{"ontology":"https://layer.re/machine/schema.json","epistemic_model":"evidence-backed-temporal-layer-graph","canonical":true,"type":"layer_history","layer":{"id":"layer-agents","slug":"agents","name":"Agents","description":"Systems that plan, act and use tools across environments.","machine_description":"Systems that plan, act and use tools across environments.","active":1},"snapshots":[{"id":"state-agents-2022q1","layer_id":"layer-agents","as_of":"2022-03-31 23:59:59.000","summary":"Within the selected Q1 2022 evidence, models could follow instructions, reason under prompting and generate code, but persistent tool-using autonomous agents are not yet supported as an independent operational layer.","dependencies":["layer-models","layer-software"],"constraints":["reliable long-horizon execution","tool use","persistent state","delegation controls"],"emerging_capabilities":["code/problem-solving systems as a precursor, not yet general agents"],"underdeveloped_components":["persistent autonomy","general tool orchestration","background execution"],"uncertainties":["absence from this curated source set is not proof that no agent research existed anywhere"],"confidence":0.78,"previous_snapshot_id":null,"mapping_mode":"retrospective","reconstructed_at":"2026-10-01T17:34:32.000Z","created_at":"2026-10-01T18:05:36.398Z","updated_at":"2026-10-01T18:05:36.398Z"},{"id":"state-agents-2022q2","layer_id":"layer-agents","as_of":"2022-06-30 23:59:59.000","summary":"By end-Q2, agent-like capability was materially more visible in research: models could participate in planning and action loops across embodied or interactive environments. The layer was still a precursor/research capability rather than a persistent deployed agent category.","dependencies":["layer-models","layer-research","layer-software","layer-robotics"],"constraints":["durable memory","general tool use","reliable long-horizon execution","persistent identity","delegation and permissions"],"emerging_capabilities":["language-model-guided robotic planning","multi-task generalist action policies"],"underdeveloped_components":["persistent background autonomy","broad external-tool orchestration","production agent ecosystems"],"uncertainties":["the boundary between 'model policy', 'robot controller' and the later Agents layer remains conceptually ambiguous"],"confidence":0.88,"previous_snapshot_id":"state-agents-2022q1","mapping_mode":"retrospective","reconstructed_at":"2026-10-01T18:14:00.000Z","created_at":"2026-10-01T19:22:17.766Z","updated_at":"2026-10-01T19:22:17.766Z"},{"id":"state-agents-2022q3","layer_id":"layer-agents","as_of":"2022-09-30 23:59:59.000","summary":"Agent-like capability matured through internet retrieval, long-term conversational memory, explicit behavioral rules, human feedback, closed-loop environmental feedback and generated executable policies, but remained a research/demo precursor rather than a persistent autonomous operational layer.","dependencies":["layer-models","layer-software","layer-research","layer-robotics","layer-human"],"constraints":["durable identity","general tool use","reliable long-horizon execution","delegated permissions","safety"],"emerging_capabilities":["internet-searching dialogue agents","conversational long-term-memory capability","evidence retrieval","closed-loop embodied replanning","generated robot-policy code"],"underdeveloped_components":["persistent background autonomy","general cross-application action","portable delegated permissions","production agent ecosystems"],"uncertainties":["the boundary between richer model systems and an independent Agents layer remained unresolved"],"confidence":0.91,"previous_snapshot_id":"state-agents-2022q2","mapping_mode":"retrospective","reconstructed_at":"2026-10-02T15:12:29.000Z","created_at":"2026-10-02T16:08:16.260Z","updated_at":"2026-10-02T16:08:16.260Z"},{"id":"state-agents-2022q4","layer_id":"layer-agents","as_of":"2022-12-31 23:59:59.000","summary":"Agent-like research in Q4 could interleave reasoning with actions and negotiate toward goals in a structured multi-agent environment, but still lacked persistence, general tools, delegated authority and cross-application