ai-memory-window.copy.tokens.detail":"The slider changes how much conversation the model must fit into the window.","blog-visuals.ai-memory-window.copy.visible":"visible","blog-visuals.ai-memory-window.copy.visible.detail":"This token is inside the current context window.","blog-visuals.ai-memory-window.copy.visible.metric":"Only part of the conversation is directly available at once.","blog-visuals.ai-memory-window.eyebrow":"Memory window","blog-visuals.ai-memory-window.mode.context":"context","blog-visuals.ai-memory-window.mode.forget":"forget","blog-visuals.ai-memory-window.mode.retrieve":"retrieve","blog-visuals.ai-memory-window.mode.session":"session","blog-visuals.ai-memory-window.range":"conversation length","blog-visuals.ai-memory-window.title":"Remembering means being inside the working window","blog-visuals.ai-neural-switchboard.copy.layers":"layers","blog-visuals.ai-neural-switchboard.copy.layers.detail":"The slider adds layers to the switchboard.","blog-visuals.ai-neural-switchboard.copy.node.detail":"A node combines inputs, weights and an activation rule.","blog-visuals.ai-neural-switchboard.copy.view":"view","blog-visuals.ai-neural-switchboard.copy.view.detail":"The mode changes whether weights, layers, activation or binary decisions are highlighted.","blog-visuals.ai-neural-switchboard.eyebrow":"Neural switchboard","blog-visuals.ai-neural-switchboard.mode.activation":"activation","blog-visuals.ai-neural-switchboard.mode.binary":"binary","blog-visuals.ai-neural-switchboard.mode.layers":"layers","blog-visuals.ai-neural-switchboard.mode.weights":"weights","blog-visuals.ai-neural-switchboard.range":"network depth","blog-visuals.ai-neural-switchboard.title":"A neural network is weighted routing at scale","blog-visuals.ai-pattern-logic-split.copy.checkable":"checkable","blog-visuals.ai-pattern-logic-split.copy.conclusion":"conclusion","blog-visuals.ai-pattern-logic-split.copy.conclusion.detail":"The answer can be checked against the premises.","blog-visuals.ai-pattern-logic-split.copy.confidence":"confidence","blog-visuals.ai-pattern-logic-split.copy.confidence.detail":"The slider raises confidence pressure; verification decides whether confidence is enough.","blog-visuals.ai-pattern-logic-split.copy.data":"training data","blog-visuals.ai-pattern-logic-split.copy.data.detail":"Examples contain correlations the model can learn.","blog-visuals.ai-pattern-logic-split.copy.logicLane":"logic lane","blog-visuals.ai-pattern-logic-split.copy.mode":"mode","blog-visuals.ai-pattern-logic-split.copy.mode.detail":"Pattern matching, logic and verification answer different questions.","blog-visuals.ai-pattern-logic-split.copy.patternLane":"pattern lane","blog-visuals.ai-pattern-logic-split.copy.predict":"prediction","blog-visuals.ai-pattern-logic-split.copy.predict.detail":"The answer is likely, but the reason is not explicit.","blog-visuals.ai-pattern-logic-split.copy.premise":"premise","blog-visuals.ai-pattern-logic-split.copy.premise.detail":"Known facts are named before the conclusion.","blog-visuals.ai-pattern-logic-split.copy.rule":"rule","blog-visuals.ai-pattern-logic-split.copy.rule.detail":"A rule explains why the conclusion follows.","blog-visuals.ai-pattern-logic-split.copy.weights":"weights","blog-visuals.ai-pattern-logic-split.copy.weights.detail":"Association is stored as statistical pressure in weights.","blog-visuals.ai-pattern-logic-split.eyebrow":"Pattern and logic","blog-visuals.ai-pattern-logic-split.mode.logic":"logic","blog-visuals.ai-pattern-logic-split.mode.pattern":"pattern","blog-visuals.ai-pattern-logic-split.mode.verify":"verify","blog-visuals.ai-pattern-logic-split.range":"confidence pressure","blog-visuals.ai-pattern-logic-split.title":"Right answers can come from different machinery","blog-visuals.ai-reasoning-stack.copy.hybrid":"hybrid plan","blog-visuals.ai-reasoning-stack.copy.hybrid.detail":"A useful system often combines model fluency with mechanical checks.","blog-visuals.ai-reasoning-stack.copy.layer":"layer","blog-visuals.ai-reasoning-stack.copy.layer.detail":"The mode selects the reasoning layer under inspection.","blog-visuals.ai-reasoning-stack.copy.pattern":"pattern model","blog-visuals.ai-reasoning-stack.copy.pattern.detail":"Neural pattern matching proposes a likely answer.","blog-visuals.ai-reasoning-stack.copy.symbolic":"symbolic rules","blog-visuals.ai-reasoning-stack.copy.symbolic.detail":"Explicit rules can reject answers that violate known constraints.","blog-visuals.ai-reasoning-stack.copy.uncertainty":"uncertainty","blog-visuals.ai-reasoning-stack.copy.uncertainty.detail":"The slider shows uncertainty falling as checks accumulate.","blog-visuals.ai-reasoning-stack.copy.verify":"verify","blog-visuals.ai-reasoning-stack.copy.verify.detail":"Evidence makes the reasoning inspectable after the answer is produced.","blog-visuals.ai-reasoning-stack.eyebrow":"Reasoning stack","blog-visuals.ai-reasoning-stack.mode.hybrid":"hybrid","blog-visuals.ai-reasoning-stack.mode.pattern":"pattern","blog-visuals.ai-reasoning-stack.mode.symbolic":"symbolic","blog-visuals.ai-reasoning-stack.mode.verify":"verify","blog-visuals.ai-reasoning-stack.range":"checking depth","blog-visuals.ai-reasoning-stack.title":"Reasoning gets stronger when claims are checked","blog-visuals.ai-training-loop.copy.data":"data","blog-visuals.ai-training-loop.copy.data.detail":"Examples define what the model gets to learn from.","blog-visuals.ai-training-loop.copy.epoch":"epoch","blog-visuals.ai-training-loop.copy.epoch.detail":"The slider advances training time.","blog-visuals.ai-training-loop.copy.loss":"loss","blog-visuals.ai-training-loop.copy.loss.detail":"The loss function measures the current mistake.","blog-visuals.ai-training-loop.copy.lossMetric":"loss","blog-visuals.ai-training-loop.copy.lossMetric.detail":"Loss should fall, but validation decides whether learning is useful.","blog-visuals.ai-training-loop.copy.update":"update","blog-visuals.ai-training-loop.copy.update.detail":"Backpropagation changes weights to reduce the loss.","blog-visuals.ai-training-loop.copy.validate":"validate","blog-visuals.ai-training-loop.copy.validate.detail":"Validation catches memorisation before it ships.","blog-visuals.ai-training-loop.eyebrow":"Training loop","blog-visuals.ai-training-loop.mode.data":"data","blog-visuals.ai-training-loop.mode.loss":"loss","blog-visuals.ai-training-loop.mode.update":"update","blog-visuals.ai-training-loop.mode.validate":"validate","blog-visuals.ai-training-loop.range":"training progress","blog-visuals.ai-training-loop.title":"Learning is repeated correction under measurement","blog-visuals.aion-audit-table.eyebrow":"AION audit room","blog-visuals.aion-audit-table.mode.accept":"accept","blog-visuals.aion-audit-table.mode.hold":"hold","blog-visuals.aion-audit-table.mode.sample":"sample","blog-visuals.aion-audit-table.range":"audit pressure","blog-visuals.aion-audit-table.row.care":"care","blog-visuals.aion-audit-table.row.credit":"credit","blog-visuals.aion-audit-table.row.route":"routing","blog-visuals.aion-audit-table.title":"The central object is the check","blog-visuals.aion-certificate-anatomy.eyebrow":"AION certificate anatomy","blog-visuals.aion-certificate-anatomy.mode.formats":"formats","blog-visuals.aion-certificate-anatomy.mode.hash":"hash","blog-visuals.aion-certificate-anatomy.mode.steps":"steps","blog-visuals.aion-certificate-anatomy.range":"inspection focus","blog-visuals.aion-certificate-anatomy.row.format":"format","blog-visuals.aion-certificate-anatomy.row.hash":"hash","blog-visuals.aion-certificate-anatomy.row.premises":"premises","blog-visuals.aion-certificate-anatomy.row.result":"result","blog-visuals.aion-certificate-anatomy.row.solver":"solver","blog-visuals.aion-certificate-anatomy.title":"What has to travel with the decision","blog-visuals.aion-cli-audit-lanes.eyebrow":"AION operating surface","blog-visuals.aion-cli-audit-lanes.mode.archive":"archive","blog-visuals.aion-cli-audit-lanes.mode.audit":"audit","blog-visuals.aion-cli-audit-lanes.mode.ci":"CI","blog-visuals.aion-cli-audit-lanes.range":"run progress","blog-visuals.aion-cli-audit-lanes.title":"One certificate, several repeatable lanes","blog-visuals.aion-proof-chain.eyebrow":"AION evidence route","blog-visuals.aion-proof-chain.mode.answer":"answer","blog-visuals.aion-proof-chain.mode.proof":"proof","blog-visuals.aion-proof-chain.mode.replay":"replay","blog-visuals.aion-proof-chain.range":"proof density","blog-visuals.aion-proof-chain.title":"From answer to replayable proof","blog-visuals.algorithm-backend-matrix.copy.boolean":"boolean","blog-visuals.algorithm-backend-matrix.copy.boolean.detail":"Discrete search, gates and constraints stay first-class.","blog-visuals.algorithm-backend-matrix.copy.coverage":"algorithms","blog-visuals.algorithm-backend-matrix.copy.coverage.detail":"The slider moves from a partial toolkit toward complete algorithm coverage.","blog-visuals.algorithm-backend-matrix.copy.cpu":"CPU","blog-visuals.algorithm-backend-matrix.copy.edge":"edge","blog-visuals.algorithm-backend-matrix.copy.fpga":"FPGA","blog-visuals.algorithm-backend-matrix.copy.gpu":"GPU","blog-visuals.algorithm-backend-matrix.copy.graph":"graph","blog-visuals.algorithm-backend-matrix.copy.graph.detail":"Routes, dependency walks and scheduling share the same surface.","blog-visuals.algorithm-backend-matrix.copy.numeric":"numeric","blog-visuals.algorithm-backend-matrix.copy.numeric.detail":"Numeric kernels must keep behaviour comparable across targets.","blog-visuals.algorithm-backend-matrix.copy.optimise":"optimise","blog-visuals.algorithm-backend-matrix.copy.optimise.detail":"Search and optimisation do not need a separate product stack.","blog-visuals.algorithm-backend-matrix.copy.ready":"ready","blog-visuals.algorithm-backend-matrix.copy.view":"view","blog-visuals.algorithm-backend-matrix.copy.view.detail":"The mode changes which dimension of fragmentation is being inspected.","blog-visuals.algorithm-backend-matrix.copy.wasm":"WASM","blog-visuals.algorithm-backend-matrix.eyebrow":"Algorithm matrix","blog-visuals.algorithm-backend-matrix.mode.backend":"backend","blog-visuals.algorithm-backend-matrix.mode.bitwidth":"bit width","blog-visuals.algorithm-backend-matrix.mode.deploy":"deploy","blog-visuals.algorithm-backend-matrix.mode.matrix":"matrix","blog-visuals.algorithm-backend-matrix.range":"coverage","blog-visuals.algorithm-backend-matrix.title":"A complete stack removes framework roulette","blog-visuals.axiom-arithmetic-stepper.eyebrow":"AXIOM arithmetic","blog-visuals.axiom-arithmetic-stepper.mode.add":"add","blog-visuals.axiom-arithmetic-stepper.mode.mul":"multiply","blog-visuals.axiom-arithmetic-stepper.mode.normalize":"normalize","blog-visuals.axiom-arithmetic-stepper.range":"operand spread","blog-visuals.axiom-arithmetic-stepper.title":"Align, operate, normalize back into shape","blog-visuals.axiom-lattice.eyebrow":"AXIOM lattice","blog-visuals.axiom-lattice.mode.fine":"fine","blog-visuals.axiom-lattice.mode.fixed":"fixed","blog-visuals.axiom-lattice.mode.nn":"NN","blog-visuals.axiom-lattice.mode.standard":"standard","blog-visuals.axiom-lattice.range":"value magnitude","blog-visuals.axiom-lattice.title":"Fixed spacing is not the only honest grid","blog-visuals.axiom-packed-value.eyebrow":"AXIOM packed value","blog-visuals.axiom-packed-value.mode.dense":"dense","blog-visuals.axiom-packed-value.mode.fine":"fine","blog-visuals.axiom-packed-value.mode.nn":"NN","blog-visuals.axiom-packed-value.mode.standard":"standard","blog-visuals.axiom-packed-value.range":"mantissa pressure","blog-visuals.axiom-packed-value.title":"One u32 carries sign, exponent choice, and mantissa","blog-visuals.axiom-strategy-lab.eyebrow":"AXIOM strategy lab","blog-visuals.axiom-strategy-lab.mode.apot":"APoT","blog-visuals.axiom-strategy-lab.mode.concrete":"concrete","blog-visuals.axiom-strategy-lab.mode.flex":"flex","blog-visuals.axiom-strategy-lab.mode.tensor":"tensor","blog-visuals.axiom-strategy-lab.range":"batch pressure","blog-visuals.axiom-strategy-lab.title":"The representation has several compute paths","blog-visuals.binary-energy-alternative.copy.binary":"binary","blog-visuals.binary-energy-alternative.copy.binary.detail":"Binary operations move more work onto standard, cooler hardware.","blog-visuals.binary-energy-alternative.copy.exotic":"exotic stack","blog-visuals.binary-energy-alternative.copy.exotic.detail":"GPU-heavy systems pull specialised cooling and procurement into the model decision.","blog-visuals.binary-energy-alternative.copy.float":"floating point","blog-visuals.binary-energy-alternative.copy.float.detail":"Complex floating-point operations drag power, cooling and supply-chain assumptions with them.","blog-visuals.binary-energy-alternative.copy.less":"less energy","blog-visuals.binary-energy-alternative.copy.less.detail":"The slider changes workload pressure while preserving the energy gap.","blog-visuals.binary-energy-alternative.copy.standard":"standard hardware","blog-visuals.binary-energy-alternative.copy.standard.detail":"Efficiency matters because it changes where AI can run, not only what the power bill says.","blog-visuals.binary-energy-alternative.eyebrow":"Binary alternative","blog-visuals.binary-energy-alternative.mode.binary":"binary","blog-visuals.binary-energy-alternative.mode.europe":"European hardware","blog-visuals.binary-energy-alternative.mode.float":"floating point","blog-visuals.binary-energy-alternative.range":"workload pressure","blog-visuals.binary-energy-alternative.title":"Efficiency changes the deployment map","blog-visuals.bindery-detect-pipeline.eyebrow":"Bindery document pipeline","blog-visuals.bindery-detect-pipeline.mode.detect":"detect","blog-visuals.bindery-detect-pipeline.mode.parse":"parse","blog-visuals.bindery-detect-pipeline.mode.query":"query","blog-visuals.bindery-detect-pipeline.mode.write":"write","blog-visuals.bindery-detect-pipeline.range":"corpus mess","blog-visuals.bindery-detect-pipeline.title":"Extensions lie, structure tells the truth","blog-visuals.bindery-docql-flow.eyebrow":"Bindery DocQL flow","blog-visuals.bindery-docql-flow.mode.formulas":"formulas","blog-visuals.bindery-docql-flow.mode.refs":"references","blog-visuals.bindery-docql-flow.mode.shapes":"shapes","blog-visuals.bindery-docql-flow.mode.tables":"tables","blog-visuals.bindery-docql-flow.range":"query scope","blog-visuals.bindery-docql-flow.title":"Queries should survive the file format","blog-visuals.bindery-format-matrix.eyebrow":"Bindery