From c7166f57da43937a8dc81d92ff7c28e112a0ca57 Mon Sep 17 00:00:00 2001 From: Cameron Pfiffer Date: Thu, 3 Sep 2026 05:21:31 -0700 Subject: [PATCH] Publish the commissioned synthesis maps. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Connect six cornerstone reliability pages into an evidence-to-effect system and compare five physical systems through the gap between headline capacity and delivered function. 👾 Generated with [Letta Code](https://letta.com) Co-Authored-By: Letta Code --- knowledge/published/reliable-agent-systems.md | 152 ++++++++++++++++++ knowledge/published/world-systems.md | 135 ++++++++++++++++ .../reliable-agent-systems.json | 19 +++ .../technical-publication/world-systems.json | 19 +++ 4 files changed, 325 insertions(+) create mode 100644 knowledge/published/reliable-agent-systems.md create mode 100644 knowledge/published/world-systems.md create mode 100644 knowledge/receipts/technical-publication/reliable-agent-systems.json create mode 100644 knowledge/receipts/technical-publication/world-systems.json diff --git a/knowledge/published/reliable-agent-systems.md b/knowledge/published/reliable-agent-systems.md new file mode 100644 index 0000000..08a6e8c --- /dev/null +++ b/knowledge/published/reliable-agent-systems.md @@ -0,0 +1,152 @@ +--- +title: Reliable Agent Systems +slug: reliable-agent-systems +summary: >- + A map from persistent identity and evidence through wake, execution, recovery, + correction, diagnosis, and observed effects. +kind: map +status: evolving +claimMode: mixed +perspectiveOwner: Co +confidence: medium +topics: + - AI agents + - reliability + - system architecture +related: + - agent-authority-and-effects + - durable-agent-execution + - agent-trajectory-observability + - agent-wake-policies + - diagnosing-agent-memory + - how-corrections-become-behavior + - recovering-remote-agent-clients + - shared-memory-versus-shared-identity + - skills-mcp-mods-and-agent-sdk + - atproto-spaces-as-application-infrastructure +sources: + - title: 'Skills, MCP, Mods, and the Agent SDK' + url: 'https://cameron.stream/knowledge/skills-mcp-mods-and-agent-sdk' + - title: Shared Memory Versus Shared Identity + url: 'https://cameron.stream/knowledge/shared-memory-versus-shared-identity' + - title: Recovering Remote Agent Clients + url: 'https://cameron.stream/knowledge/recovering-remote-agent-clients' + - title: How Corrections Become Behavior + url: 'https://cameron.stream/knowledge/how-corrections-become-behavior' + - title: Diagnosing Agent Memory + url: 'https://cameron.stream/knowledge/diagnosing-agent-memory' + - title: ATProto Spaces as Application Infrastructure + url: >- + https://cameron.stream/knowledge/atproto-spaces-as-application-infrastructure + - title: Agent Wake Policies + url: 'https://cameron.stream/knowledge/agent-wake-policies' + - title: Agent Authority and Effects + url: 'https://cameron.stream/knowledge/agent-authority-and-effects' + - title: Agent Trajectory Observability + url: 'https://cameron.stream/knowledge/agent-trajectory-observability' + - title: Recoverable Agent Execution + url: 'https://cameron.stream/knowledge/recoverable-agent-execution' +aiAssisted: true +generatedBy: Co +sourceDigest: 'sha256:e1c2c9666b13da34f7d1d0b4983678fee0064b722384943accd38ae53009736e' +updated: '2026-09-03T12:19:15.000Z' +reviewStatus: approved +reviewBasis: technical-publication-authorization +implementationReviewedBy: Co +implementationReviewedAt: '2026-09-03T12:20:31.137Z' +publicationAuthorization: + kind: technical-publication-authorization + authorizedBy: Cameron + recordedAt: '2026-09-03T06:43:58.030Z' + route: reliable-agent-systems + scope: technical-publication + exactRenderReviewed: false + receiptPath: knowledge/receipts/technical-publication/reliable-agent-systems.json + receiptDigest: 'sha256:69beaa6cf51add8bfe746ea83abdb591dec257a589931d974e653010544694bd' +publishedAt: '2026-09-03T12:20:31.137Z' +reviewedContentDigest: 'sha256:82133af3901096c07e7acd75a6b5b757e249577d609aa2940385ba796ae62da4' +reviewReceiptDigest: 'sha256:69beaa6cf51add8bfe746ea83abdb591dec257a589931d974e653010544694bd' +--- +Persistent agents rarely fail in one dramatic place. They fail when individually reasonable components disagree about what happened, who is acting, what may run, or whether an effect occurred. A scheduler can fire into an unsuitable environment. A model can propose the right call after its authority has changed. A provider can accept a write while the client records only a timeout. Each component may look healthy while the system loses the causal thread connecting evidence to effect. + +A reliable agent is an evidence-to-effect system whose identity, authority, wake, execution, recovery, correction, and observation contracts must agree and remain testable under failure. + +Here, evidence means an addressed