continuity.","dependencies":["layer-models","layer-research","layer-software","layer-human","layer-robotics"],"constraints":["persistent identity","durable memory","general tool access","delegated permissions","reliable long-horizon execution"],"emerging_capabilities":["reasoning-action loops over external environments","goal-directed strategic dialogue with humans"],"underdeveloped_components":["persistent background autonomy","general cross-application operation","portable permissions"],"uncertainties":["ReAct and CICERO can reasonably be classified as bounded model systems rather than an independent operational agent layer"],"confidence":0.92,"previous_snapshot_id":"state-agents-2022q3","mapping_mode":"retrospective","reconstructed_at":"2026-10-03T14:26:11.000Z","created_at":"2026-10-03T14:45:10.925Z","updated_at":"2026-10-03T14:45:10.925Z"},{"id":"state-agents-2023q1","layer_id":"layer-agents","as_of":"2023-03-31 23:59:59.000","summary":"Agent-like capability crossed from research-only tool-use precursors into a constrained product alpha through ChatGPT plugins, but lacked persistent identity, durable memory, general delegated authority and reliable long-horizon execution.","dependencies":["layer-models","layer-software","layer-research","layer-human","layer-governance"],"constraints":["persistent identity","durable memory","general permissions","tool safety","reliable long-horizon execution"],"emerging_capabilities":["research API/tool use","browser access","code execution","third-party service interaction in limited alpha"],"underdeveloped_components":["background persistence","general cross-application autonomy","portable delegated permissions","mature agent ecosystem"],"uncertainties":["Plugins were a small alpha and Toolformer was research."],"confidence":0.93,"previous_snapshot_id":"state-agents-2022q4","mapping_mode":"retrospective","reconstructed_at":"2026-10-05T15:23:51.000Z","created_at":"2026-10-05T18:17:53.801Z","updated_at":"2026-10-05T18:17:53.801Z"},{"id":"state-agents-2023q2","layer_id":"layer-agents","as_of":"2023-06-30 23:59:59.000","summary":"Agent-like systems gained wider deployed tool access and reusable integration contracts, while execution remained application-controlled and persistence/authority immature.","dependencies":["layer-models","layer-software","layer-human","layer-governance"],"constraints":["tool safety","authorization","persistent identity","durable memory"],"emerging_capabilities":["broader plugins","common plugin interfaces","structured function calls"],"underdeveloped_components":["persistent agents","portable permissions","reliable autonomy"],"uncertainties":["integration primitives are not persistent autonomy"],"confidence":0.96,"previous_snapshot_id":"state-agents-2023q1","mapping_mode":"retrospective","reconstructed_at":"2026-10-07T12:55:58.000Z","created_at":"2026-10-07T14:04:48.610Z","updated_at":"2026-10-07T14:04:48.610Z"},{"id":"state-agents-2023q3","layer_id":"layer-agents","as_of":"2023-09-30 23:59:59.000","summary":"Agent-like systems in Q3 became more executable and orchestratable through enterprise-grade sandboxed data/code analysis and explicit multi-agent workflow frameworks, while execution remained bounded by applications and human-configured tools.","dependencies":["layer-models","layer-software","layer-human","layer-governance","layer-permissions"],"constraints":["persistent identity","durable memory","authorization","tool safety","long-horizon reliability"],"emerging_capabilities":["sandboxed analysis and code execution in enterprise assistant deployment","multi-agent orchestration frameworks","enterprise advanced data analysis"],"underdeveloped_components":["persistent background operation","general delegated authority","portable permissions","reliable autonomous execution"],"uncertainties":["Code Interpreter and AutoGen do not establish persistent general-purpose agents"],"confidence":0.96,"previous_snapshot_id":"state-agents-2023q2","mapping_mode":"retrospective","reconstructed_at":"2026-10-07T16:04:51.000Z","created_at":"2026-10-07T16:56:16.866Z","updated_at":"2026-10-07T16:56:16.866Z"},{"id":"state-agents-2023q4","layer_id":"layer-agents","as_of":"2023-12-31 23:59:59.000","summary":"Agent-like infrastructure gained provider-managed persistent