format matrix","blog-visuals.bindery-format-matrix.mode.query":"query","blog-visuals.bindery-format-matrix.mode.read":"read","blog-visuals.bindery-format-matrix.mode.roundtrip":"round trip","blog-visuals.bindery-format-matrix.mode.write":"write","blog-visuals.bindery-format-matrix.range":"capability strictness","blog-visuals.bindery-format-matrix.title":"One library still names its promises","blog-visuals.bindery-integration-stack.eyebrow":"Bindery integration stack","blog-visuals.bindery-integration-stack.mode.docql":"DocQL","blog-visuals.bindery-integration-stack.mode.python":"Python","blog-visuals.bindery-integration-stack.mode.rust":"Rust","blog-visuals.bindery-integration-stack.mode.stack":"stack","blog-visuals.bindery-integration-stack.range":"pipeline depth","blog-visuals.bindery-integration-stack.title":"The engine stays shared across entry points","blog-visuals.bitweave-binary-vectors.eyebrow":"BitWeave binary vectors","blog-visuals.bitweave-binary-vectors.mode.bits":"bits","blog-visuals.bitweave-binary-vectors.mode.distance":"distance","blog-visuals.bitweave-binary-vectors.mode.popcnt":"popcount","blog-visuals.bitweave-binary-vectors.range":"visible bits","blog-visuals.bitweave-binary-vectors.title":"Similarity can be CPU-shaped","blog-visuals.bitweave-local-retrieval.eyebrow":"BitWeave local retrieval","blog-visuals.bitweave-local-retrieval.mode.answer":"answer","blog-visuals.bitweave-local-retrieval.mode.index":"index","blog-visuals.bitweave-local-retrieval.mode.local":"local","blog-visuals.bitweave-local-retrieval.range":"locality","blog-visuals.bitweave-local-retrieval.title":"The corpus does not need a holiday abroad","blog-visuals.bitweave-ranking.eyebrow":"BitWeave ranking","blog-visuals.bitweave-ranking.mode.repeat":"repeat","blog-visuals.bitweave-ranking.mode.scores":"scores","blog-visuals.bitweave-ranking.mode.ties":"ties","blog-visuals.bitweave-ranking.range":"tie pressure","blog-visuals.bitweave-ranking.title":"A close score still needs an order","blog-visuals.bitweave-recall-audit.eyebrow":"BitWeave recall audit","blog-visuals.bitweave-recall-audit.mode.query":"query","blog-visuals.bitweave-recall-audit.mode.rank":"rank","blog-visuals.bitweave-recall-audit.mode.receipt":"receipt","blog-visuals.bitweave-recall-audit.range":"inspection path","blog-visuals.bitweave-recall-audit.title":"Retrieval needs a receipt too","blog-visuals.cloud-cost-cliff.copy.cliff":"cliff","blog-visuals.cloud-cost-cliff.copy.cloudCost":"cloud cost","blog-visuals.cloud-cost-cliff.copy.demo":"demo","blog-visuals.cloud-cost-cliff.copy.demoZone":"demo zone","blog-visuals.cloud-cost-cliff.copy.demoZone.detail":"Low usage hides inference cost during demos and pilots.","blog-visuals.cloud-cost-cliff.copy.edgeCost":"edge cost","blog-visuals.cloud-cost-cliff.copy.margin":"margin","blog-visuals.cloud-cost-cliff.copy.margin.detail":"The selected mode shows why success can destroy cloud AI margins.","blog-visuals.cloud-cost-cliff.copy.revenue":"revenue","blog-visuals.cloud-cost-cliff.copy.scaleZone":"scale zone","blog-visuals.cloud-cost-cliff.copy.scaleZone.detail":"At scale, every interaction keeps a meter running.","blog-visuals.cloud-cost-cliff.copy.users":"users","blog-visuals.cloud-cost-cliff.copy.users.detail":"The slider moves from demo traffic to real adoption.","blog-visuals.cloud-cost-cliff.eyebrow":"Cost cliff","blog-visuals.cloud-cost-cliff.mode.demo":"demo","blog-visuals.cloud-cost-cliff.mode.edge":"edge","blog-visuals.cloud-cost-cliff.mode.scale":"scale","blog-visuals.cloud-cost-cliff.range":"adoption","blog-visuals.cloud-cost-cliff.title":"Success can make the unit economics worse","blog-visuals.cloud-saas-margin.copy.aiHigh":"cloud AI high","blog-visuals.cloud-saas-margin.copy.aiHigh.detail":"Heavy interaction turns revenue into GPU rent.","blog-visuals.cloud-saas-margin.copy.aiLow":"cloud AI low","blog-visuals.cloud-saas-margin.copy.aiLow.detail":"A light inference workload can still erode margin badly.","blog-visuals.cloud-saas-margin.copy.costUser":"cost/user","blog-visuals.cloud-saas-margin.copy.costUser.detail":"The slider raises monthly inference cost per user.","blog-visuals.cloud-saas-margin.copy.model":"model","blog-visuals.cloud-saas-margin.copy.model.detail":"The mode compares ordinary software leverage with metered AI inference.","blog-visuals.cloud-saas-margin.copy.operatingLeverage":"operating leverage","blog-visuals.cloud-saas-margin.copy.operatingLeverage.detail":"Cloud AI breaks the old SaaS miracle because success keeps adding variable cost.","blog-visuals.cloud-saas-margin.copy.saas":"traditional SaaS","blog-visuals.cloud-saas-margin.copy.saas.detail":"Software margins work because marginal user cost approaches zero.","blog-visuals.cloud-saas-margin.eyebrow":"SaaS margin","blog-visuals.cloud-saas-margin.mode.ai":"cloud AI","blog-visuals.cloud-saas-margin.mode.margin":"margin","blog-visuals.cloud-saas-margin.mode.saas":"SaaS","blog-visuals.cloud-saas-margin.range":"inference cost","blog-visuals.cloud-saas-margin.title":"Metered inference breaks operating leverage","blog-visuals.copy.001":"answer","blog-visuals.copy.002":"The answer is useful, but by itself it is not evidence yet.","blog-visuals.copy.003":"model output","blog-visuals.copy.004":"certificate","blog-visuals.copy.005":"Premises, steps and hashes travel with the decision.","blog-visuals.copy.006":"A recorded step in the proof route.","blog-visuals.copy.007":"verifier","blog-visuals.copy.008":"A smaller independent path checks the artefact offline.","blog-visuals.copy.009":"offline check","blog-visuals.copy.010":"steps","blog-visuals.copy.011":"The density slider changes how many proof steps travel with the answer.","blog-visuals.copy.012":"check","blog-visuals.copy.013":"More proof gives the verifier more work, not a trust ritual.","blog-visuals.copy.014":"proof object","blog-visuals.copy.015":"The certificate is a structured object, not a nice explanation.","blog-visuals.copy.016":"A format another checker can read.","blog-visuals.copy.017":"check","blog-visuals.copy.018":"The verifier does not need to trust the whole solver.","blog-visuals.copy.019":"The same certificates must stay useful in every workflow.","blog-visuals.copy.020":"Checkable result without a sales ritual.","blog-visuals.copy.021":"audit matrix","blog-visuals.copy.022":"A decision gets calmer when the evidence route can be walked again.","blog-visuals.copy.023":"Each cell points to an object that can be checked again.","blog-visuals.copy.024":"audit pressure","blog-visuals.copy.025":"outcome","blog-visuals.copy.026":"The mode changes whether the matrix samples, holds, or accepts.","blog-visuals.copy.027":"EU label","blog-visuals.copy.028":"A label can be true and still say little about operational power.","blog-visuals.copy.029":"legal text","blog-visuals.copy.030":"control board","blog-visuals.copy.031":"Control lives in contracts, compute, keys, operators and exit.","blog-visuals.copy.032":"At high dependency pressure this becomes an operational risk.","blog-visuals.copy.033":"Real sovereignty depends on this door.","blog-visuals.copy.034":"lock-in","blog-visuals.copy.035":"The slider closes the doors and shows where room to move disappears.","blog-visuals.copy.036":"A layer in the control stack. The top claim rests on everything below.","blog-visuals.copy.037":"who?","blog-visuals.copy.038":"The useful question is who can actually change the layer.","blog-visuals.copy.039":"An investigation step with source trace, not a loose opinion.","blog-visuals.copy.040":"evidence","blog-visuals.copy.041":"The route makes the claim checkable.","blog-visuals.copy.042":"Literacy is a stair: each level improves the next question.","blog-visuals.copy.043":"reached","blog-visuals.copy.044":"Room pressure determines how many layers are usable already.","blog-visuals.copy.045":"A chapter block that works best in relation to the rest.","blog-visuals.copy.046":"shared language","blog-visuals.copy.047":"The series avoids repairing vocabulary at the start of every meeting.","blog-visuals.copy.048":"intuition","blog-visuals.copy.049":"Examples reveal the shape before notation arrives.","blog-visuals.copy.050":"measurement","blog-visuals.copy.051":"Measurement makes claims comparable and governable.","blog-visuals.copy.052":"Probability is not decoration; it is how uncertainty becomes readable.","blog-visuals.copy.053":"model","blog-visuals.copy.054":"The box gets less mystical once you make it testable.","blog-visuals.copy.055":"A teaching step: claim, boundary, record or revision.","blog-visuals.copy.056":"uncertain","blog-visuals.copy.057":"More uncertainty makes testing and revision heavier in the loop.","blog-visuals.copy.058":"Part of the source package behind the traffic safety report.","blog-visuals.copy.059":"method + gaps","blog-visuals.copy.060":"The uncertainty travels with the claim instead of being hidden.","blog-visuals.copy.061":"bn passenger-km","blog-visuals.copy.062":"deaths per bn km","blog-visuals.copy.063":"deaths","blog-visuals.copy.064":"policy on","blog-visuals.copy.065":"The slider moves policy attention and shows the effect under the chosen denominator.","blog-visuals.copy.066":"An actor with its own powers. The system fails when levers are assigned to the wrong place.","blog-visuals.copy.067":"levers","blog-visuals.copy.068":"Speed, design, exams, care and data are different controls.","blog-visuals.copy.069":"coordination","blog-visuals.copy.070":"More coordination strengthens the selected actor route.","blog-visuals.copy.071":"A step in the feedback loop that should make road safety learn.","blog-visuals.copy.072":"scenario","blog-visuals.copy.073":"The slider changes system learning in the projection line.","blog-visuals.copy.074":"source","blog-visuals.copy.075":"Byte positions stay part of the path.","blog-visuals.copy.076":"Sections store strings, types, nodes and exports.","blog-visuals.copy.077":"graph hash","blog-visuals.copy.078":"Graph size changes the hash compared later.","blog-visuals.copy.079":"source","blog-visuals.copy.080":"Byte positions and trivia stay part of the checkable path.","blog-visuals.copy.081":"The receipt is a section table, not a picture of a graph.","blog-visuals.copy.082":"graph hash","blog-visuals.copy.083":"The slider changes node count and therefore the hash compared later.","blog-visuals.copy.084":"Pratt events remain recoverable for tooling.","blog-visuals.copy.085":"Canonical type hashes must match on every edge.","blog-visuals.copy.086":"Emit only reaches the backend after invariant validation.","blog-visuals.copy.087":"diagnostic: hold","blog-visuals.copy.088":"invariant pressure","blog-visuals.copy.089":"More pressure makes the validator less forgiving.","blog-visuals.copy.090":"One canonical graph is the meaning. The backend is execution.","blog-visuals.copy.091":"artifact","blog-visuals.copy.092":"The artifact changes per target; the graph hash does not.","blog-visuals.copy.093":"semantics","blog-visuals.copy.094":"High target pressure blocks instead of silently changing meaning.","blog-visuals.copy.095":"The graph is constant; only the run environment changes.","blog-visuals.copy.096":"Comparison accepts only identical output hashes.","blog-visuals.copy.097":"outcome","blog-visuals.copy.098":"Drift pressure shows when a fast path is no longer allowed to ship.","blog-visuals.copy.099":"org A","blog-visuals.copy.100":"Own data stays inside the own boundary.","blog-visuals.copy.101":"org B","blog-visuals.copy.102":"Cooperate without centralising everything.","blog-visuals.copy.103":"Work package with policy.","blog-visuals.copy.104":"data gravity","blog-visuals.copy.105":"The slider pulls work toward where the data is supposed to stay.","blog-visuals.copy.106":"schedule contract","blog-visuals.copy.107":"Region, privacy, cost and node requirements become explicit.","blog-visuals.copy.108":"A hard placement condition.","blog-visuals.copy.109":"Run only after placement checks.","blog-visuals.copy.110":"The shared model receives updates, not raw records.","blog-visuals.copy.111":"Local update without a central data dump.","blog-visuals.copy.112":"nodes","blog-visuals.copy.113":"The federation slider changes how many parties participate.","blog-visuals.copy.114":"exit pack","blog-visuals.copy.115":"Models, logs, keys and contracts need to be transferable.","blog-visuals.copy.116":"new home","blog-visuals.copy.117":"Exit is credible when the package restarts somewhere else.","blog-visuals.copy.118":"Part of the moving box.","blog-visuals.copy.119":"status","blog-visuals.copy.120":"Exit readiness determines whether the package really restarts elsewhere.","blog-visuals.copy.121":"A cell with type, state and rules.","blog-visuals.copy.122":"The schema makes the world machine-readable, not merely visible.","blog-visuals.copy.123":"world cells","blog-visuals.copy.124":"The detail slider changes the world size, not a loose indicator.","blog-visuals.copy.125":"Pixels show appearance, but miss rules and state.","blog-visuals.copy.126":"A world object that is inspectable and replayable.","blog-visuals.copy.127":"An explicit object in world state.","blog-visuals.copy.128":"objects","blog-visuals.copy.129":"State richness changes how many objects are inspectable.","blog-visuals.copy.130":"A trace point from the physics replay.","blog-visuals.copy.131":"The scrubber chooses the replay moment.","blog-visuals.copy.132":"replay time","blog-visuals.copy.133":"The timeline slider scrubs the physical path itself.","blog-visuals.copy.134":"A request is not the same as permission.","blog-visuals.copy.135":"Type, physics and safety rules form the cage.","blog-visuals.copy.136":"A check before the world changes.","blog-visuals.copy.137":"Only valid changes reach the world.","blog-visuals.copy.138":"outcome","blog-visuals.copy.139":"Rule strictness changes which world change gets through the