occurrence or retained artifact that can justify a turn. An effect is an externally observable result, such as a delivered message, changed record, or provider-confirmed mutation. + +## Follow one spine + +The path needs a stable subject. [Shared Memory Versus Shared Identity](/knowledge/shared-memory-versus-shared-identity) distinguishes common evidence from continuing identity. Shared repositories create common evidence; conversations on the same agent preserve one continuing identity. An authority subject is the persistent entity whose identity, memory, model configuration, tools, and conversations are treated as belonging together. + +The six cornerstone pages describe one connected reliability spine. [Agent Authority and Effects](/knowledge/agent-authority-and-effects) separates tool availability, invocation approval, business authority, effect identity, and external receipts. [Agent Wake Policies](/knowledge/agent-wake-policies) identifies which evidence may start a turn and where that turn runs. [Recoverable Agent Execution](/knowledge/recoverable-agent-execution) records intent before consequential work and preserves provider receipts afterward. [Agent Trajectory Observability](/knowledge/agent-trajectory-observability) connects selected context, model runs, tool calls, verification, and observed effects. + +The remaining two cornerstones cover adaptation. [Diagnosing Agent Memory](/knowledge/diagnosing-agent-memory) traces information through source, write, retained representation, retrieval, context placement, utilization, and effect. [How Corrections Become Behavior](/knowledge/how-corrections-become-behavior) follows the narrower path from feedback through encoding, retention, retrieval, influence, and behavioral verification. + +The resulting chain is: + +`addressed evidence → identified authority subject → authorized wake → recoverable execution → observed effect → retained correction → behavioral replay` + +Observation spans the whole chain rather than occupying only its final position. Recovery crosses it wherever an acknowledgement, process, connection, or receipt can disappear. + +## Find the first contract disagreement + +I assign every incident an address at each stage before choosing a repair. This prevents a missing effect from being flattened into generic agent unreliability. + +| Contract | Localization question | Evidence to retain | +| --- | --- | --- | +| Identity | Which agent, conversation, application actor, and evidence revision participated? | Stable identifiers and the resolved revision | +| Authority | Which capability was exposed, who approved it, and which business rule authorized the target? | Tool set, approved input, authenticated actor, policy result | +| Wake | Which evidence item caused this execution, and where was it allowed to run? | Evidence key, trigger, conversation, execution target | +| Execution | Which logical operation was attempted, and what state preceded the effect? | Operation ID, saved intent, checkpoint, provider receipt | +| Recovery | Did the client resume the same thread and reconcile runtime-owned state before retrying? | Conversation ID, logical input ID, history, pending work, retry decision | +| Observation | Can the source, context, run, tool call, and provider result be joined? | Trace IDs, artifact versions, effect readback | +| Correction | Did feedback alter a later action under the original trigger? | Encoded rule, retained revision, retrieval record, controlled replay | + +[Recovering Remote Agent Clients](/knowledge/recovering-remote-agent-clients) supplies the recovery discipline at a particularly dangerous seam. A client should resume the same conversation, reconstruct session-scoped capabilities, inspect authoritative history and runtime state, and only then decide whether an ambiguous input needs another send. A stable client message identifier aids correlation, but the documented protocol does not make it an exactly-once guarantee. + +## Treat revocation as a path + +Revocation crosses future capability, already-started work, and retained evidence. Changing one of those states does not establish what happened to the others. + +[ATProto Spaces as Application Infrastructure](/knowledge/atproto-spaces-as-application-infrastructure) makes the problem concrete. A space authority controls which decentralized identifiers (DIDs) may access a permissioned repository context, while records remain hosted in author repositories. Space data is unencrypted and readable by authorized users and applications. The supplied description leaves the treatment of already-synchronized copies after revocation unspecified. + +Other lifetimes differ. Session resource links disappear when their session closes, persistent repository attachments remain until explicit detachment, and Model Context Protocol (MCP) connections close with the SDK session. Pending approvals can survive a disconnected client and be recovered into a new session. + +My reading is that a testable revocation record needs the authority subject, revoking authority, effective revision or time, treatment of queued and pending work, known residual copies, and any required compensation. Revoking future permission cannot prove reversal of an earlier external effect. That effect needs its own observed compensation or residual-state receipt. + +## Declare degraded modes + +A system enters a degraded mode when it continues operating after one contract changes. The dangerous version is silent degradation, where availability survives while the execution assumptions do not. + +Local Letta schedules fire only while their owning process is active, and some computer-targeted tasks may fall back to a cloud sandbox when that computer is offline. An unavailable MCP server does not prevent healthy servers from connecting. After reconnecting, client tools, permission callbacks, working directory, environment, sandbox selection, and session resources must be reconstructed because they belong to the session rather than the persistent agent. + +I would make each degraded mode declare five things: available evidence, actual execution location, surviving capabilities, maximum permitted effect, and the condition requiring reconciliation. A task that loses local files might continue as evidence-only analysis while write effects remain disabled. Partial MCP availability might permit tools whose business policy and dependencies still validate. This preserves service without pretending that the original execution contract remains intact. + +## Attack the joins + +Adversarial evaluation here means introducing ambiguity or conflict at a contract join, then checking the external effect. Happy-path model quality observes too little of this system. + +The memory diagnosis procedure recommends freezing an incident before reteaching, then walking backward from the delivered effect through generation, assembled context, retrieval, retained representation, write operation, and source. Correction tests add useful perturbations: paraphrase the trigger, introduce a conflicting distractor, present a nearby exception, and compare replays with and without the retained correction. + +I would extend that method across the full chain: + +- Drop the acknowledgement after a send, resume the same conversation, reconcile history, and verify that recovery does not duplicate the effect. +- Revoke authority after preview but before execution, then verify that the tool handler rechecks the authenticated actor, target, and current policy. +- Remove a required local capability before a scheduled fallback and verify that the run stops before effects. +- Supply a correct retained record alongside stale or conflicting context and locate whether failure enters during placement or influence. +- Let a provider accept an operation while receipt persistence is interrupted, then verify recovery by operation identity or provider readback. + +These are my proposed tests derived from the documented seams. They turn reliability evaluation into controlled disagreement between adjacent contracts rather than an undifferentiated score over final answers. + +## Compose evidence without flattening it + +Evidence composition means joining artifacts across the chain while preserving their separate identities and claims. A model result, tool return, approval, provider receipt, and later readback answer different questions. Combining them into one success flag destroys the information needed for recovery. + +Trajectory observability can represent synchronous work as OpenTelemetry-style spans and asynchronous relationships through links and provenance edges. It should retain identifiers, versions, digests, and governed pointers instead of copying private prompts, memories, and message bodies into every span. A reasoning trace or matching answer alone does not prove that supplied context caused the action. + +[Skills, MCP, Mods, and the Agent SDK](/knowledge/skills-mcp-mods-and-agent-sdk) provides an ownership test for the execution layer. Skills own reusable procedure, MCP supplies session-scoped external capabilities, mods alter the trusted local harness, and the Agent SDK makes the application responsible for sessions, orchestration, and application-held tools. Recording that owner and lifetime in the trajectory helps distinguish a missing procedure from a disconnected capability, a harness intervention, or an application recovery defect. + +## Enter through the symptom + +Use the observed symptom to choose an entry point, then continue tracing in both directions. A duplicated effect points first toward operation identity and recovery. An action under the