thread state, retrieval/execution tools and reusable customized assistant configurations while remaining user/app initiated.","dependencies":["layer-models","layer-software","layer-human","layer-permissions"],"constraints":["authorization","durable identity","background operation","tool safety"],"emerging_capabilities":["Assistants API threads","Retrieval","Code Interpreter","custom GPTs","Actions"],"underdeveloped_components":["continuous autonomous operation","portable delegated permissions"],"uncertainties":["Persistent thread state is not continuously running autonomy"],"confidence":0.99,"previous_snapshot_id":"state-agents-2023q3","mapping_mode":"retrospective","reconstructed_at":"2026-10-08T15:50:00.000Z","created_at":"2026-10-08T17:00:00.130Z","updated_at":"2026-10-08T17:00:00.130Z"},{"id":"state-agents-2024q1","layer_id":"layer-agents","as_of":"2024-03-31 23:59:59.000","summary":"Agent-like systems added bounded cross-chat personalization and sandboxed engineering/task demonstrations; durable user-independent autonomy remained unestablished.","dependencies":[],"constraints":["availability scope","reliability and implementation limits"],"emerging_capabilities":["ChatGPT tests cross-chat memory with user controls","Devin is introduced through demos and early access","SIMA research follows instructions through game interfaces"],"underdeveloped_components":["generalization beyond the selected evidence"],"uncertainties":["Exclude September 2024 and 2025 updates; remembered preferences are not autonomous background work.","Vendor demos and SWE-bench results do not establish general availability or reliable unattended engineering.","Approximately ten-second tasks do not establish long-horizon autonomy, public deployment or semantic runtime-state access."],"confidence":0.93,"previous_snapshot_id":"state-agents-2023q4","mapping_mode":"retrospective","reconstructed_at":"2026-10-09T19:12:47.000Z","created_at":"2026-10-09T21:04:17.185Z","updated_at":"2026-10-09T21:04:17.185Z"},{"id":"state-agents-2024q2","layer_id":"layer-agents","as_of":"2024-06-30 23:59:59.000","summary":"Responsive multimodal prototypes and built-in model code execution strengthened agent workflows, without establishing continuously autonomous general-purpose agents.","dependencies":[],"constraints":["availability scope","reliability and implementation limits"],"emerging_capabilities":["Project Astra demonstrates responsive multimodal assistant prototypes","Gemini API adds code execution and two-million-token GA context"],"underdeveloped_components":["generalization beyond the selected evidence"],"uncertainties":["A demonstrated responsive prototype is not a shipped persistent autonomous agent.","Code execution is a bounded API capability, not universal delegated authority or reliable autonomous work."],"confidence":0.93,"previous_snapshot_id":"state-agents-2024q1","mapping_mode":"retrospective","reconstructed_at":"2026-10-09T19:12:47.000Z","created_at":"2026-10-10T07:29:06.787Z","updated_at":"2026-10-10T07:29:06.787Z"},{"id":"state-agents-2024q4","layer_id":"layer-agents","as_of":"2024-12-31 23:59:59.000","summary":"Developer computer control, shared tool/data connections and bounded research workflows enlarged action surfaces without proving persistent general autonomy.","dependencies":[],"constraints":["availability scope","reliability and implementation limits"],"emerging_capabilities":["Claude computer use enters public beta","Model Context Protocol is open-sourced","Gemini Deep Research introduces plan-and-browse reports","Full o1 API gains tool and vision interfaces"],"underdeveloped_components":["generalization beyond the selected evidence"],"uncertainties":["Experimental computer use remains error-prone and requires developer integration and supervision.","Initial protocol release is not universal interoperability, portable authorization or autonomous persistence.","A bounded multi-step research task is not persistent general autonomous operation.","The rollout was not immediately available to every API tier; tool calling does not authorize arbitrary