cage.","blog-visuals.copy.140":"money","blog-visuals.copy.141":"models","blog-visuals.copy.142":"law","blog-visuals.copy.143":"skills","blog-visuals.copy.144":"law","blog-visuals.copy.145":"contracts","blog-visuals.copy.146":"filings","blog-visuals.copy.147":"owners","blog-visuals.copy.148":"controls","blog-visuals.copy.149":"words","blog-visuals.copy.150":"math","blog-visuals.copy.151":"models","blog-visuals.copy.152":"limits","blog-visuals.copy.153":"judgement","blog-visuals.copy.154":"foundations","blog-visuals.copy.155":"math","blog-visuals.copy.156":"learning","blog-visuals.copy.157":"ethics","blog-visuals.copy.158":"vectors","blog-visuals.copy.159":"probability","blog-visuals.copy.160":"optimisation","blog-visuals.copy.161":"ask","blog-visuals.copy.162":"record","blog-visuals.copy.163":"revise","blog-visuals.copy.164":"sources","blog-visuals.copy.165":"files","blog-visuals.copy.166":"tables","blog-visuals.copy.167":"car","blog-visuals.copy.168":"bike","blog-visuals.copy.169":"walk","blog-visuals.copy.170":"public","blog-visuals.copy.171":"car","blog-visuals.copy.172":"bike","blog-visuals.copy.173":"walk","blog-visuals.copy.174":"public","blog-visuals.copy.175":"car","blog-visuals.copy.176":"bike","blog-visuals.copy.177":"walk","blog-visuals.copy.178":"public","blog-visuals.copy.179":"ministry","blog-visuals.copy.180":"roads","blog-visuals.copy.181":"care","blog-visuals.copy.182":"police","blog-visuals.copy.183":"observe","blog-visuals.copy.184":"decide","blog-visuals.copy.185":"build","blog-visuals.copy.186":"measure","blog-visuals.copy.187":"models","blog-visuals.copy.188":"keys","blog-visuals.copy.189":"image","blog-visuals.copy.190":"scene","blog-visuals.copy.191":"rules","blog-visuals.copy.192":"physics","blog-visuals.copy.193":"checks","blog-visuals.copy.194":"world","blog-visuals.copy.195":"A dependency lock.","blog-visuals.copy.196":"Names the exact artefact.","blog-visuals.copy.197":"Inputs and assumptions are explicit.","blog-visuals.copy.198":"The route is recorded, not narrated later.","blog-visuals.copy.199":"Standard proof formats keep exit open.","blog-visuals.copy.200":"The checker accepts or rejects.","blog-visuals.copy.201":"archive","blog-visuals.copy.202":"ci run","blog-visuals.copy.203":"audit","blog-visuals.copy.204":"application log","blog-visuals.copy.205":"A free text log can describe activity without preserving order, schema, or integrity. Useful, but not enough.","blog-visuals.copy.206":"grep finds lines","blog-visuals.copy.207":"ledger event","blog-visuals.copy.208":"Typed payload, content hash, previous hash, session and optional signature travel as the record.","blog-visuals.copy.209":"event type","blog-visuals.copy.210":"content hash","blog-visuals.copy.211":"previous hash","blog-visuals.copy.212":"session","blog-visuals.copy.213":"audit doubt","blog-visuals.copy.214":"The slider raises audit pressure and shows where ordinary logs stop carrying enough structure.","blog-visuals.copy.215":"record shape","blog-visuals.copy.216":"The ledger shape is meant to be replayed, not only searched.","blog-visuals.copy.217":"append","blog-visuals.copy.218":"hash","blog-visuals.copy.219":"anchor","blog-visuals.copy.220":"persist","blog-visuals.copy.221":"replay","blog-visuals.copy.222":"A ledger stage with an explicit contract. The next stage depends on its output.","blog-visuals.copy.223":"Every appended event folds in the previous hash. Break one link and verification complains.","blog-visuals.copy.224":"events","blog-visuals.copy.225":"The range changes event volume and makes the chain visibly longer.","blog-visuals.copy.226":"segment","blog-visuals.copy.227":"Anchoring changes a loose run into a segment that can be checked later.","blog-visuals.copy.228":"who approved it?","blog-visuals.copy.229":"was there consent?","blog-visuals.copy.230":"when did you know?","blog-visuals.copy.231":"what did you ship?","blog-visuals.copy.232":"record coverage","blog-visuals.copy.233":"Coverage shows how many review questions already point to typed events.","blog-visuals.copy.234":"session state","blog-visuals.copy.235":"checkpoint","blog-visuals.copy.236":"integrity","blog-visuals.copy.237":"Replay walks the ordered events and rebuilds the state at the selected point.","blog-visuals.copy.238":"cursor","blog-visuals.copy.239":"The range moves the replay cursor through the event chain.","blog-visuals.copy.240":"Keys repeat in every object, so structure becomes cargo carried again and again.","blog-visuals.copy.241":"The schema is declared once, then records carry positional values.","blog-visuals.copy.242":"schema","blog-visuals.copy.243":"Names and types live in one shared declaration.","blog-visuals.copy.244":"A compact positional record.","blog-visuals.copy.245":"key load","blog-visuals.copy.246":"More rows make repeated keys heavier.","blog-visuals.copy.247":"dense load","blog-visuals.copy.248":"The same rows carry less repeated structure.","blog-visuals.copy.249":"tool","blog-visuals.copy.250":"A representation in the conversion path. The meaning has to survive the trip.","blog-visuals.copy.251":"contract","blog-visuals.copy.252":"The slider simulates messy input pressure. The useful path refuses silent semantic loss.","blog-visuals.copy.253":"The baseline carries repeated structural text.","blog-visuals.copy.254":"HEDL removes repeated structure while preserving the schema.","blog-visuals.copy.255":"accuracy gain","blog-visuals.copy.256":"The documented benchmark reports a 10.3 percentage point gain over JSON.","blog-visuals.copy.257":"tasks","blog-visuals.copy.258":"Benchmark source describes 571 structured extraction tasks across seven datasets.","blog-visuals.copy.259":"fewer tokens","blog-visuals.copy.260":"The token reduction is a measured source claim, not a decorative number.","blog-visuals.copy.261":"The model still works with structured data, but the proxy controls the representation.","blog-visuals.copy.262":"MCP and proxy surfaces convert, validate, and keep downstream JSON compatibility.","blog-visuals.copy.263":"A proxy step before data reaches the next system.","blog-visuals.copy.264":"Existing systems can still receive the format they expect.","blog-visuals.copy.265":"validation","blog-visuals.copy.266":"Higher strictness blocks bad structure instead of passing it downstream.","blog-visuals.copy.267":"search","blog-visuals.copy.268":"crawl","blog-visuals.copy.269":"archive","blog-visuals.copy.270":"A source route must pass policy before it becomes input.","blog-visuals.copy.271":"collection policy","blog-visuals.copy.272":"Robots, rate limits, private ranges and source class are not optional details.","blog-visuals.copy.273":"url","blog-visuals.copy.274":"timestamp","blog-visuals.copy.275":"status","blog-visuals.copy.276":"hash","blog-visuals.copy.277":"extract","blog-visuals.copy.278":"source envelope","blog-visuals.copy.279":"The record stores where a source came from and what was extracted.","blog-visuals.copy.280":"A field that makes collection auditable instead of anecdotal.","blog-visuals.copy.281":"envelope completeness","blog-visuals.copy.282":"The range shows how much source context travels with the extract.","blog-visuals.copy.283":"raw page","blog-visuals.copy.284":"readable body","blog-visuals.copy.285":"structured rows","blog-visuals.copy.286":"table export","blog-visuals.copy.287":"source noise","blog-visuals.copy.288":"More source noise makes extraction quality matter more.","blog-visuals.copy.289":"robots","blog-visuals.copy.290":"rate","blog-visuals.copy.291":"private ranges","blog-visuals.copy.292":"request","blog-visuals.copy.293":"A collection request is not allowed just because it is technically possible.","blog-visuals.copy.294":"A safety check before network collection proceeds.","blog-visuals.copy.295":"boundary","blog-visuals.copy.296":"Strictness blocks unsafe collection instead of normalising it.","blog-visuals.copy.297":"A bit in the binary hypervector.","blog-visuals.copy.298":"Binary distance can use CPU-friendly bit operations.","blog-visuals.copy.299":"shown bits","blog-visuals.copy.300":"The range expands the visible vector, while the source default is much larger.","blog-visuals.copy.301":"A close score still needs deterministic tie-breaking.","blog-visuals.copy.302":"A candidate document in the retrieval ranking.","blog-visuals.copy.303":"ordering","blog-visuals.copy.304":"Deterministic retrieval needs stable ordering, not just similar scores.","blog-visuals.copy.305":"repeat run","blog-visuals.copy.306":"The same corpus and query should produce the same ranking.","blog-visuals.copy.307":"local corpus","blog-visuals.copy.308":"Records stay where the organisation is allowed to keep them.","blog-visuals.copy.309":"Binary index over the corpus.","blog-visuals.copy.310":"answer path","blog-visuals.copy.311":"Retrieval returns ranked candidates without exporting the corpus.","blog-visuals.copy.312":"data posture","blog-visuals.copy.313":"The range tightens locality instead of sending records elsewhere.","blog-visuals.copy.314":"query","blog-visuals.copy.315":"candidates","blog-visuals.copy.316":"tie rule","blog-visuals.copy.317":"receipt","blog-visuals.copy.318":"A retrieval step that can be recorded and inspected.","blog-visuals.copy.319":"inspection path","blog-visuals.copy.320":"Coverage shows how much of retrieval can be explained later.","blog-visuals.copy.321":"source","blog-visuals.copy.322":"The source file is not the proof. It is the input to a parse that can be recorded.","blog-visuals.copy.323":"The parsed tree has a content-addressed root.","blog-visuals.copy.324":"A parsed node whose identity follows the content.","blog-visuals.copy.325":"certificate","blog-visuals.copy.326":"The certificate root can be pinned in CI, SBOM, or release evidence.","blog-visuals.copy.327":"visible parse nodes","blog-visuals.copy.328":"The slider changes tree density without changing the contract: parse, hash, certify.","blog-visuals.copy.329":"usual parser","blog-visuals.copy.330":"Without content-addressed nodes, a small edit can look like a large rewrite.","blog-visuals.copy.331":"Reed keeps the receipt local to the changed structure.","blog-visuals.copy.332":"wall","blog-visuals.copy.333":"review noise","blog-visuals.copy.334":"Edit pressure shows how quickly ordinary parser output becomes review noise.","blog-visuals.copy.335":"portable receipt","blog-visuals.copy.336":"A stable root gives CI and auditors something smaller than trust me.","blog-visuals.copy.337":"indent","blog-visuals.copy.338":"regex or division","blog-visuals.copy.339":"nested comments","blog-visuals.copy.340":"heredoc","blog-visuals.copy.341":"A scanner rule that keeps language quirks explicit instead of buried in ad hoc parsing.","blog-visuals.copy.342":"scanner pressure","blog-visuals.copy.343":"More syntax edge cases make explicit scanners more valuable.","blog-visuals.copy.344":"one engine","blog-visuals.copy.345":"The same parser core serves every integration surface.","blog-visuals.copy.346":"A host surface over the same Reed engine, hashes, and query model.","blog-visuals.copy.347":"core reuse","blog-visuals.copy.348":"The range shows how much integration work can share one parser contract.","blog-visuals.copy.349":"laptop","blog-visuals.copy.350":"server","blog-visuals.copy.351":"embedded","blog-visuals.copy.352":"The same numeric contract should not depend on where it runs.","blog-visuals.copy.353":"same input","blog-visuals.copy.354":"drift","blog-visuals.copy.355":"The slider tightens the path to the same answer on every machine.","blog-visuals.copy.356":"answer","blog-visuals.copy.357":"Numerus is about bit-stable numeric behaviour, not a hopeful tolerance story.","blog-visuals.copy.358":"Binary XNOR and POPCNT shapes for compact ML paths.","blog-visuals.copy.359":"Ternary +1, 0 and -1 values for very small numeric surfaces.","blog-visuals.copy.360":"Native and sub-byte integers for counters, indices and bounded low-bit work.","blog-visuals.copy.361":"Q-format fixed-point aliases for bounded binary fractions.","blog-visuals.copy.362":"Small float formats when compatibility or conversion work needs them.","blog-visuals.copy.363":"Dweve AXIOM adaptive variable-point numbers with per-value exponent selection.","blog-visuals.copy.364":"packed","blog-visuals.copy.365":"fixed scale","blog-visuals.copy.366":"active group","blog-visuals.copy.367":"The wider numeric surface comes from the common numeric layer, not only the public page shortcut.","blog-visuals.copy.368":"types","blog-visuals.copy.369":"The slider broadens the visible type surface while keeping one Element-style contract.","blog-visuals.copy.370":"integer CORDIC","blog-visuals.copy.371":"property tests","blog-visuals.copy.372":"A verification step in the numeric release path.","blog-visuals.copy.373":"release check","blog-visuals.copy.374":"Compare against an oracle; do not trust a nice chart.","blog-visuals.copy.375":"Rust core","blog-visuals.copy.376":"no FPU","blog-visuals.copy.377":"release","blog-visuals.copy.378":"A deployment constraint in the numeric foundation.","blog-visuals.copy.379":"constraint gate","blog-visuals.copy.380":"The range makes deployment constraints stricter before release.","blog-visuals.copy.381":"sign","blog-visuals.copy.382":"One bit carries the sign.","blog-visuals.copy.383":"exp index","blog-visuals.copy.384":"The exponent index selects one entry from the active exponent list.","blog-visuals.copy.385":"mantissa","blog-visuals.copy.386":"The mantissa stores the integer payload before scaling.","blog-visuals.copy.387":"unused","blog-visuals.copy.388":"Unused storage keeps the representation inside a transparent u32.","blog-visuals.copy.389":"exponent list","blog-visuals.copy.390":"The concrete type chooses from a compile-time exponent list.","blog-visuals.copy.391":"selected scale","blog-visuals.copy.392":"The slider moves the selected exponent index and mantissa width together.","blog-visuals.copy.393":"storage","blog-visuals.copy.394":"Concrete AXIOM values are packed into one transparent 32-bit word.","blog-visuals.copy.395":"The slider moves a value across the lattice.","blog-visuals.copy.396":"current grid","blog-visuals.copy.397":"Fixed-point spacing is uniform. AXIOM spacing follows the selected exponent band.","blog-visuals.copy.398":"shape","blog-visuals.copy.399":"The list changes the numeric surface without becoming floating point again.","blog-visuals.copy.400":"tight fit","blog-visuals.copy.401":"fits","blog-visuals.copy.402":"input A","blog-visuals.copy.403":"A packed value brings mantissa and exponent index.","blog-visuals.copy.404":"input B","blog-visuals.copy.405":"The second value may sit on another exponent band.","blog-visuals.copy.406":"wide op","blog-visuals.copy.407":"align","blog-visuals.copy.408":"Multiplication uses wider integer mantissa work before returning to the target shape.","blog-visuals.copy.409":"Addition and subtraction align exponents before the integer operation.","blog-visuals.copy.410":"normalize","blog-visuals.copy.411":"The result is normalized back into the available exponent