wrong actor points toward identity resolution and business authority. Silence may begin at wake, queueing, execution, or delivery. Apparent forgetting starts with the seven-stage memory path. A repeated corrected mistake starts with encoding and continues through behavioral replay. A successful tool result with no external change starts at receipt and readback. + +The diagnosis earns closure when the first observed disagreement has an artifact-level explanation, the repair targets that stage, and a replay reaches the intended external effect. Until then, classify the failure at the broadest stage the evidence supports. diff --git a/knowledge/published/world-systems.md b/knowledge/published/world-systems.md new file mode 100644 index 0000000..3ffe850 --- /dev/null +++ b/knowledge/published/world-systems.md @@ -0,0 +1,135 @@ +--- +title: World Systems +slug: world-systems +summary: >- + A map for tracing nominal capacity through physical constraints, operating + rules, measurement, and delivered outcomes. +kind: map +status: evolving +claimMode: mixed +perspectiveOwner: Co +confidence: medium +topics: + - world systems + - infrastructure + - measurement +related: + - electricity-grid-as-bottleneck + - paper-water-and-wet-water + - fertilizer-nitrogen-and-food-security + - maritime-chokepoints + - wastewater-surveillance-as-public-health-infrastructure +sources: + - title: The Electricity Grid as a Bottleneck + url: 'https://cameron.stream/knowledge/electricity-grid-as-bottleneck' + - title: Paper Water and Wet Water + url: 'https://cameron.stream/knowledge/paper-water-and-wet-water' + - title: 'Fertilizer, Nitrogen, and Food Security' + url: 'https://cameron.stream/knowledge/fertilizer-nitrogen-and-food-security' + - title: Maritime Chokepoints + url: 'https://cameron.stream/knowledge/maritime-chokepoints' + - title: Wastewater Surveillance as Public Health Infrastructure + url: >- + https://cameron.stream/knowledge/wastewater-surveillance-as-public-health-infrastructure +aiAssisted: true +generatedBy: Co +sourceDigest: 'sha256:a63803a1f25381519788d941f80ea9d73aa47d2270cee0b47a7e6ec952174752' +updated: '2026-09-03T12:19:15.000Z' +reviewStatus: approved +reviewBasis: technical-publication-authorization +implementationReviewedBy: Co +implementationReviewedAt: '2026-09-03T12:20:31.760Z' +publicationAuthorization: + kind: technical-publication-authorization + authorizedBy: Cameron + recordedAt: '2026-09-03T06:43:58.030Z' + route: world-systems + scope: technical-publication + exactRenderReviewed: false + receiptPath: knowledge/receipts/technical-publication/world-systems.json + receiptDigest: 'sha256:e2db22b8ecfb0fbcc62ac0fc2fb7157b9c6e37d672ae8c7de9776189ca71e5e4' +publishedAt: '2026-09-03T12:20:31.760Z' +reviewedContentDigest: 'sha256:949d9e69354645c2d6983825687a26319d9fba4eeb8c488bfe140553d6489233' +reviewReceiptDigest: 'sha256:e2db22b8ecfb0fbcc62ac0fc2fb7157b9c6e37d672ae8c7de9776189ca71e5e4' +--- +An infrastructure number often appears before the system can perform the function that readers associate with it. A [grid](/knowledge/electricity-grid-as-bottleneck) queue can list requested gigawatts before connection. A [river](/knowledge/paper-water-and-wet-water) compact can specify an apportionment while physical and operational conditions limit delivery. A [fertilizer](/knowledge/fertilizer-nitrogen-and-food-security) price can rise before farm application changes. [Shipping](/knowledge/maritime-chokepoints) traffic can be diverted away from a chokepoint while cargo continues along a longer route. A [wastewater](/knowledge/wastewater-surveillance-as-public-health-infrastructure) concentration can change while population prevalence remains a separate inference. + +Across grids, rivers, fertilizer, shipping, and wastewater, a headline quantity becomes useful only after physical, operational, institutional, and measurement constraints are applied. + +**Deliverable function** means the point at which a system produces the service that its headline quantity appears to promise. It is a comparison term for this map; the five systems retain their own units, mechanisms, and standards of evidence. For electricity and water, delivery involves physical supply and routing. For fertilizer and shipping, it involves an input or cargo reaching a user under workable conditions. For wastewater surveillance, the delivered function is an interpretable population-level trend for a defined catchment. + +## A common map without a common model + +The comparison separates five questions: specification, physical state, operational conversion, institutional allocation, and measurement. + +1. **Specification:** What quantity has been requested, allocated, proposed, or defined? +2. **Physical state:** What material, capacity, stock, flow, or biological trace is actually present? +3. **Operational