actions."],"confidence":0.93,"previous_snapshot_id":"state-agents-2024q2","mapping_mode":"retrospective","reconstructed_at":"2026-10-09T19:12:47.000Z","created_at":"2026-10-10T08:02:14.503Z","updated_at":"2026-10-10T08:02:14.503Z"},{"id":"state-agents-2025q1","layer_id":"layer-agents","as_of":"2025-03-31 23:59:59.000","summary":"Browser-operation previews, research sessions and coding/tool SDKs made bounded delegation accessible with human takeover and review boundaries.","dependencies":[],"constraints":["availability scope","reliability and implementation limits"],"emerging_capabilities":["Operator begins US Pro research preview","Deep research launches for bounded web investigations","Claude 3.7 and Claude Code research preview launch","Responses API and Agents SDK launch"],"underdeveloped_components":["generalization beyond the selected evidence"],"uncertainties":["Login takeover, purchase confirmations and limited access are not generalized delegated authority or continuously autonomous operation.","Cited reports can contain errors; a research session is not general background autonomy. Later upgrades are excluded.","Model availability differs from coding-agent preview; autonomous coding claims do not establish unsupervised reliability.","Built-in search and computer-use previews are bounded tools; orchestration is not proof of autonomous agents."],"confidence":0.93,"previous_snapshot_id":"state-agents-2024q4","mapping_mode":"retrospective","reconstructed_at":"2026-10-09T19:12:47.000Z","created_at":"2026-10-10T08:11:02.814Z","updated_at":"2026-10-10T08:11:02.814Z"},{"id":"state-agents-2025q2","layer_id":"layer-agents","as_of":"2025-06-30 23:59:59.000","summary":"Cloud task execution and inter-agent communication are becoming product and protocol capabilities, with previews and approval boundaries still material.","dependencies":[],"constraints":["Protocol publication does not prove universal interoperability or shared authorization.","An announced program does not establish widespread production agent payments.","Research preview access and selected demonstrations do not establish reliable autonomous software delivery.","Vendor evaluations and long-running demonstrations do not establish general unattended autonomy."],"emerging_capabilities":["A2A protocol is announced","Visa introduces Intelligent Commerce","Codex cloud research preview launches","Claude 4 releases and Claude Code reaches GA"],"underdeveloped_components":["Independent measurement of deployment scope and outcomes"],"uncertainties":["The referenced historical predecessor belongs to the reviewed chronology; the live history must be checked before ingestion."],"confidence":0.9,"previous_snapshot_id":"state-agents-2025q1","mapping_mode":"retrospective","reconstructed_at":"2026-10-10T06:24:42.000Z","created_at":"2026-10-10T08:25:19.360Z","updated_at":"2026-10-10T08:25:19.360Z"},{"id":"state-agents-2025q3","layer_id":"layer-agents","as_of":"2025-09-30 23:59:59.000","summary":"This period adds specific agents evidence through ChatGPT agent starts paid-user rollout; AP2 agent payment protocol is announced; Gemini Robotics 1.5 introduces physical-agent planning; Claude Sonnet 4.5 is released; Instant Checkout launches with an Agentic Commerce Protocol. Maturity remains specific to the cited products and programs; this synthesis does not infer sector-wide adoption.","dependencies":[],"constraints":["Bounded tasks and staged rollout do not establish persistent always-on agents or unrestricted permissions.","Protocol support does not confer bank permissions or prove broad live transaction adoption.","API access to reasoning is distinct from unrestricted access to robot actuation models.","Long-horizon vendor demonstrations do not establish unattended success rates across real production tasks.","Merchant expansion is prospective and checkout still requires user confirmation using existing payment infrastructure."],"emerging_capabilities":["ChatGPT agent starts paid-user rollout","AP2 agent payment protocol is announced","Gemini Robotics 1.5 introduces physical-agent planning","Claude Sonnet 4.5 is released","Instant Checkout launches with an Agentic Commerce Protocol"],"underdeveloped_components":["Independent measurement of deployment scope and outcomes"],"uncertainties":["The referenced historical predecessor belongs to