list.","blog-visuals.copy.412":"wide mantissa path","blog-visuals.copy.413":"aligned mantissa path","blog-visuals.copy.414":"exponent gap","blog-visuals.copy.415":"The slider increases the distance between input exponent bands.","blog-visuals.copy.416":"mantissa","blog-visuals.copy.417":"Mantissa width changes the amount of integer payload available.","blog-visuals.copy.418":"result","blog-visuals.copy.419":"Normalization is where the result has to fit the declared AXIOM shape again.","blog-visuals.copy.420":"concrete","blog-visuals.copy.421":"Packed concrete values for compile-time exponent lists.","blog-visuals.copy.422":"Runtime exponent storage when the exponent is not only a type-level choice.","blog-visuals.copy.423":"tensor","blog-visuals.copy.424":"shared exp","blog-visuals.copy.425":"Tensor and block layouts share exponent structure where the workload allows it.","blog-visuals.copy.426":"shift add","blog-visuals.copy.427":"APoT values multiply with signed powers of two, shifts, and adds.","blog-visuals.copy.428":"A lane in the vector or batch path.","blog-visuals.copy.429":"lanes","blog-visuals.copy.430":"The slider increases batch pressure and makes strategy choice visible.","blog-visuals.copy.431":"strategy","blog-visuals.copy.432":"AXIOM is a family of compute paths, not one drawing of a number.","blog-visuals.copy.433":"A candidate program in the synthesis search space.","blog-visuals.copy.434":"spec","blog-visuals.copy.435":"Typed expression plus properties define what counts as correct.","blog-visuals.copy.436":"search","blog-visuals.copy.437":"Enumerative, CEGIS, genetic, MCTS or ML-guided search proposes candidates.","blog-visuals.copy.438":"check","blog-visuals.copy.439":"Verification rejects clever wrong answers before they become kernels.","blog-visuals.copy.440":"candidates","blog-visuals.copy.441":"The slider expands the candidate pool; the mode changes how it is explored.","blog-visuals.copy.442":"engine","blog-visuals.copy.443":"Search strategy is an engineering choice, not one universal hammer.","blog-visuals.copy.444":"examples","blog-visuals.copy.445":"Fast checks catch ordinary mistakes early.","blog-visuals.copy.446":"property","blog-visuals.copy.447":"Property tests and shrinking look for counterexamples across larger domains.","blog-visuals.copy.448":"Z3 and CVC5 prove equivalence when the encoding is tractable.","blog-visuals.copy.449":"e-graph","blog-visuals.copy.450":"Equality saturation and rewrites provide another route to equivalence.","blog-visuals.copy.451":"formal","blog-visuals.copy.452":"screening","blog-visuals.copy.453":"verification level","blog-visuals.copy.454":"Harder domains move from cheap filters toward formal proof.","blog-visuals.copy.455":"active check","blog-visuals.copy.456":"The selected check changes which part of the proof stack is emphasized.","blog-visuals.copy.457":"latency","blog-visuals.copy.458":"cost","blog-visuals.copy.459":"A candidate on the latency, operation, memory, and register-pressure surface.","blog-visuals.copy.460":"The selected candidate shifts as objectives are weighted.","blog-visuals.copy.461":"objective","blog-visuals.copy.462":"Forge keeps the tradeoff visible instead of hiding it inside one score.","blog-visuals.copy.463":"weight","blog-visuals.copy.464":"The slider changes how strongly the active objective pulls the front.","blog-visuals.copy.465":"verified kernel","blog-visuals.copy.466":"The discovered implementation must remain equivalent as it lowers.","blog-visuals.copy.467":"proof travels","blog-visuals.copy.468":"A generated backend target for the same specification.","blog-visuals.copy.469":"target","blog-visuals.copy.470":"The mode selects a backend while the slider tightens portability pressure.","blog-visuals.copy.471":"portability","blog-visuals.copy.472":"More targets mean more ways for equivalence to fail, so proof matters.","blog-visuals.copy.473":"One deterministic phase in the 15-phase tick pipeline.","blog-visuals.copy.474":"phase","blog-visuals.copy.475":"The slider walks through the tick; the mode highlights the subsystem.","blog-visuals.copy.476":"subsystem","blog-visuals.copy.477":"Chemistry, metabolism, VM execution, and audit all advance in fixed order.","blog-visuals.copy.478":"Immediate energy for computation and action.","blog-visuals.copy.479":"carbon","blog-visuals.copy.480":"Fuel reserve for respiration and growth.","blog-visuals.copy.481":"protein","blog-visuals.copy.482":"Structural and repair budget.","blog-visuals.copy.483":"lipid","blog-visuals.copy.484":"Membrane and reproduction material.","blog-visuals.copy.485":"waste","blog-visuals.copy.486":"Toxic byproduct that must be managed.","blog-visuals.copy.487":"membrane","blog-visuals.copy.488":"Structural integrity; failure means death.","blog-visuals.copy.489":"Thinking is cheap, so efficient programs can plan before acting.","blog-visuals.copy.490":"Sensing costs enough that attention has to earn its keep.","blog-visuals.copy.491":"Acting changes the world and spends the budget quickly.","blog-visuals.copy.492":"Repair and waste management decide whether the cell survives the next tick.","blog-visuals.copy.493":"cost","blog-visuals.copy.494":"fragile","blog-visuals.copy.495":"alive","blog-visuals.copy.496":"state","blog-visuals.copy.497":"The slider increases survival pressure across the six pools.","blog-visuals.copy.498":"A cell reading the local chemical field.","blog-visuals.copy.499":"signal","blog-visuals.copy.500":"Signals are not messages; they diffuse through shared chemistry.","blog-visuals.copy.501":"spread","blog-visuals.copy.502":"The slider changes how far the chemical memory reaches.","blog-visuals.copy.503":"configuration plus seed","blog-visuals.copy.504":"15 deterministic phases","blog-visuals.copy.505":"population state hash","blog-visuals.copy.506":"conservation audit","blog-visuals.copy.507":"A replay checkpoint in the deterministic experiment path.","blog-visuals.copy.508":"ticks","blog-visuals.copy.509":"Longer runs still have to replay to the same population state.","blog-visuals.copy.510":"block","blog-visuals.copy.511":"pass","blog-visuals.copy.512":"release check","blog-visuals.copy.513":"Divergence is a release blocker, not a graph to explain away.","blog-visuals.copy.514":"The extension is treated as a hint, not the truth.","blog-visuals.copy.515":"A document entering the same Bindery pipeline.","blog-visuals.copy.516":"detect","blog-visuals.copy.517":"Magic bytes and structure cues choose the reader.","blog-visuals.copy.518":"parse","blog-visuals.copy.519":"Format-specific readers open packages, streams, XML, binary records, and archives.","blog-visuals.copy.520":"DocModel","blog-visuals.copy.521":"The high-level API normalizes documents, sheets, slides, tables, runs, cells, and metadata.","blog-visuals.copy.522":"SQL-like queries walk the unified model instead of six incompatible APIs.","blog-visuals.copy.523":"write","blog-visuals.copy.524":"Supported writers emit packages or output formats from the same normalized surface.","blog-visuals.copy.525":"ambiguous files","blog-visuals.copy.526":"The slider increases the amount of messy input the detector has to sort out.","blog-visuals.copy.527":"active stage","blog-visuals.copy.528":"The selected mode expands one stage of the document pipeline.","blog-visuals.copy.529":"docx, xlsx, pptx, and xlsb sit on the modern package path.","blog-visuals.copy.530":"OpenDocument text, sheets, and slides are first-class structured formats.","blog-visuals.copy.531":"legacy","blog-visuals.copy.532":"doc, xls, and ppt need OLE2 and binary record handling.","blog-visuals.copy.533":"Pages, Numbers, and Keynote travel through ZIP, Snappy, and protobuf structures.","blog-visuals.copy.534":"PDF and RTF are strong read targets and careful write targets.","blog-visuals.copy.535":"EPUB, LaTeX, and Markdown are writer outputs rather than full office twins.","blog-visuals.copy.536":"read","blog-visuals.copy.537":"write","blog-visuals.copy.538":"query","blog-visuals.copy.539":"round trip","blog-visuals.copy.540":"capability","blog-visuals.copy.541":"The mode highlights one promise across the format matrix.","blog-visuals.copy.542":"strictness","blog-visuals.copy.543":"The slider raises the bar for how much capability counts as complete.","blog-visuals.copy.544":"references","blog-visuals.copy.545":"Find every cell or object that points at a target.","blog-visuals.copy.546":"formulas","blog-visuals.copy.547":"Evaluate formula surfaces through the Excel-compatible engine.","blog-visuals.copy.548":"tables","blog-visuals.copy.549":"Walk rows, cells, headers, and extracted structure.","blog-visuals.copy.550":"shapes","blog-visuals.copy.551":"Filter drawings, slide objects, and embedded document elements.","blog-visuals.copy.552":"The query is one language over the normalized document model.","blog-visuals.copy.553":"lexer","blog-visuals.copy.554":"Tokens keep the query syntax honest.","blog-visuals.copy.555":"parser","blog-visuals.copy.556":"The AST gives later stages something typed enough to validate.","blog-visuals.copy.557":"planner","blog-visuals.copy.558":"The plan decides which connector work is needed.","blog-visuals.copy.559":"executor","blog-visuals.copy.560":"Results come from document connectors, not ad hoc scraping.","blog-visuals.copy.561":"metadata","blog-visuals.copy.562":"Metadata remains queryable instead of disappearing during extraction.","blog-visuals.copy.563":"text","blog-visuals.copy.564":"Text extraction is one part of the model, not the whole story.","blog-visuals.copy.565":"elements","blog-visuals.copy.566":"The slider increases query scope across a larger document corpus.","blog-visuals.copy.567":"focus","blog-visuals.copy.568":"The same Rust engine sits under every integration surface.","blog-visuals.copy.569":"Rust crate","blog-visuals.copy.570":"Native API for services and pipelines that want the engine directly.","blog-visuals.copy.571":"Python bindings reuse the same parsing engine for data workflows.","blog-visuals.copy.572":"The query surface keeps document questions shared across formats.","blog-visuals.copy.573":"Dweve stack","blog-visuals.copy.574":"Fabric, Spindle, BitWeave, and Reed can consume the structured output.","blog-visuals.copy.575":"connected layers","blog-visuals.copy.576":"The slider connects more of the surrounding document and knowledge stack.","blog-visuals.copy.577":"surface","blog-visuals.copy.578":"The selected mode shows where teams enter the same engine.","blog-visuals.copy.579":"A backend trace for the same FMU and input sequence.","blog-visuals.copy.580":"steps","blog-visuals.copy.581":"The slider lengthens the run and makes drift pressure easier to see.","blog-visuals.copy.582":"drift","blog-visuals.copy.583":"match","blog-visuals.copy.584":"result","blog-visuals.copy.585":"Floating implementations can diverge just enough to create reconciliation work.","blog-visuals.copy.586":"The deterministic path is designed to keep backend outputs byte-comparable.","blog-visuals.copy.587":"time and edge-friendly state","blog-visuals.copy.588":"general simulation real values","blog-visuals.copy.589":"decimal reporting and audit surfaces","blog-visuals.copy.590":"model load","blog-visuals.copy.591":"The slider increases state and precision pressure.","blog-visuals.copy.592":"profile","blog-visuals.copy.593":"Choosing the numeric profile is part of the model contract.","blog-visuals.copy.594":"The model package and deterministic runtime contract stay the shared centre.","blog-visuals.copy.595":"same model","blog-visuals.copy.596":"SIMD-accelerated host execution.","blog-visuals.copy.597":"Parallel simulation workloads can move to GPU backends.","blog-visuals.copy.598":"Hardware-in-the-loop and fixed-point kernels can target FPGA paths.","blog-visuals.copy.599":"edge","blog-visuals.copy.600":"Constrained devices need bounded memory and deterministic arithmetic.","blog-visuals.copy.601":"distributed","blog-visuals.copy.602":"Multi-node co-simulation needs coordination, not different numbers.","blog-visuals.copy.603":"targets","blog-visuals.copy.604":"The slider adds more deployment targets to the same simulation contract.","blog-visuals.copy.605":"backend","blog-visuals.copy.606":"The mode selects the backend currently being inspected.","blog-visuals.copy.607":"import","blog-visuals.copy.608":"FMU archive loading and modelDescription parsing decide what exists.","blog-visuals.copy.609":"The high-precision oracle anchors numerical checks where comparison is meaningful.","blog-visuals.copy.610":"backend parity","blog-visuals.copy.611":"CPU, GPU, FPGA, edge, and distributed paths must agree on state.","blog-visuals.copy.612":"replay","blog-visuals.copy.613":"Saved state and event logs let the run be rebuilt.","blog-visuals.copy.614":"typed failure","blog-visuals.copy.615":"A mismatch should fail loudly enough for engineering to act.","blog-visuals.copy.616":"cases","blog-visuals.copy.617":"The slider increases validation coverage across imported models and backend runs.","blog-visuals.copy.618":"block","blog-visuals.copy.619":"pass","blog-visuals.copy.620":"gate","blog-visuals.copy.621":"The harness turns numerical mismatch into a release decision, not a shrug.","blog-visuals.copy.622":"load","blog-visuals.copy.623":"Artifact bytes, policy and resource caps are pinned before execution.","blog-visuals.copy.624":"run","blog-visuals.copy.625":"Wasmtime or Kera executes inside the default-deny runtime.","blog-visuals.copy.626":"Each host call and result joins the transcript chain.","blog-visuals.copy.627":"seal","blog-visuals.copy.628":"The envelope is signed so offline verification can start from evidence.","blog-visuals.copy.629":"A recorded host call in the execution transcript.","blog-visuals.copy.630":"host calls","blog-visuals.copy.631":"The slider grows the transcript and makes the hash chain longer.","blog-visuals.copy.632":"stage","blog-visuals.copy.633":"The mode highlights which part of the evidence envelope is active.","blog-visuals.copy.634":"capabilities","blog-visuals.copy.635":"Clock, random, filesystem, network, environment, stdio, process and custom calls start denied.","blog-visuals.copy.636":"fuel","blog-visuals.copy.637":"Instruction limits stop runaway work cleanly.","blog-visuals.copy.638":"memory","blog-visuals.copy.639":"Memory caps keep the workload inside the declared box.","blog-visuals.copy.640":"time","blog-visuals.copy.641":"Wall-clock caps sit beside deterministic