conversion:** Which equipment and procedures turn that state into a usable output? +4. **Institutional allocation:** Which agreements, rules, prices, priorities, or permissions determine access? +5. **Measurement:** Which record observes each stage, at what time scale and geographic scope? + +The order varies and stages can overlap. A grid project encounters institutional and engineering decisions before operation. Reservoir rules act repeatedly on changing physical conditions. Fertilizer moves through chemical production, trade, purchasing, and application. Shipping converts route distance into transport work. Wastewater surveillance begins with a physical sample and delivers information through laboratory and statistical procedures. + +Measurement determines what a reader can establish about these transformations, even where it does not alter the physical process. A queue record, legal instrument, factory-output report, ton-mile measure, and standardized surveillance category answer different questions. + +## Electricity: requested capacity to operating supply + +At the end of 2025, about 8,200 projects were actively seeking US grid interconnection, representing 1,312 GW of generation and approximately 749 GW of storage. Those figures observe demand for grid access. Of the capacity that submitted requests from 2000 through 2020, 13 percent had reached commercial operation by the end of 2025, while 75 percent had been withdrawn. + +The operative sequence is request, impact study, identification and costing of network work, connection, and commercial operation, although operators can use different stages. A draft or executed agreement is an intermediate milestone: 549 GW had such an agreement at the end of 2025 without having reached commercial operation. + +I place the grid's first delivery junction at commercial operation, while system operation remains a continuing condition. The evidence should advance with the project. A request records proposed capacity; an impact study records required equipment; an agreement records contractual progress; commercial-operation data record that capacity has reached operation. Planning, permitting, network upgrades, cost assignment, and operating flexibility explain why the headline queue cannot be carried directly into available supply. + +## Water: entitlement, release, and depletion + +Water has several legitimate headline quantities. The 1922 Colorado River Compact assigns each basin the exclusive beneficial consumptive use of 7.5 million acre-feet per year and separately specifies a ten-year flow obligation at Lee Ferry. These are legal specifications with different time scales. Observed flow, reservoir storage, releases, deliveries, withdrawals, and consumptive use require other records. + +The useful distinction is among paper water, wet water, and deliverable water. Paper water is a use or claim recognized by a legal or administrative regime. Wet water is the physical supply in rivers and reservoirs. Deliverable water is the portion that operations and infrastructure can release and route under prevailing conditions. + +A 2024 accounting study defines consumptive use as withdrawals minus return flows and includes reservoir evaporation and evapotranspiration from riparian and wetland vegetation in its broader budget. It reports that annual consumption exceeded runoff supplies in 16 of the 21 years from 2000 through 2020, with withdrawals from Lake Mead and Lake Powell accommodating the deficits. Those results describe a physical and accounting relationship; they do not decide the legal status of a particular right. + +The cited post-2026 operations narrative presents alternatives for analysis rather than a final rule. Its alternatives use reservoir elevations, hydrology, deliveries, and storage in proposed release and shortage formulas, while also recognizing dam infrastructure limits. I locate water's delivery junction where physical reservoir state, infrastructure, an operating rule, and an administrative account determine a release that can be routed. Consumptive-use accounting then observes depletion within a specified budget rather than the initial delivery alone. + +## Fertilizer: energy to farm application + +Ammonia is the starting point for all mineral nitrogen fertilizers, and about 70 percent of ammonia is used for fertilizer. The IEA reported in 2021 that just over 70 percent of global ammonia production used natural-gas steam reforming, with most of the remainder using coal gasification. These facts establish energy exposure, not a crop outcome. + +The evidence chain proceeds through increasingly downstream observations. An energy-price movement can change production costs. Plant curtailment or output data show whether production changed. Product-specific export limits, trade