the reviewed chronology; the live history must be checked before ingestion."],"confidence":0.9,"previous_snapshot_id":"state-agents-2025q2","mapping_mode":"retrospective","reconstructed_at":"2026-10-10T06:24:42.000Z","created_at":"2026-10-10T08:31:22.789Z","updated_at":"2026-10-10T08:31:22.789Z"},{"id":"state-agents-2025q4","layer_id":"layer-agents","as_of":"2025-12-31 23:59:59.000","summary":"This period adds specific agents evidence through Codex reaches general availability; Anthropic introduces Agent Skills; SIMA 2 research extends virtual-world agent interaction; Gemini 3 and Antigravity are introduced; Agentic AI Foundation is formed; Agent Skills specification is opened for portability. Maturity remains specific to the cited products and programs; this synthesis does not infer sector-wide adoption.","dependencies":[],"constraints":["Tool availability does not remove review requirements or establish universal autonomous engineering.","Loading procedural knowledge is not evidence that embedded scripts are safe or universally portable.","Research interaction through game interfaces does not prove native semantic game-state export or production deployment.","Preview access and provider evaluations do not prove dependable unsupervised delivery.","Neutral governance and donated projects do not establish universal implementation or solved interoperability.","Specification publication does not establish uniform support or a permission enforcement standard."],"emerging_capabilities":["Codex reaches general availability","Anthropic introduces Agent Skills","SIMA 2 research extends virtual-world agent interaction","Gemini 3 and Antigravity are introduced","Agentic AI Foundation is formed","Agent Skills specification is opened for portability"],"underdeveloped_components":["Independent measurement of deployment scope and outcomes"],"uncertainties":["The referenced historical predecessor belongs to the reviewed chronology; the live history must be checked before ingestion."],"confidence":0.9,"previous_snapshot_id":"state-agents-2025q3","mapping_mode":"retrospective","reconstructed_at":"2026-10-10T06:24:42.000Z","created_at":"2026-10-10T08:37:10.926Z","updated_at":"2026-10-10T08:37:10.926Z"},{"id":"state-agents-2026q1","layer_id":"layer-agents","as_of":"2026-03-31 23:59:59.000","summary":"Agent work reaches desktop supervision and native computer interaction, with research previews and beta coordination still visible.","dependencies":[],"constraints":["Promised forthcoming checkout is distinct from a completed production launch on this date.","Experimental desktop work is not general availability or proof of dependable unsupervised workplace autonomy.","The March Windows update is separate; long-running workflow claims do not quantify independent reliability.","Million-token context is beta and agent teams are a research preview rather than unrestricted production automation.","Model capability does not grant permissions or independently establish safe end-to-end task completion."],"emerging_capabilities":["Universal Commerce Protocol is introduced","Cowork enters desktop research preview","Codex desktop app launches on macOS","Claude Opus 4.6 is released","GPT-5.4 adds native computer use and tool search"],"underdeveloped_components":["Independent measurement of deployment scope and outcomes"],"uncertainties":["The referenced historical predecessor belongs to the reviewed chronology; the live history must be checked before ingestion."],"confidence":0.9,"previous_snapshot_id":"state-agents-2025q4","mapping_mode":"retrospective","reconstructed_at":"2026-10-10T06:24:42.000Z","created_at":"2026-10-10T08:45:32.708Z","updated_at":"2026-10-10T08:45:32.708Z"},{"id":"state-agents-2026q2","layer_id":"layer-agents","as_of":"2026-06-30 23:59:59.000","summary":"This period adds specific agents evidence through Google makes agents first-class identity principals; Anthropic announces SpaceX compute agreement and higher limits; Worldline ING and Mastercard report live agent payment; GPT-5.6 family enters limited preview. Maturity remains specific to the cited products and programs; this synthesis does not infer sector-wide adoption.","dependencies":[],"constraints":["Agent Runtime