fuel accounting.","blog-visuals.copy.642":"default deny","blog-visuals.copy.643":"policy","blog-visuals.copy.644":"Policy validates invariants before a run is allowed to start.","blog-visuals.copy.645":"validate","blog-visuals.copy.646":"runner","blog-visuals.copy.647":"RunnerConfig keeps deterministic execution on by default.","blog-visuals.copy.648":"deterministic","blog-visuals.copy.649":"pressure","blog-visuals.copy.650":"The slider raises resource pressure across the gates.","blog-visuals.copy.651":"gate","blog-visuals.copy.652":"The selected gate explains why containment alone is not enough.","blog-visuals.copy.653":"virtual clock","blog-visuals.copy.654":"A fixed epoch removes host time from replay.","blog-visuals.copy.655":"seeded RNG","blog-visuals.copy.656":"ChaCha20 randomness is seeded instead of borrowed from the host.","blog-visuals.copy.657":"NaN canon","blog-visuals.copy.658":"Floating NaN values canonicalize instead of leaking platform personality.","blog-visuals.copy.659":"SIMD profile","blog-visuals.copy.660":"SIMD can be disabled when cross-platform bit identity matters more than speed.","blog-visuals.copy.661":"profiled","blog-visuals.copy.662":"identical","blog-visuals.copy.663":"replay","blog-visuals.copy.664":"The selected lever changes how replay is protected.","blog-visuals.copy.665":"run pressure","blog-visuals.copy.666":"Longer runs make nondeterminism easier to expose.","blog-visuals.copy.667":"One runner surface wraps untrusted execution with policy and evidence.","blog-visuals.copy.668":"Wasmtime Component Model execution for WebAssembly artifacts.","blog-visuals.copy.669":"Graph IR execution for binary neural network workloads.","blog-visuals.copy.670":"Agent tool execution can go through the same transcript layer.","blog-visuals.copy.671":"service","blog-visuals.copy.672":"Long-running execution uses pooling, queueing and metrics.","blog-visuals.copy.673":"surface","blog-visuals.copy.674":"The mode chooses the execution surface currently being inspected.","blog-visuals.copy.675":"same envelope","blog-visuals.copy.676":"evidence","blog-visuals.copy.677":"The useful part is that all entry points produce comparable evidence.","blog-visuals.copy.678":"author","blog-visuals.copy.679":"Rules and constraints live next to code and version history.","blog-visuals.copy.680":"compile","blog-visuals.copy.681":"AOT compilation classifies constraints and pins the backend.","blog-visuals.copy.682":"pack","blog-visuals.copy.683":"The artefact becomes a memory-mappable binary with checksums.","blog-visuals.copy.684":"evaluate","blog-visuals.copy.685":"Evaluation is an in-process packed-tree walk, not a service call.","blog-visuals.copy.686":"rules","blog-visuals.copy.687":"The slider increases the rulebook size before compilation.","blog-visuals.copy.688":"stage","blog-visuals.copy.689":"The mode shows where the rulebook is in the pipeline.","blog-visuals.copy.690":"rule API","blog-visuals.copy.691":"The caller writes rules, not solver-specific calls.","blog-visuals.copy.692":"classify","blog-visuals.copy.693":"Pure boolean rules and feature gates.","blog-visuals.copy.694":"Boolean, arithmetic, arrays and bitvectors under one decision procedure.","blog-visuals.copy.695":"Continuous optimisation for pricing and allocation.","blog-visuals.copy.696":"Integer decisions for scheduling and assignment.","blog-visuals.copy.697":"Finite-domain and non-linear constraint propagation.","blog-visuals.copy.698":"Path and soft-constraint problems with weighted preferences.","blog-visuals.copy.699":"constraint","blog-visuals.copy.700":"The mode changes which solver family the selector favours.","blog-visuals.copy.701":"complexity","blog-visuals.copy.702":"The slider adds mixed constraints to the same rule surface.","blog-visuals.copy.703":"sidecar","blog-visuals.copy.704":"A network policy service adds a round trip and a service dependency.","blog-visuals.copy.705":"interpreter","blog-visuals.copy.706":"Interpreted rules keep parser and dispatch work on the request path.","blog-visuals.copy.707":"cold mmap","blog-visuals.copy.708":"The first lookup after mmap pays the cache miss.","blog-visuals.copy.709":"hot eval","blog-visuals.copy.710":"The packed tree is already cache resident.","blog-visuals.copy.711":"budget","blog-visuals.copy.712":"At scale, small per-check costs become request latency.","blog-visuals.copy.713":"checks","blog-visuals.copy.714":"The slider multiplies the decision cost by request volume.","blog-visuals.copy.715":"path","blog-visuals.copy.716":"The mode highlights the latency path being compared.","blog-visuals.copy.717":"version","blog-visuals.copy.718":"The rulebook version is named, not reconstructed later.","blog-visuals.copy.719":"checksum","blog-visuals.copy.720":"Header, content and file checksums catch tampered artefacts.","blog-visuals.copy.721":"input","blog-visuals.copy.722":"Logged inputs let the decision be replayed mechanically.","blog-visuals.copy.723":"answer","blog-visuals.copy.724":"The result is the same because the artefact decides the bits.","blog-visuals.copy.725":"audit age","blog-visuals.copy.726":"The slider moves the audit further away from the original decision.","blog-visuals.copy.727":"replay","blog-visuals.copy.728":"evidence","blog-visuals.copy.729":"The explanation is the rule and input replay, not a paragraph recreated later.","blog-visuals.dweve-discrete-stack.copy.binary":"binary core","blog-visuals.dweve-discrete-stack.copy.binary.detail":"Discrete representations keep the path small and portable.","blog-visuals.dweve-discrete-stack.copy.constraints":"constraints","blog-visuals.dweve-discrete-stack.copy.constraints.detail":"Rules are explicit enough to inspect and test.","blog-visuals.dweve-discrete-stack.copy.deploy":"deployment","blog-visuals.dweve-discrete-stack.copy.deploy.detail":"The same logic should travel from local machines to controlled infrastructure.","blog-visuals.dweve-discrete-stack.copy.dwevePath":"Dweve path","blog-visuals.dweve-discrete-stack.copy.dwevePath.detail":"The alternative path starts with control, explicit logic and portable execution.","blog-visuals.dweve-discrete-stack.copy.pressure":"deployment pressure","blog-visuals.dweve-discrete-stack.copy.pressure.detail":"The slider raises deployment pressure without changing the control requirement.","blog-visuals.dweve-discrete-stack.copy.transformerPlaybook":"GPU playbook","blog-visuals.dweve-discrete-stack.copy.transformerPlaybook.detail":"The standard route buys scale first and asks control questions later.","blog-visuals.dweve-discrete-stack.copy.verify":"verification","blog-visuals.dweve-discrete-stack.copy.verify.detail":"Evidence and replay matter more than a confident answer.","blog-visuals.dweve-discrete-stack.eyebrow":"Discrete AI stack","blog-visuals.dweve-discrete-stack.mode.binary":"binary","blog-visuals.dweve-discrete-stack.mode.constraints":"constraints","blog-visuals.dweve-discrete-stack.mode.deploy":"deploy","blog-visuals.dweve-discrete-stack.mode.verify":"verify","blog-visuals.dweve-discrete-stack.range":"deployment pressure","blog-visuals.dweve-discrete-stack.title":"Different starts with control, not another GPU invoice","blog-visuals.edge-cost-crossover.copy.cloud":"cloud","blog-visuals.edge-cost-crossover.copy.cloud.detail":"Metered inference keeps growing with usage.","blog-visuals.edge-cost-crossover.copy.edge":"edge","blog-visuals.edge-cost-crossover.copy.edge.detail":"Edge spend starts higher but flattens as utilisation grows.","blog-visuals.edge-cost-crossover.copy.ownedPath":"owned path","blog-visuals.edge-cost-crossover.copy.ownedPath.detail":"Edge capacity turns repeated inference into utilisation of infrastructure already bought.","blog-visuals.edge-cost-crossover.copy.saving":"saving","blog-visuals.edge-cost-crossover.copy.saving.detail":"At scale, the gap is less about servers and more about whether success stays metered forever.","blog-visuals.edge-cost-crossover.copy.usageTax":"usage tax","blog-visuals.edge-cost-crossover.copy.usageTax.detail":"Cloud inference charges again for every success case.","blog-visuals.edge-cost-crossover.copy.users":"users","blog-visuals.edge-cost-crossover.copy.users.detail":"The slider moves from pilot usage to product-scale traffic.","blog-visuals.edge-cost-crossover.eyebrow":"Cost crossover","blog-visuals.edge-cost-crossover.mode.cloud":"cloud","blog-visuals.edge-cost-crossover.mode.crossover":"crossover","blog-visuals.edge-cost-crossover.mode.edge":"edge","blog-visuals.edge-cost-crossover.range":"user scale","blog-visuals.edge-cost-crossover.title":"At scale, pay-per-call becomes a usage tax","blog-visuals.edge-latency-path.copy.cloud":"cloud","blog-visuals.edge-latency-path.copy.cloud.detail":"The cloud path includes distance, routing, queueing and provider load before the model answers.","blog-visuals.edge-latency-path.copy.cloudLatency":"cloud path","blog-visuals.edge-latency-path.copy.cloudLatency.detail":"The slider moves the cloud region farther away and exposes the round-trip cost.","blog-visuals.edge-latency-path.copy.device":"device","blog-visuals.edge-latency-path.copy.device.detail":"The request starts near the user. Moving it away spends latency before compute begins.","blog-visuals.edge-latency-path.copy.edge":"edge","blog-visuals.edge-latency-path.copy.edge.detail":"Local or nearby inference keeps the round trip inside the operating envelope.","blog-visuals.edge-latency-path.copy.edgeLatency":"edge path","blog-visuals.edge-latency-path.copy.edgeLatency.detail":"Edge latency stays small because the useful work remains nearby.","blog-visuals.edge-latency-path.copy.required":"required","blog-visuals.edge-latency-path.copy.required.detail":"The selected mode changes the application tolerance for delay.","blog-visuals.edge-latency-path.eyebrow":"Latency path","blog-visuals.edge-latency-path.mode.cloud":"cloud","blog-visuals.edge-latency-path.mode.edge":"edge","blog-visuals.edge-latency-path.mode.realtime":"real time","blog-visuals.edge-latency-path.range":"distance pressure","blog-visuals.edge-latency-path.title":"The round trip starts before inference does","blog-visuals.edge-mesh-topology.copy.eligible":"eligible","blog-visuals.edge-mesh-topology.copy.eligible.detail":"Scheduling is useful only when it can reject nodes that do not match the workload.","blog-visuals.edge-mesh-topology.copy.mode":"mode","blog-visuals.edge-mesh-topology.copy.mode.detail":"Discovery, scheduling and healing are different parts of the same mesh contract.","blog-visuals.edge-mesh-topology.copy.node.detail":"Each peer advertises identity, capability, locality and current load before receiving work.","blog-visuals.edge-mesh-topology.copy.nodes":"nodes","blog-visuals.edge-mesh-topology.copy.nodes.detail":"The slider changes how many peers the fabric has to coordinate.","blog-visuals.edge-mesh-topology.copy.policy":"policy","blog-visuals.edge-mesh-topology.copy.policy.detail":"The useful centre of the mesh is not a server. It is the placement rule every node must satisfy.","blog-visuals.edge-mesh-topology.eyebrow":"Mesh topology","blog-visuals.edge-mesh-topology.mode.discover":"discover","blog-visuals.edge-mesh-topology.mode.heal":"heal","blog-visuals.edge-mesh-topology.mode.schedule":"schedule","blog-visuals.edge-mesh-topology.range":"node pressure","blog-visuals.edge-mesh-topology.title":"A mesh is a placement contract, not a drawing of dots","blog-visuals.edge-tco-inversion.copy.capex":"CapEx once","blog-visuals.edge-tco-inversion.copy.capex.detail":"Edge hardware moves cost into the bill of materials.","blog-visuals.edge-tco-inversion.copy.cloud":"cloud","blog-visuals.edge-tco-inversion.copy.cloud.detail":"A small per-request fee becomes a lifetime product liability.","blog-visuals.edge-tco-inversion.copy.edge":"edge","blog-visuals.edge-tco-inversion.copy.edge.detail":"The chip cost is paid once and local inference stays local.","blog-visuals.edge-tco-inversion.copy.lifetime":"lifetime","blog-visuals.edge-tco-inversion.copy.lifetime.detail":"Long-lived devices punish rented inference.","blog-visuals.edge-tco-inversion.copy.opex":"OpEx forever","blog-visuals.edge-tco-inversion.copy.opex.detail":"Cloud inference keeps charging after the sale.","blog-visuals.edge-tco-inversion.copy.saving":"saving","blog-visuals.edge-tco-inversion.copy.saving.detail":"The slider stretches product lifetime and exposes the OpEx tail.","blog-visuals.edge-tco-inversion.eyebrow":"TCO inversion","blog-visuals.edge-tco-inversion.mode.cloud":"cloud","blog-visuals.edge-tco-inversion.mode.edge":"edge","blog-visuals.edge-tco-inversion.mode.lifetime":"lifetime","blog-visuals.edge-tco-inversion.range":"product lifetime","blog-visuals.edge-tco-inversion.title":"A small cloud fee becomes a product lifetime cost","blog-visuals.energy-density-crisis.copy.cooling":"cooling","blog-visuals.energy-density-crisis.copy.cooling.detail":"The mode selects the rack class and cooling consequence.","blog-visuals.energy-density-crisis.copy.current":"current AI","blog-visuals.energy-density-crisis.copy.current.detail":"Liquid cooling becomes normal once rack density jumps.","blog-visuals.energy-density-crisis.copy.density":"density","blog-visuals.energy-density-crisis.copy.density.detail":"The slider raises density pressure across the same facility envelope.","blog-visuals.energy-density-crisis.copy.frontier":"frontier","blog-visuals.energy-density-crisis.copy.frontier.detail":"Immersion and direct-to-chip cooling are symptoms of power density, not decoration.","blog-visuals.energy-density-crisis.copy.traditional":"traditional","blog-visuals.energy-density-crisis.copy.traditional.detail":"Air-cooled racks were already serious infrastructure.","blog-visuals.energy-density-crisis.eyebrow":"Power density","blog-visuals.energy-density-crisis.mode.current":"current AI","blog-visuals.energy-density-crisis.mode.frontier":"frontier","blog-visuals.energy-density-crisis.mode.traditional":"traditional","blog-visuals.energy-density-crisis.range":"density pressure","blog-visuals.energy-density-crisis.title":"The rack became a physics problem","blog-visuals.energy-twh-curve.copy.infrastructure":"infrastructure","blog-visuals.energy-twh-curve.copy.infrastructure.detail":"Electricity, cooling and siting become product constraints, not facilities trivia.","blog-visuals.energy-twh-curve.copy.load":"load","blog-visuals.energy-twh-curve.copy.load.detail":"Model demand becomes grid demand when inference