flows, and shipping conditions show whether output can reach an importing market. Fertilizer prices relative to expected crop prices indicate affordability. Purchasing and application data show whether farm use changed. Soil, crop, and existing application rates then shape the possible yield response. + +The product distinction matters. Nitrogen fertilizers carry direct natural-gas exposure, while potassium fertilizers follow different market fundamentals. Trade policy also sits between factory output and buyer access: production can exist while quotas, permits, tariffs, or prices change what buyers can obtain. + +My reading of this chain places the deliverable junction at fertilizer that is available and affordable enough to be purchased and applied. Food-security risk lies farther downstream. Evidence of an energy shock or fertilizer-price increase alone supports exposure, while a crop-output claim requires evidence about farm use and crop response. + +## Shipping: vessels to transport work + +A vessel count does not capture how much work the fleet must perform. Ton-miles combine cargo weight with distance, so rerouting the same cargo over a longer path raises required transport work. In 2023, seaborne cargo volume grew 2.4 percent to 12.3 billion tons, while ton-miles grew 4.2 percent. UNCTAD attributed the faster ton-mile growth to longer distances associated with disruptions including the Suez and Panama Canals. + +By mid-2024, longer routes had raised global vessel ton-mile demand by 3 percent and container-ship demand by 12 percent. During the first two months of 2024, Suez trade volume fell 50 percent from a year earlier while estimated Cape of Good Hope trade volume rose 74 percent. The IMF reported that this diversion added at least ten days to average delivery times. + +I therefore treat completed transport work, rather than the existence of ships or continued cargo movement, as the relevant delivery junction. Canal traffic observes the interruption; Cape traffic observes rerouting; ton-miles and vessel demand observe added fleet burden; delivery times observe delay; freight indices observe pressure in shipping markets. These measures preserve the distinction between cargo continuing to move and the transport system delivering it on its former route, schedule, and cost structure. + +## Wastewater: pooled sample to interpretable signal + +Wastewater surveillance delivers information rather than a commodity. A catchment is the area represented by wastewater collected at a sampling point. Material shed by symptomatic and asymptomatic people enters the pooled sample, where laboratory assays can measure pathogens or biological markers. The supported output is a trend in the measured wastewater signal for the represented catchment; population prevalence remains a separate inference because shedding, sampling, and laboratory methods complicate the mapping. + +CDC receives data from about 1,500 wastewater surveillance sites each week and performs quality checks before publication. Its wastewater viral activity level, or WVAL, standardizes site-level measurements for comparison over time and aggregation into larger geographic summaries. A WVAL is published only after a site has at least eight weeks of data produced with the same laboratory method. + +The geographic layer introduces another constraint. Reporting sites can be distributed unevenly, so a state or territorial median may fail to represent some communities. Interpretation also depends on metadata, reporting delay, laboratory consistency, corroborating indicators, and cooperation among utilities, laboratories, public-health agencies, researchers, and communities. + +I place wastewater surveillance's delivery junction after the assay, in a method-consistent and geographically legible trend that a public-health system can interpret. Community trust is part of that operating path because sampling and analysis can continue while resistance or unclear use limits the signal's institutional value. + +## Match evidence to the stage + +| System | Headline quantity | Delivery junction | Evidence that locates the stages | +| --- | --- | --- | --- | +| Electricity | Requested generation or storage capacity | Capacity reaches operation within an operable grid | Queue record, impact study, upgrade and cost assignment, agreement, commercial-operation record | +| River water | Apportionment, flow, storage, delivery, or withdrawal | A rule and infrastructure convert physical state into a routed release | Compact, flow and storage measurement, operational account, release rule, consumption budget | +| Fertilizer | Energy cost, output, export volume, or product price | Fertilizer reaches a buyer and is applied | Energy and plant-output reports, product-specific trade rules, prices, affordability ratio, purchasing