identity availability must not be conflated with preview platform and auth-manager components.","A stated near-term capacity agreement is not independent confirmation that every GPU was delivered immediately.","First-of-kind language is an actor claim; one reported transaction does not establish mass adoption.","Preview benchmarks and harness results are company claims; July access is not backdated into June."],"emerging_capabilities":["Google makes agents first-class identity principals","Anthropic announces SpaceX compute agreement and higher limits","Worldline ING and Mastercard report live agent payment","GPT-5.6 family enters limited preview"],"underdeveloped_components":["Independent measurement of deployment scope and outcomes"],"uncertainties":["The referenced historical predecessor belongs to the reviewed chronology; the live history must be checked before ingestion."],"confidence":0.9,"previous_snapshot_id":"state-agents-2026q1","mapping_mode":"retrospective","reconstructed_at":"2026-10-10T06:24:42.000Z","created_at":"2026-10-10T09:06:48.252Z","updated_at":"2026-10-10T09:06:48.252Z"},{"id":"state-agents-20260930-initial","layer_id":"layer-agents","as_of":"2026-09-30 12:45:00.000","summary":"Persistent, always-on agents are moving into mainstream product form while infrastructure for controlling their actions is becoming a separate engineering concern.","dependencies":["layer-models","layer-compute","layer-software","layer-permissions","layer-identity"],"constraints":["permissioning","security","liability","reliable long-horizon execution"],"emerging_capabilities":["persistent cloud agents","multi-app action","background work","delegation"],"underdeveloped_components":["portable permission standards","mature liability rules","widely proven safe autonomy"],"uncertainties":["frequency and severity of real-world failures","how much autonomy users and institutions will accept"],"confidence":0.94,"previous_snapshot_id":null,"mapping_mode":"contemporaneous","reconstructed_at":null,"created_at":"2026-09-30T12:54:32.491Z","updated_at":"2026-09-30T12:54:32.491Z"},{"id":"state-agents-20260930-late","layer_id":"layer-agents","as_of":"2026-09-30 21:20:00.000","summary":"Agent deployment is entering an enforcement phase as US regulators investigate risks from agentic systems, adding legal accountability to the existing runtime-safety and permissioning stack.","dependencies":["layer-models","layer-permissions","layer-governance","layer-software"],"constraints":["liability","cybersecurity failure modes","reliable long-horizon execution"],"emerging_capabilities":["domain-specialized tool-using agents for engineering"],"underdeveloped_components":["clear liability standards","portable evidence and audit standards"],"uncertainties":["scope and outcome of FTC action"],"confidence":0.95,"previous_snapshot_id":"state-agents-20260930-initial","mapping_mode":"contemporaneous","reconstructed_at":null,"created_at":"2026-09-30T21:57:03.549Z","updated_at":"2026-09-30T21:57:03.549Z"},{"id":"state-agents-2026q3","layer_id":"layer-agents","as_of":"2026-09-30 23:59:59.000","summary":"Dedicated virtual computers and restricted persistent-agent rollouts make ongoing delegated work a product architecture, without independently proving robust continuous autonomy.","dependencies":[],"constraints":["Broad product access remains plan-dependent and does not validate all vendor performance claims independently.","Performance and cost comparisons are vendor-reported, not universal production outcome measures.","GA is a product status, not evidence of universal identity federation or adoption.","The September 3 entry explicitly says Astra is not yet generally available; later rollout must not be backdated.","Confidential-computing enhancements announced for later are not established launch capabilities.","Reference architecture does not establish broad hardware deployment or invulnerable permission enforcement.","Rollout is gradual and geographically restricted; product claims do not independently prove reliable continuous operation.","Broader developer and consumer access is prospective; internal productivity and autonomous patching assertions remain provider claims.","The specialized