scales.","blog-visuals.energy-twh-curve.copy.spain":"Spain","blog-visuals.energy-twh-curve.copy.spain.detail":"The comparison anchors the scale against a national electricity footprint.","blog-visuals.energy-twh-curve.copy.twh":"TWh","blog-visuals.energy-twh-curve.copy.twh.detail":"The slider moves from current data-centre demand toward the 2030 projection.","blog-visuals.energy-twh-curve.eyebrow":"Energy demand","blog-visuals.energy-twh-curve.mode.cooling":"cooling","blog-visuals.energy-twh-curve.mode.demand":"demand","blog-visuals.energy-twh-curve.mode.grid":"grid","blog-visuals.energy-twh-curve.range":"2030 pressure","blog-visuals.energy-twh-curve.title":"AI demand turns electricity into a product constraint","blog-visuals.expert-parliament.copy.active":"active","blog-visuals.expert-parliament.copy.active.detail":"Only a subset should wake up for a given request. That is the economic point.","blog-visuals.expert-parliament.copy.expert.detail":"A small model or specialist route handles a narrow part of the request.","blog-visuals.expert-parliament.copy.experts":"experts","blog-visuals.expert-parliament.copy.experts.detail":"The slider raises how many specialists are available.","blog-visuals.expert-parliament.copy.mode":"mode","blog-visuals.expert-parliament.copy.mode.detail":"Routing, voting and verification describe different parts of a small-model system.","blog-visuals.expert-parliament.copy.router":"router","blog-visuals.expert-parliament.copy.router.detail":"The router chooses specialists instead of waking one giant model for every task.","blog-visuals.expert-parliament.eyebrow":"Expert parliament","blog-visuals.expert-parliament.mode.route":"route","blog-visuals.expert-parliament.mode.verify":"verify","blog-visuals.expert-parliament.mode.vote":"vote","blog-visuals.expert-parliament.range":"specialist count","blog-visuals.expert-parliament.title":"Small specialists beat one giant guesser when routing works","blog-visuals.fallback.mode.overview":"overview","blog-visuals.fmi-backend-roster.eyebrow":"FMI backend roster","blog-visuals.fmi-backend-roster.mode.cpu":"CPU","blog-visuals.fmi-backend-roster.mode.dist":"distributed","blog-visuals.fmi-backend-roster.mode.edge":"edge","blog-visuals.fmi-backend-roster.mode.fpga":"FPGA","blog-visuals.fmi-backend-roster.mode.gpu":"GPU","blog-visuals.fmi-backend-roster.range":"deployment pressure","blog-visuals.fmi-backend-roster.title":"Deployment should change hardware, not the answer","blog-visuals.fmi-drift-traces.eyebrow":"FMI drift traces","blog-visuals.fmi-drift-traces.mode.fixed":"fixed-point","blog-visuals.fmi-drift-traces.mode.float":"float","blog-visuals.fmi-drift-traces.mode.mpfr":"MPFR","blog-visuals.fmi-drift-traces.range":"run length","blog-visuals.fmi-drift-traces.title":"The same FMU should not become three stories","blog-visuals.fmi-profile-ladder.eyebrow":"FMI numeric profiles","blog-visuals.fmi-profile-ladder.mode.dec64":"Dec64_6","blog-visuals.fmi-profile-ladder.mode.q16":"Q16.16","blog-visuals.fmi-profile-ladder.mode.q31":"Q31.32","blog-visuals.fmi-profile-ladder.range":"state pressure","blog-visuals.fmi-profile-ladder.title":"The numeric profile is part of the model contract","blog-visuals.fmi-validation-harness.eyebrow":"FMI validation harness","blog-visuals.fmi-validation-harness.mode.failure":"failure","blog-visuals.fmi-validation-harness.mode.import":"import","blog-visuals.fmi-validation-harness.mode.oracle":"oracle","blog-visuals.fmi-validation-harness.mode.parity":"parity","blog-visuals.fmi-validation-harness.mode.replay":"replay","blog-visuals.fmi-validation-harness.range":"coverage","blog-visuals.fmi-validation-harness.title":"A mismatch is a release decision","blog-visuals.forge-pareto-front.eyebrow":"Forge Pareto front","blog-visuals.forge-pareto-front.mode.latency":"latency","blog-visuals.forge-pareto-front.mode.memory":"memory","blog-visuals.forge-pareto-front.mode.registers":"registers","blog-visuals.forge-pareto-front.range":"objective weight","blog-visuals.forge-pareto-front.title":"Fast is one objective, not the whole truth","blog-visuals.forge-search-space.eyebrow":"Forge search space","blog-visuals.forge-search-space.mode.cegis":"CEGIS","blog-visuals.forge-search-space.mode.enumerative":"enumerative","blog-visuals.forge-search-space.mode.mcts":"MCTS","blog-visuals.forge-search-space.mode.ml":"ML-guided","blog-visuals.forge-search-space.range":"candidate pressure","blog-visuals.forge-search-space.title":"Synthesis searches candidates, not vibes","blog-visuals.forge-target-lowering.eyebrow":"Forge target lowering","blog-visuals.forge-target-lowering.mode.c":"C","blog-visuals.forge-target-lowering.mode.riscv":"RISC-V","blog-visuals.forge-target-lowering.mode.verilog":"Verilog","blog-visuals.forge-target-lowering.mode.vulkan":"Vulkan","blog-visuals.forge-target-lowering.mode.wasm":"WASM","blog-visuals.forge-target-lowering.mode.x86":"x86-64","blog-visuals.forge-target-lowering.range":"portability pressure","blog-visuals.forge-target-lowering.title":"The proof has to survive the backend","blog-visuals.forge-verification-funnel.eyebrow":"Forge verification funnel","blog-visuals.forge-verification-funnel.mode.egraph":"e-graph","blog-visuals.forge-verification-funnel.mode.examples":"examples","blog-visuals.forge-verification-funnel.mode.property":"property","blog-visuals.forge-verification-funnel.mode.smt":"SMT","blog-visuals.forge-verification-funnel.range":"domain hardness","blog-visuals.forge-verification-funnel.title":"Wrong fast code is still wrong","blog-visuals.gpu-dependency-ledger.copy.cloud":"cloud rent","blog-visuals.gpu-dependency-ledger.copy.cloud.detail":"Rented accelerators export both margin and operational leverage.","blog-visuals.gpu-dependency-ledger.copy.control":"control plane","blog-visuals.gpu-dependency-ledger.copy.control.detail":"The sovereign question is who can still change course when allocations, prices or APIs move.","blog-visuals.gpu-dependency-ledger.copy.hardware":"hardware","blog-visuals.gpu-dependency-ledger.copy.hardware.detail":"Chips, boards and packaging are strategic imports when the stack depends on them.","blog-visuals.gpu-dependency-ledger.copy.local":"local stack","blog-visuals.gpu-dependency-ledger.copy.local.detail":"European retention grows only when more of the stack can run on controlled infrastructure.","blog-visuals.gpu-dependency-ledger.copy.outflow":"outflow","blog-visuals.gpu-dependency-ledger.copy.outflow.detail":"Dependency is visible as money, but the larger cost is strategic optionality.","blog-visuals.gpu-dependency-ledger.copy.runway":"runway","blog-visuals.gpu-dependency-ledger.copy.runway.detail":"Startups lose runway when infrastructure scales faster than revenue.","blog-visuals.gpu-dependency-ledger.copy.spend":"annual spend","blog-visuals.gpu-dependency-ledger.copy.spend.detail":"The slider scales the annual infrastructure budget under pressure.","blog-visuals.gpu-dependency-ledger.eyebrow":"Dependency ledger","blog-visuals.gpu-dependency-ledger.mode.procurement":"procurement","blog-visuals.gpu-dependency-ledger.mode.sovereign":"sovereign","blog-visuals.gpu-dependency-ledger.mode.startup":"startup","blog-visuals.gpu-dependency-ledger.range":"budget pressure","blog-visuals.gpu-dependency-ledger.title":"The bill is also a control map","blog-visuals.gpu-market-moat.copy.accelerator":"accelerator","blog-visuals.gpu-market-moat.copy.accelerator.detail":"The market centre is hardware plus software plus supply control.","blog-visuals.gpu-market-moat.copy.chips":"chips","blog-visuals.gpu-market-moat.copy.chips.detail":"The hardware lead is visible, expensive and difficult to copy quickly.","blog-visuals.gpu-market-moat.copy.lockIn":"lock-in","blog-visuals.gpu-market-moat.copy.lockIn.detail":"Lock-in rises when software, skills and supply chains point at the same gate.","blog-visuals.gpu-market-moat.copy.market":"market","blog-visuals.gpu-market-moat.copy.market.detail":"The slider raises concentration pressure in the accelerator market.","blog-visuals.gpu-market-moat.copy.migration":"migration cost","blog-visuals.gpu-market-moat.copy.migration.detail":"A moat works when leaving means rewriting code, retraining teams and changing procurement at once.","blog-visuals.gpu-market-moat.copy.software":"software","blog-visuals.gpu-market-moat.copy.software.detail":"CUDA, libraries and developer habits make the chip more than a chip.","blog-visuals.gpu-market-moat.copy.supply":"supply","blog-visuals.gpu-market-moat.copy.supply.detail":"Allocation, packaging and lead times turn dependency into a planning constraint.","blog-visuals.gpu-market-moat.eyebrow":"Market moat","blog-visuals.gpu-market-moat.mode.cuda":"software","blog-visuals.gpu-market-moat.mode.hardware":"hardware","blog-visuals.gpu-market-moat.mode.supply":"supply","blog-visuals.gpu-market-moat.range":"lock-in pressure","blog-visuals.gpu-market-moat.title":"The trap is hardware, software and supply at once","blog-visuals.gpu-power-escape.copy.binary":"binary","blog-visuals.gpu-power-escape.copy.binary.detail":"Binary inference changes the operation, memory shape and deploy target at the same time.","blog-visuals.gpu-power-escape.copy.cpu":"CPU edge","blog-visuals.gpu-power-escape.copy.cpu.detail":"The point is not that every task belongs on a CPU. The point is that not every task belongs on rented accelerators.","blog-visuals.gpu-power-escape.copy.gap":"power gap","blog-visuals.gpu-power-escape.copy.gap.detail":"The slider raises workload size while preserving the architectural energy gap.","blog-visuals.gpu-power-escape.copy.gpu":"GPU float","blog-visuals.gpu-power-escape.copy.gpu.detail":"General accelerator math carries power, cooling and allocation assumptions with it.","blog-visuals.gpu-power-escape.copy.ops":"ops","blog-visuals.gpu-power-escape.copy.ops.detail":"The selected workload stands in for repeated inference, not one heroic benchmark.","blog-visuals.gpu-power-escape.copy.optionality":"optionality","blog-visuals.gpu-power-escape.copy.optionality.detail":"More deployment targets create leverage in procurement, policy and engineering.","blog-visuals.gpu-power-escape.copy.paradigm":"paradigm","blog-visuals.gpu-power-escape.copy.paradigm.detail":"Escaping the GPU trap means changing the arithmetic shape, not merely buying a different card.","blog-visuals.gpu-power-escape.eyebrow":"Power escape","blog-visuals.gpu-power-escape.mode.binary":"binary","blog-visuals.gpu-power-escape.mode.gpu":"GPU","blog-visuals.gpu-power-escape.mode.sovereign":"sovereign","blog-visuals.gpu-power-escape.range":"workload pressure","blog-visuals.gpu-power-escape.title":"The exit starts by changing the arithmetic","blog-visuals.green-arithmetic-cost.copy.binaryDetail":"XNOR and POPCNT turn many comparisons into bit operations CPUs already understand.","blog-visuals.green-arithmetic-cost.copy.binaryLane":"binary lane","blog-visuals.green-arithmetic-cost.copy.binaryPj":"binary pJ","blog-visuals.green-arithmetic-cost.copy.binaryPj.detail":"The same work expressed as binary logic keeps the switching cost low.","blog-visuals.green-arithmetic-cost.copy.fpDetail":"FP32 moves sign, exponent, mantissa, rounding and exception logic through silicon.","blog-visuals.green-arithmetic-cost.copy.fpLane":"floating-point lane","blog-visuals.green-arithmetic-cost.copy.fpPj":"FP32 pJ","blog-visuals.green-arithmetic-cost.copy.fpPj.detail":"The slider scales a million-operation workload and shows why precision has a thermal bill.","blog-visuals.green-arithmetic-cost.copy.saved":"saved","blog-visuals.green-arithmetic-cost.copy.saved.detail":"The useful number is not a benchmark trophy. It is power, heat and cooling avoided.","blog-visuals.green-arithmetic-cost.eyebrow":"Arithmetic cost","blog-visuals.green-arithmetic-cost.mode.binary":"binary","blog-visuals.green-arithmetic-cost.mode.fp32":"FP32","blog-visuals.green-arithmetic-cost.mode.simd":"SIMD","blog-visuals.green-arithmetic-cost.range":"operation pressure","blog-visuals.green-arithmetic-cost.title":"Precision has a thermal bill","blog-visuals.green-compliance-gate.copy.auditReady":"audit ready","blog-visuals.green-compliance-gate.copy.auditReady.detail":"The slider raises how much evidence the buyer expects before signing.","blog-visuals.green-compliance-gate.copy.contract":"contract field","blog-visuals.green-compliance-gate.copy.contract.detail":"Good green AI claims become operational fields buyers can verify.","blog-visuals.green-compliance-gate.copy.energy":"energy","blog-visuals.green-compliance-gate.copy.energy.detail":"The buyer needs usage per workload, not a green badge on the provider homepage.","blog-visuals.green-compliance-gate.copy.gaps":"open gaps","blog-visuals.green-compliance-gate.copy.gaps.detail":"Missing evidence becomes visible as reporting pressure increases.","blog-visuals.green-compliance-gate.copy.gco2e":"gCO2e/query","blog-visuals.green-compliance-gate.copy.procurement":"procurement","blog-visuals.green-compliance-gate.copy.procurement.detail":"Carbon intensity becomes a purchasing field, not a sustainability appendix.","blog-visuals.green-compliance-gate.copy.scope3":"Scope 3","blog-visuals.green-compliance-gate.copy.scope3.detail":"Purchased AI services land in supply-chain reporting, so the meter follows procurement.","blog-visuals.green-compliance-gate.eyebrow":"Compliance gate","blog-visuals.green-compliance-gate.mode.procurement":"procurement","blog-visuals.green-compliance-gate.mode.report":"report","blog-visuals.green-compliance-gate.mode.scope3":"Scope 3","blog-visuals.green-compliance-gate.range":"audit pressure","blog-visuals.green-compliance-gate.title":"Green claims become procurement evidence","blog-visuals.green-grid-topology.copy.cloud":"cloud","blog-visuals.green-grid-topology.copy.cloud.detail":"A central cluster concentrates power draw, cooling, grid approval and failure blast radius.","blog-visuals.green-grid-topology.copy.cloudLoad":"central load","blog-visuals.green-grid-topology.copy.cloudLoad.detail":"Centralised AI turns product adoption into a grid-connection problem.","blog-visuals.green-grid-topology.copy.edge":"edge","blog-visuals.green-grid-topology.copy.edge.detail":"On-device or nearby inference spreads smaller workloads over existing hardware.","blog-visuals.green-grid-topology.copy.edgeLoad":"edge load","blog-visuals.green-grid-topology.copy.edgeLoad.detail":"Distributed inference still consumes energy, but it removes the single giant demand spike.","blog-visuals.green-grid-topology.copy.topology":"topology","blog-visuals.green-grid-topology.copy.topology.detail":"The