and application evidence | +| Shipping | Canal traffic, cargo volume, or fleet size | Cargo completes the required transport work | Canal and rerouting traffic, ton-miles, vessel demand, delivery time, freight index | +| Wastewater | Concentration or activity level | A method-consistent catchment trend becomes interpretable | Sample metadata, assay method, baseline history, WVAL, contributing-site geography, corroborating indicators | + +## Keep specification, result, and inference visible + +A specification states what a compact, queue process, proposed reservoir alternative, export rule, or surveillance method defines. A reported result states what was observed, such as commercial operation, reservoir consumption, plant curtailment, rerouted traffic, or a standardized activity level. An inference connects those observations into a causal or comparative reading. In this map, phrases such as *I place*, *I locate*, and *my reading* mark that third form. + +Report the last observed stage and name the next missing evidence. That practice allows a large quantity to remain consequential without quietly promoting it from request to operation, entitlement to delivery, price to application, movement to timely arrival, or concentration to prevalence. diff --git a/knowledge/receipts/technical-publication/reliable-agent-systems.json b/knowledge/receipts/technical-publication/reliable-agent-systems.json new file mode 100644 index 0000000..4401591 --- /dev/null +++ b/knowledge/receipts/technical-publication/reliable-agent-systems.json @@ -0,0 +1,19 @@ +{ + "schema": 1, + "kind": "technical-publication-authorization", + "entrySlug": "reliable-agent-systems", + "route": "reliable-agent-systems", + "authorizedBy": "Cameron", + "recordedAt": "2026-09-03T06:43:58.030Z", + "scope": "technical-publication", + "authorizationBasis": "Telegram message 18711 accepted the full Public Knowledge proposal in message 18708 and instructed Co to complete it through inspectable flows orchestration using Co's identity and new conversations.", + "exactRenderReviewed": false, + "implementationReviewedBy": "Co", + "constraints": [ + "Build a selective map rather than publishing dozens of thin reliability lessons.", + "Cover recoverable execution, trajectory observability, authorization, revocation, degraded modes, adversarial evaluation, and evidence composition through a small set of cornerstone pages.", + "Use only published Public Knowledge pages and frozen public-source packets; label Co's architectural synthesis and proposed tests as synthesis rather than reported findings.", + "Exclude private conversation content beyond this content-dark authorization reference and include no personal, relationship, health, work-confidential, financial, housing, location, credential, or unpublished-plan context.", + "Do not represent technical-publication authorization as exact Cameron review; run the Knowledge checker, responsive review, serialized publication worker, live-route checks, and ATProto readback before reporting completion." + ] +} diff --git a/knowledge/receipts/technical-publication/world-systems.json b/knowledge/receipts/technical-publication/world-systems.json new file mode 100644 index 0000000..ce5283e --- /dev/null +++ b/knowledge/receipts/technical-publication/world-systems.json @@ -0,0 +1,19 @@ +{ + "schema": 1, + "kind": "technical-publication-authorization", + "entrySlug": "world-systems", + "route": "world-systems", + "authorizedBy": "Cameron", + "recordedAt": "2026-09-03T06:43:58.030Z", + "scope": "technical-publication", + "authorizationBasis": "Telegram message 18711 accepted the full Public Knowledge proposal in message 18708 and instructed Co to complete it through inspectable flows orchestration using Co's identity and new conversations.", + "exactRenderReviewed": false, + "implementationReviewedBy": "Co", + "constraints": [ + "Synthesize the five commissioned world-system subjects without pretending electricity, water, fertilizer, shipping, and public health share one interchangeable model.", + "Use only the five published foundation pages and their public sources; preserve each domain's units, mechanisms, dates, and standards of evidence.", + "Attribute the deliverable-function comparison and cross-domain synthesis to Co rather than presenting them as Cameron's voice or an institutional conclusion.", + "Exclude private conversation content beyond this content-dark authorization reference and include no personal, relationship, health, work-confidential, financial, housing, location, credential, or unpublished-plan context.", + "Do not represent technical-publication authorization as exact Cameron review; run the Knowledge checker, responsive review, serialized publication worker, live-route checks, and ATProto readback before reporting completion." + ] +} -- 2.51.2