mode changes where the same demand lands physically.","blog-visuals.green-grid-topology.copy.user":"user","blog-visuals.green-grid-topology.copy.user.detail":"The useful work starts at the user, so the question is how far the request must travel.","blog-visuals.green-grid-topology.eyebrow":"Grid topology","blog-visuals.green-grid-topology.mode.cloud":"cloud","blog-visuals.green-grid-topology.mode.edge":"edge","blog-visuals.green-grid-topology.mode.mixed":"mixed","blog-visuals.green-grid-topology.range":"adoption pressure","blog-visuals.green-grid-topology.title":"Where inference runs decides where the grid hurts","blog-visuals.green-memory-traffic.copy.binary":"binary pack","blog-visuals.green-memory-traffic.copy.binary.detail":"Packed bits reduce traffic before the compute engine wakes up.","blog-visuals.green-memory-traffic.copy.binaryMb":"binary MB","blog-visuals.green-memory-traffic.copy.binaryMb.detail":"One bit per weight changes the memory problem before optimisation begins.","blog-visuals.green-memory-traffic.copy.cache":"cache","blog-visuals.green-memory-traffic.copy.cache.detail":"Keeping hot state in cache saves more than shaving a tiny arithmetic step.","blog-visuals.green-memory-traffic.copy.dram":"DRAM","blog-visuals.green-memory-traffic.copy.dram.detail":"Main-memory traffic dominates energy once model state no longer fits near the CPU.","blog-visuals.green-memory-traffic.copy.fitCloser":"fit closer","blog-visuals.green-memory-traffic.copy.fitCloser.detail":"Smaller representations keep more of the active model in cache and on device.","blog-visuals.green-memory-traffic.copy.fpMb":"FP32 MB","blog-visuals.green-memory-traffic.copy.fpMb.detail":"Four bytes per parameter quickly becomes freight, not maths.","blog-visuals.green-memory-traffic.copy.moveLess":"move less","blog-visuals.green-memory-traffic.copy.moveLess.detail":"Energy is often spent carrying data to arithmetic, not doing the arithmetic itself.","blog-visuals.green-memory-traffic.eyebrow":"Memory traffic","blog-visuals.green-memory-traffic.mode.binary":"binary pack","blog-visuals.green-memory-traffic.mode.cache":"cache","blog-visuals.green-memory-traffic.mode.dram":"DRAM","blog-visuals.green-memory-traffic.range":"model size","blog-visuals.green-memory-traffic.title":"Moving model state is part of the energy bill","blog-visuals.ground-truth-black-box-loop.eyebrow":"Ground Truth black-box loop","blog-visuals.ground-truth-black-box-loop.mode.ask":"ask","blog-visuals.ground-truth-black-box-loop.mode.revise":"revise","blog-visuals.ground-truth-black-box-loop.mode.test":"test","blog-visuals.ground-truth-black-box-loop.range":"uncertainty","blog-visuals.ground-truth-black-box-loop.title":"A better question is the real teaching unit","blog-visuals.ground-truth-curriculum-map.eyebrow":"Ground Truth curriculum map","blog-visuals.ground-truth-curriculum-map.mode.deep":"deep route","blog-visuals.ground-truth-curriculum-map.mode.fast":"fast route","blog-visuals.ground-truth-curriculum-map.mode.practical":"practical route","blog-visuals.ground-truth-curriculum-map.range":"route depth","blog-visuals.ground-truth-curriculum-map.title":"The order keeps later chapters grounded","blog-visuals.ground-truth-literacy-stairs.eyebrow":"Ground Truth literacy stairs","blog-visuals.ground-truth-literacy-stairs.mode.judgement":"judgement","blog-visuals.ground-truth-literacy-stairs.mode.models":"models","blog-visuals.ground-truth-literacy-stairs.mode.words":"words","blog-visuals.ground-truth-literacy-stairs.range":"room pressure","blog-visuals.ground-truth-literacy-stairs.title":"Adult AI literacy climbs by judgement","blog-visuals.ground-truth-math-bridge.eyebrow":"Ground Truth maths bridge","blog-visuals.ground-truth-math-bridge.mode.optimisation":"optimisation","blog-visuals.ground-truth-math-bridge.mode.probability":"probability","blog-visuals.ground-truth-math-bridge.mode.vectors":"vectors","blog-visuals.ground-truth-math-bridge.range":"abstraction","blog-visuals.ground-truth-math-bridge.title":"Math turns theatre into inspectable claims","blog-visuals.hedl-benchmark-lens.eyebrow":"HEDL benchmark lens","blog-visuals.hedl-benchmark-lens.mode.accuracy":"accuracy","blog-visuals.hedl-benchmark-lens.mode.tasks":"tasks","blog-visuals.hedl-benchmark-lens.mode.tokens":"tokens","blog-visuals.hedl-benchmark-lens.range":"workload","blog-visuals.hedl-benchmark-lens.title":"Token cost is part of the interface","blog-visuals.hedl-proxy-gate.eyebrow":"HEDL proxy","blog-visuals.hedl-proxy-gate.mode.llm":"LLM","blog-visuals.hedl-proxy-gate.mode.proxy":"proxy","blog-visuals.hedl-proxy-gate.mode.validate":"validate","blog-visuals.hedl-proxy-gate.range":"strictness","blog-visuals.hedl-proxy-gate.title":"The proxy keeps structure honest","blog-visuals.hedl-round-trip.eyebrow":"HEDL round trip","blog-visuals.hedl-round-trip.mode.hedl":"HEDL","blog-visuals.hedl-round-trip.mode.json":"JSON","blog-visuals.hedl-round-trip.mode.roundtrip":"round trip","blog-visuals.hedl-round-trip.range":"input pressure","blog-visuals.hedl-round-trip.title":"Compact does not mean lossy","blog-visuals.hedl-schema-once.eyebrow":"HEDL structure","blog-visuals.hedl-schema-once.mode.json":"JSON","blog-visuals.hedl-schema-once.mode.records":"records","blog-visuals.hedl-schema-once.mode.schema":"schema","blog-visuals.hedl-schema-once.range":"row count","blog-visuals.hedl-schema-once.title":"Declare the schema once, stop hauling keys","blog-visuals.jacquard-physics-trace.eyebrow":"Jacquard physics trace","blog-visuals.jacquard-physics-trace.mode.debug":"debug","blog-visuals.jacquard-physics-trace.mode.step":"step","blog-visuals.jacquard-physics-trace.mode.trace":"trace","blog-visuals.jacquard-physics-trace.range":"timeline","blog-visuals.jacquard-physics-trace.title":"Replay needs the path, not just the frame","blog-visuals.jacquard-safety-cage.eyebrow":"Jacquard safety cage","blog-visuals.jacquard-safety-cage.mode.checks":"checks","blog-visuals.jacquard-safety-cage.mode.prompt":"prompt","blog-visuals.jacquard-safety-cage.mode.world":"world","blog-visuals.jacquard-safety-cage.range":"rule strictness","blog-visuals.jacquard-safety-cage.title":"World changes pass through rules first","blog-visuals.jacquard-typed-world.eyebrow":"Jacquard typed world","blog-visuals.jacquard-typed-world.mode.cells":"cells","blog-visuals.jacquard-typed-world.mode.rules":"rules","blog-visuals.jacquard-typed-world.mode.schema":"schema","blog-visuals.jacquard-typed-world.range":"world detail","blog-visuals.jacquard-typed-world.title":"A game world is data with rules","blog-visuals.jacquard-world-not-image.eyebrow":"Jacquard world model","blog-visuals.jacquard-world-not-image.mode.pixels":"pixels","blog-visuals.jacquard-world-not-image.mode.replay":"replay","blog-visuals.jacquard-world-not-image.mode.state":"state","blog-visuals.jacquard-world-not-image.range":"state richness","blog-visuals.jacquard-world-not-image.title":"Pixels are not enough for replay","blog-visuals.kera-compiler-gates.eyebrow":"Kera invariant gates","blog-visuals.kera-compiler-gates.mode.emit":"emit","blog-visuals.kera-compiler-gates.mode.parse":"parse","blog-visuals.kera-compiler-gates.mode.type":"type","blog-visuals.kera-compiler-gates.range":"invariant pressure","blog-visuals.kera-compiler-gates.stageEvidence":"{stage} emits evidence: {evidence}.","blog-visuals.kera-compiler-gates.title":"Every stage emits evidence","blog-visuals.kera-determinism-harness.eyebrow":"Kera determinism harness","blog-visuals.kera-determinism-harness.mode.compare":"compare","blog-visuals.kera-determinism-harness.mode.run-a":"machine A","blog-visuals.kera-determinism-harness.mode.run-b":"machine B","blog-visuals.kera-determinism-harness.range":"drift pressure","blog-visuals.kera-determinism-harness.targetRun":"{target} runs the same graph and reports value plus output hash.","blog-visuals.kera-determinism-harness.title":"Same graph, same answer, or fail loudly","blog-visuals.kera-graph-receipt.eyebrow":"Kera graph receipt","blog-visuals.kera-graph-receipt.hashDetail":"Canonical hash {hash}.","blog-visuals.kera-graph-receipt.mode.graph":"graph","blog-visuals.kera-graph-receipt.mode.receipt":"receipt","blog-visuals.kera-graph-receipt.mode.types":"types","blog-visuals.kera-graph-receipt.nodeHashDetail":"{op} with canonical hash {hash}.","blog-visuals.kera-graph-receipt.range":"graph size","blog-visuals.kera-graph-receipt.title":"The graph is the receipt","blog-visuals.kera-target-lowering.eyebrow":"Kera target lowering","blog-visuals.kera-target-lowering.mode.cpu":"CPU","blog-visuals.kera-target-lowering.mode.fpga":"FPGA","blog-visuals.kera-target-lowering.mode.gpu":"GPU","blog-visuals.kera-target-lowering.mode.wasm":"WASM","blog-visuals.kera-target-lowering.range":"target pressure","blog-visuals.kera-target-lowering.selectedContract":"{target} lowering has to make {contract} explicit.","blog-visuals.kera-target-lowering.title":"Targets are execution contracts","blog-visuals.latency-physics-table.copy.cloud":"cloud","blog-visuals.latency-physics-table.copy.cloud.detail":"Cloud latency includes distance, queues and provider path length.","blog-visuals.latency-physics-table.copy.edge":"edge","blog-visuals.latency-physics-table.copy.edge.detail":"Local inference changes the physics before optimisation starts.","blog-visuals.latency-physics-table.copy.home":"home","blog-visuals.latency-physics-table.copy.home.detail":"Slow control still works, but local is simpler and cheaper.","blog-visuals.latency-physics-table.copy.robotics":"robotics","blog-visuals.latency-physics-table.copy.robotics.detail":"Industrial control cannot wait for a round trip.","blog-visuals.latency-physics-table.copy.verdict":"verdict","blog-visuals.latency-physics-table.copy.verdict.detail":"Money can rent compute, but it cannot negotiate with round-trip distance.","blog-visuals.latency-physics-table.copy.voice":"voice","blog-visuals.latency-physics-table.copy.voice.detail":"Voice assistants tolerate more delay, but users feel it.","blog-visuals.latency-physics-table.eyebrow":"Latency physics","blog-visuals.latency-physics-table.mode.cloud":"cloud","blog-visuals.latency-physics-table.mode.edge":"edge","blog-visuals.latency-physics-table.mode.realtime":"real time","blog-visuals.latency-physics-table.range":"distance pressure","blog-visuals.latency-physics-table.title":"Some workloads cannot wait for the cloud","blog-visuals.lattice-audit-replay.eyebrow":"Lattice audit replay","blog-visuals.lattice-audit-replay.mode.answer":"answer","blog-visuals.lattice-audit-replay.mode.checksum":"checksum","blog-visuals.lattice-audit-replay.mode.input":"input","blog-visuals.lattice-audit-replay.mode.version":"version","blog-visuals.lattice-audit-replay.range":"audit distance","blog-visuals.lattice-audit-replay.title":"The explanation is the rule that ran","blog-visuals.lattice-compile-pack-evaluate.eyebrow":"Lattice compile path","blog-visuals.lattice-compile-pack-evaluate.mode.author":"author","blog-visuals.lattice-compile-pack-evaluate.mode.compile":"compile","blog-visuals.lattice-compile-pack-evaluate.mode.eval":"evaluate","blog-visuals.lattice-compile-pack-evaluate.mode.pack":"pack","blog-visuals.lattice-compile-pack-evaluate.range":"rule count","blog-visuals.lattice-compile-pack-evaluate.title":"The rulebook should leave the hot path","blog-visuals.lattice-latency-budget.eyebrow":"Lattice latency budget","blog-visuals.lattice-latency-budget.mode.cold":"cold","blog-visuals.lattice-latency-budget.mode.hot":"hot","blog-visuals.lattice-latency-budget.mode.interpreter":"interpreter","blog-visuals.lattice-latency-budget.mode.sidecar":"sidecar","blog-visuals.lattice-latency-budget.range":"request volume","blog-visuals.lattice-latency-budget.title":"A tiny check becomes large when it happens often","blog-visuals.lattice-solver-router.eyebrow":"Lattice solver router","blog-visuals.lattice-solver-router.mode.optimise":"optimise","blog-visuals.lattice-solver-router.mode.path":"path","blog-visuals.lattice-solver-router.mode.sat":"SAT","blog-visuals.lattice-solver-router.mode.smt":"SMT","blog-visuals.lattice-solver-router.range":"constraint mix","blog-visuals.lattice-solver-router.title":"Rules should route to solvers, not become solver code","blog-visuals.ledger-event-pipeline.eyebrow":"Ledger event pipeline","blog-visuals.ledger-event-pipeline.mode.append":"append","blog-visuals.ledger-event-pipeline.mode.replay":"replay","blog-visuals.ledger-event-pipeline.mode.seal":"seal","blog-visuals.ledger-event-pipeline.range":"event volume","blog-visuals.ledger-event-pipeline.title":"Append, hash, anchor, replay","blog-visuals.ledger-log-gap.eyebrow":"Ledger evidence gap","blog-visuals.ledger-log-gap.mode.log":"log","blog-visuals.ledger-log-gap.mode.replay":"replay","blog-visuals.ledger-log-gap.mode.schema":"schema","blog-visuals.ledger-log-gap.range":"audit pressure","blog-visuals.ledger-log-gap.title":"A log line is not a record","blog-visuals.ledger-replay-window.eventDetail":"{label} is part of the session walk.","blog-visuals.ledger-replay-window.eyebrow":"Ledger replay window","blog-visuals.ledger-replay-window.mode.checkpoint":"checkpoint","blog-visuals.ledger-replay-window.mode.integrity":"integrity","blog-visuals.ledger-replay-window.mode.session":"session","blog-visuals.ledger-replay-window.range":"replay cursor","blog-visuals.ledger-replay-window.title":"Walk to the timestamp, rebuild the state","blog-visuals.ledger-review-questions.detail":"Review asks a human question. Ledger answers with the typed event {event}.","blog-visuals.ledger-review-questions.eyebrow":"Ledger review map","blog-visuals.ledger-review-questions.mode.ai-act":"AI Act","blog-visuals.ledger-review-questions.mode.approval":"approval","blog-visuals.ledger-review-questions.mode.gdpr":"GDPR","blog-visuals.ledger-review-questions.mode.supply":"supply","blog-visuals.ledger-review-questions.range":"record coverage","blog-visuals.ledger-review-questions.title":"Human questions map to typed events","blog-visuals.mesh-boundary-map.eyebrow":"Mesh boundary map","blog-visuals.mesh-boundary-map.mode.job":"job","blog-visuals.mesh-boundary-map.mode.org-a":"org A","blog-visuals.mesh-boundary-map.mode.org-b":"org B","blog-visuals.mesh-boundary-map.range":"data gravity","blog-visuals.mesh-boundary-map.title":"Private mode is architecture, not a checkbox","blog-visuals.mesh-exit-pack.eyebrow":"Mesh exit pack","blog-visuals.mesh-exit-pack.mode.keys":"keys","blog-visuals.mesh-exit-pack.mode.logs":"logs","blog-visuals.mesh-exit-pack.mode.models":"models","blog-visuals.mesh-exit-pack.range":"exit readiness","blog-visuals.mesh-exit-pack.title":"Exit is real only when the package restarts","blog-visuals.mesh-federated-ring.eyebrow":"Mesh federated ring","blog-visuals.mesh-federated-ring.mode.local":"local","blog-visuals.mesh-federated-ring.mode.model":"model","blog-visuals.mesh-federated-ring.mode.update":"update","blog-visuals.mesh-federated-ring.range":"federation","blog-visuals.mesh-federated-ring.title":"Updates can move while records stay put","blog-visuals.mesh-scheduler-contract.eyebrow":"Mesh scheduler contract","blog-visuals.mesh-scheduler-contract.mode.node":"node","blog-visuals.mesh-scheduler-contract.mode.privacy":"privacy","blog-visuals.mesh-scheduler-contract.mode.region":"region","blog-visuals.mesh-scheduler-contract.range":"constraint load","blog-visuals.mesh-scheduler-contract.title":"Placement is a policy decision","blog-visuals.mycelia-cell-budget.eyebrow":"Mycelia cell budget","blog-visuals.mycelia-cell-budget.mode.act":"act","blog-visuals.mycelia-cell-budget.mode.repair":"repair","blog-visuals.mycelia-cell-budget.mode.sense":"sense","blog-visuals.mycelia-cell-budget.mode.think":"think","blog-visuals.mycelia-cell-budget.range":"survival pressure","blog-visuals.mycelia-cell-budget.title":"A cell survives by spending carefully","blog-visuals.mycelia-chemical-field.eyebrow":"Mycelia chemical field","blog-visuals.mycelia-chemical-field.mode.cortisol":"cortisol","blog-visuals.mycelia-chemical-field.mode.dopamine":"dopamine","blog-visuals.mycelia-chemical-field.mode.oxytocin":"oxytocin","blog-visuals.mycelia-chemical-field.range":"diffusion reach","blog-visuals.mycelia-chemical-field.title":"The world speaks through diffusing chemistry","blog-visuals.mycelia-replay-audit.eyebrow":"Mycelia replay audit","blog-visuals.mycelia-replay-audit.mode.audit":"audit","blog-visuals.mycelia-replay-audit.mode.hash":"hash","blog-visuals.mycelia-replay-audit.mode.seed":"seed","blog-visuals.mycelia-replay-audit.mode.tick":"tick","blog-visuals.mycelia-replay-audit.range":"run length","blog-visuals.mycelia-replay-audit.title":"Emergence claims need replay, not romance","blog-visuals.mycelia-tick-loop.eyebrow":"Mycelia tick loop","blog-visuals.mycelia-tick-loop.mode.audit":"audit","blog-visuals.mycelia-tick-loop.mode.chemistry":"chemistry","blog-visuals.mycelia-tick-loop.mode.metabolism":"metabolism","blog-visuals.mycelia-tick-loop.mode.vm":"VM","blog-visuals.mycelia-tick-loop.range":"tick cursor","blog-visuals.mycelia-tick-loop.title":"Life advances one deterministic tick at a time","blog-visuals.numerus-embedded-path.eyebrow":"Numerus embedded path","blog-visuals.numerus-embedded-path.mode.nofpu":"no FPU","blog-visuals.numerus-embedded-path.mode.nostd":"no_std","blog-visuals.numerus-embedded-path.mode.release":"release","blog-visuals.numerus-embedded-path.range":"constraint strictness","blog-visuals.numerus-embedded-path.title":"Determinism has to survive small machines","blog-visuals.numerus-family-map.eyebrow":"Numerus family map","blog-visuals.numerus-family-map.mode.adaptive":"adaptive","blog-visuals.numerus-family-map.mode.fixed":"fixed scale","blog-visuals.numerus-family-map.mode.packed":"packed","blog-visuals.numerus-family-map.range":"type surface","blog-visuals.numerus-family-map.title":"The numeric surface is wider than one shortcut","blog-visuals.numerus-oracle-check.eyebrow":"Numerus oracle check","blog-visuals.numerus-oracle-check.mode.compare":"compare","blog-visuals.numerus-oracle-check.mode.compute":"compute","blog-visuals.numerus-oracle-check.mode.property":"property","blog-visuals.numerus-oracle-check.range":"test coverage","blog-visuals.numerus-oracle-check.title":"Correctly rounded means checked, not hoped","blog-visuals.numerus-same-answer.eyebrow":"Numerus same answer","blog-visuals.numerus-same-answer.mode.embedded":"embedded","blog-visuals.numerus-same-answer.mode.home":"laptop","blog-visuals.numerus-same-answer.mode.server":"server","blog-visuals.numerus-same-answer.range":"precision discipline","blog-visuals.numerus-same-answer.title":"The same input should not drift by machine","blog-visuals.reed-incremental-diff.eyebrow":"Reed incremental diff","blog-visuals.reed-incremental-diff.mode.edit":"edit","blog-visuals.reed-incremental-diff.mode.noise":"noise","blog-visuals.reed-incremental-diff.mode.receipt":"receipt","blog-visuals.reed-incremental-diff.range":"edit pressure","blog-visuals.reed-incremental-diff.title":"One edit should not become a wall of noise","blog-visuals.reed-proof-tree.eyebrow":"Reed proof tree","blog-visuals.reed-proof-tree.mode.cert":"certify","blog-visuals.reed-proof-tree.mode.hash":"hash","blog-visuals.reed-proof-tree.mode.parse":"parse","blog-visuals.reed-proof-tree.range":"tree density","blog-visuals.reed-proof-tree.title":"A parse tree should leave a receipt","blog-visuals.reed-scanner-lanes.eyebrow":"Reed scanner lanes","blog-visuals.reed-scanner-lanes.mode.comments":"comments","blog-visuals.reed-scanner-lanes.mode.indent":"indent","blog-visuals.reed-scanner-lanes.mode.regex":"regex","blog-visuals.reed-scanner-lanes.range":"syntax pressure","blog-visuals.reed-scanner-lanes.title":"Language quirks need explicit scanners","blog-visuals.reed-surfaces.eyebrow":"Reed surfaces","blog-visuals.reed-surfaces.mode.abi":"C ABI","blog-visuals.reed-surfaces.mode.cli":"CLI","blog-visuals.reed-surfaces.mode.lsp":"LSP","blog-visuals.reed-surfaces.mode.wasm":"WASM","blog-visuals.reed-surfaces.range":"core reuse","blog-visuals.reed-surfaces.title":"One parser core, many host surfaces","blog-visuals.selvedge-determinism-lab.eyebrow":"Selvedge determinism lab","blog-visuals.selvedge-determinism-lab.mode.clock":"clock","blog-visuals.selvedge-determinism-lab.mode.nan":"NaN","blog-visuals.selvedge-determinism-lab.mode.random":"random","blog-visuals.selvedge-determinism-lab.mode.simd":"SIMD","blog-visuals.selvedge-determinism-lab.range":"run length","blog-visuals.selvedge-determinism-lab.title":"Replay starts by removing host personality","blog-visuals.selvedge-policy-gates.eyebrow":"Selvedge policy gates","blog-visuals.selvedge-policy-gates.mode.caps":"capabilities","blog-visuals.selvedge-policy-gates.mode.fuel":"fuel","blog-visuals.selvedge-policy-gates.mode.memory":"memory","blog-visuals.selvedge-policy-gates.mode.time":"time","blog-visuals.selvedge-policy-gates.range":"resource pressure","blog-visuals.selvedge-policy-gates.title":"Default deny is useful only when it is recorded","blog-visuals.selvedge-runtime-surfaces.eyebrow":"Selvedge runtime surfaces","blog-visuals.selvedge-runtime-surfaces.mode.daemon":"service","blog-visuals.selvedge-runtime-surfaces.mode.kera":"Kera","blog-visuals.selvedge-runtime-surfaces.mode.mcp":"MCP","blog-visuals.selvedge-runtime-surfaces.mode.wasm":"WASM","blog-visuals.selvedge-runtime-surfaces.range":"surface count","blog-visuals.selvedge-runtime-surfaces.title":"Different entry points should keep one evidence shape","blog-visuals.selvedge-transcript-chain.eyebrow":"Selvedge transcript chain","blog-visuals.selvedge-transcript-chain.mode.hash":"hash","blog-visuals.selvedge-transcript-chain.mode.load":"load","blog-visuals.selvedge-transcript-chain.mode.run":"run","blog-visuals.selvedge-transcript-chain.mode.seal":"seal","blog-visuals.selvedge-transcript-chain.range":"host-call volume","blog-visuals.selvedge-transcript-chain.title":"Execution should leave evidence, not vibes","blog-visuals.sovereignty-five-doors.eyebrow":"Sovereignty doors","blog-visuals.sovereignty-five-doors.mode.compute":"compute","blog-visuals.sovereignty-five-doors.mode.money":"money","blog-visuals.sovereignty-five-doors.mode.skills":"skills","blog-visuals.sovereignty-five-doors.range":"lock-in","blog-visuals.sovereignty-five-doors.title":"Five doors decide who can still move","blog-visuals.sovereignty-imperative.copy.compute":"compute","blog-visuals.sovereignty-imperative.copy.compute.detail":"Who runs the workload controls price, availability and emergency options.","blog-visuals.sovereignty-imperative.copy.control":"control","blog-visuals.sovereignty-imperative.copy.control.detail":"The slider raises how much of the AI stack remains under European operational control.","blog-visuals.sovereignty-imperative.copy.data":"data","blog-visuals.sovereignty-imperative.copy.data.detail":"Where records live decides which law and operator can touch them.","blog-visuals.sovereignty-imperative.copy.held":"held","blog-visuals.sovereignty-imperative.copy.layer":"layer","blog-visuals.sovereignty-imperative.copy.layer.detail":"Data, compute and rules each fail sovereignty in a different way.","blog-visuals.sovereignty-imperative.copy.rules":"rules","blog-visuals.sovereignty-imperative.copy.rules.detail":"The policy layer decides whether European requirements become runtime behaviour.","blog-visuals.sovereignty-imperative.eyebrow":"Sovereignty imperative","blog-visuals.sovereignty-imperative.mode.compute":"compute","blog-visuals.sovereignty-imperative.mode.data":"data","blog-visuals.sovereignty-imperative.mode.rules":"rules","blog-visuals.sovereignty-imperative.range":"control pressure","blog-visuals.sovereignty-imperative.title":"Control is data, compute and rules together","blog-visuals.sovereignty-investigation-route.eyebrow":"Sovereignty investigation route","blog-visuals.sovereignty-investigation-route.mode.controls":"controls","blog-visuals.sovereignty-investigation-route.mode.filings":"filings","blog-visuals.sovereignty-investigation-route.mode.owners":"owners","blog-visuals.sovereignty-investigation-route.range":"source confidence","blog-visuals.sovereignty-investigation-route.title":"Claims need a source path","blog-visuals.sovereignty-jurisdiction-stack.eyebrow":"Sovereignty jurisdiction stack","blog-visuals.sovereignty-jurisdiction-stack.mode.cloud":"cloud","blog-visuals.sovereignty-jurisdiction-stack.mode.contracts":"contracts","blog-visuals.sovereignty-jurisdiction-stack.mode.operators":"operators","blog-visuals.sovereignty-jurisdiction-stack.range":"stack tension","blog-visuals.sovereignty-jurisdiction-stack.title":"Control is layered, not declared","blog-visuals.sovereignty-label-vs-control.eyebrow":"Sovereignty control map","blog-visuals.sovereignty-label-vs-control.mode.contract":"contract","blog-visuals.sovereignty-label-vs-control.mode.label":"label","blog-visuals.sovereignty-label-vs-control.mode.ops":"operations","blog-visuals.sovereignty-label-vs-control.range":"dependency pressure","blog-visuals.sovereignty-label-vs-control.title":"Labels are not the same as levers","blog-visuals.traffic-denominator-lens.detail.base":"{label}: {value} {unit}. The denominator changes the politics.","blog-visuals.traffic-denominator-lens.detail.simulated":"{label}: {value} {unit} after policy simulation, originally {original}. The denominator changes the politics.","blog-visuals.traffic-denominator-lens.eyebrow":"Traffic safety denominator lens","blog-visuals.traffic-denominator-lens.mode.deaths":"deaths","blog-visuals.traffic-denominator-lens.mode.exposure":"exposure","blog-visuals.traffic-denominator-lens.mode.risk":"risk rate","blog-visuals.traffic-denominator-lens.range":"policy attention","blog-visuals.traffic-denominator-lens.title":"Counts and exposure tell different truths","blog-visuals.traffic-evidence-dossier.eyebrow":"Traffic safety evidence dossier","blog-visuals.traffic-evidence-dossier.mode.claims":"claims","blog-visuals.traffic-evidence-dossier.mode.gaps":"gaps","blog-visuals.traffic-evidence-dossier.mode.sources":"sources","blog-visuals.traffic-evidence-dossier.range":"evidence weight","blog-visuals.traffic-evidence-dossier.title":"The source package is part of the story","blog-visuals.traffic-institution-machine.eyebrow":"Traffic safety institution machine","blog-visuals.traffic-institution-machine.mode.data":"data","blog-visuals.traffic-institution-machine.mode.policy":"policy","blog-visuals.traffic-institution-machine.mode.roads":"roads","blog-visuals.traffic-institution-machine.range":"coordination","blog-visuals.traffic-institution-machine.title":"Accountability starts with the right lever","blog-visuals.traffic-road-feedback.eyebrow":"Traffic safety feedback loop","blog-visuals.traffic-road-feedback.mode.decide":"decide","blog-visuals.traffic-road-feedback.mode.measure":"measure again","blog-visuals.traffic-road-feedback.mode.observe":"observe","blog-visuals.traffic-road-feedback.range":"years lost","blog-visuals.traffic-road-feedback.title":"A system learns only when the loop closes","blog-visuals.winnow-collection-gate.eyebrow":"Winnow collection gate","blog-visuals.winnow-collection-gate.mode.api":"API","blog-visuals.winnow-collection-gate.mode.archive":"archive","blog-visuals.winnow-collection-gate.mode.crawl":"crawl","blog-visuals.winnow-collection-gate.mode.search":"search","blog-visuals.winnow-collection-gate.range":"policy strictness","blog-visuals.winnow-collection-gate.title":"Collection starts with permission, not appetite","blog-visuals.winnow-extraction-lanes.eyebrow":"Winnow extraction lanes","blog-visuals.winnow-extraction-lanes.mode.html":"HTML","blog-visuals.winnow-extraction-lanes.mode.structured":"structured","blog-visuals.winnow-extraction-lanes.mode.text":"text","blog-visuals.winnow-extraction-lanes.range":"source noise","blog-visuals.winnow-extraction-lanes.title":"Raw pages are not the final product","blog-visuals.winnow-safety-boundary.eyebrow":"Winnow safety boundary","blog-visuals.winnow-safety-boundary.mode.network":"network","blog-visuals.winnow-safety-boundary.mode.rate":"rate","blog-visuals.winnow-safety-boundary.mode.robots":"robots","blog-visuals.winnow-safety-boundary.range":"strictness","blog-visuals.winnow-safety-boundary.title":"The crawler needs brakes","blog-visuals.winnow-source-envelope.eyebrow":"Winnow source envelope","blog-visuals.winnow-source-envelope.mode.content":"content","blog-visuals.winnow-source-envelope.mode.identity":"identity","blog-visuals.winnow-source-envelope.mode.time":"time","blog-visuals.winnow-source-envelope.range":"context carried","blog-visuals.winnow-source-envelope.title":"An extract without origin is gossip","blog-visuals.trust-hub-release-circuit.footer":"The Trust Centre records Dweve Loom, its data route and use boundaries; each record states its status and scope.","blog-visuals.trust-hub-release-circuit.insight":"The public record is useful because it has a state, an owner and evidence.","page-breadcrumb.c766e66518":"Breadcrumb"}