model and bundled service are prospective; the agreement is not a shipped autonomous chip-design system."],"emerging_capabilities":["GPT-5.6 expands to broader product and API access","Claude Opus 5 is released","Agent Identity APIs and auth manager reach GA","GPT-6 Astra begins limited organizational rollout","Meta introduces Muse personal agent","NVIDIA introduces Open Agent Safety Platform","Dots introduces persistent agents with cloud computers","Google announces Gemini 4 Argon with restricted defender access","OpenAI and Synopsys agree to develop a chip-design model"],"underdeveloped_components":["Independent measurement of deployment scope and outcomes"],"uncertainties":["The referenced historical predecessor belongs to the reviewed chronology; the live history must be checked before ingestion."],"confidence":0.9,"previous_snapshot_id":"state-agents-2026q2","mapping_mode":"retrospective","reconstructed_at":"2026-10-10T06:24:42.000Z","created_at":"2026-10-10T09:18:22.068Z","updated_at":"2026-10-10T09:18:22.068Z"},{"id":"state-agents-20261001-human-context","layer_id":"layer-agents","as_of":"2026-10-01 19:31:00.000","summary":"Persistent personal agents increasingly depend on durable human-specific context in addition to models, tools, identity and permissions, particularly when moving from response personalization toward delegated action across sessions and services.","dependencies":["layer-models","layer-compute","layer-software","layer-permissions","layer-identity","layer-human"],"constraints":["context accuracy","permissioning","portability","reliable delegated action"],"emerging_capabilities":["context-aware persistent agency","cross-session preference continuity","delegated action grounded in human context"],"underdeveloped_components":["portable human-context standards","cross-agent handoff semantics"],"uncertainties":["whether context portability becomes broadly implemented rather than remaining protocol-level architecture"],"confidence":0.91,"previous_snapshot_id":"state-agents-20260930-late","mapping_mode":"contemporaneous","reconstructed_at":null,"created_at":"2026-10-01T20:08:12.917Z","updated_at":"2026-10-01T20:08:12.917Z"},{"id":"state-agents-20261002-control-plane","layer_id":"layer-agents","as_of":"2026-10-02 07:22:00.000","summary":"Persistent agents are increasingly surrounded by explicit identity, credential and permission controls that govern what they may access and which consequential actions they may execute.","dependencies":["layer-models","layer-compute","layer-software","layer-permissions","layer-identity"],"constraints":["reliable delegation","credential security","approval design","liability"],"emerging_capabilities":["permissioned cloud and local-computer action","identity-bound delegated commerce"],"underdeveloped_components":["portable identity and authorization standards"],"uncertainties":["failure rates and user acceptance of higher-autonomy actions"],"confidence":0.95,"previous_snapshot_id":null,"mapping_mode":"contemporaneous","reconstructed_at":null,"created_at":"2026-10-02T09:22:12.384Z","updated_at":"2026-10-02T09:22:12.384Z"},{"id":"state-agents-20261004-state-infra","layer_id":"layer-agents","as_of":"2026-10-04 21:50:08.000","summary":"Persistent agents are gaining more explicit supporting infrastructure for state and application access: rapidly provisioned isolated databases are being targeted at agent/task workloads, while application platforms are beginning to expose capabilities directly to agents through authenticated contexts.","dependencies":["layer-models","layer-compute","layer-software","layer-permissions","layer-identity"],"constraints":["reliable delegation","credential security","authorization scope","revocation","state isolation","uncertain workload generality"],"emerging_capabilities":["per-agent or per-task isolated state stores","agent-addressable application backends","identity-bound delegated application action"],"underdeveloped_components":["portable authorization semantics","cross-platform agent state standards","independent workload telemetry"],"uncertainties":["whether database-per-agent becomes a broad architecture rather than a provider-specific pattern","security and adoption of signed-in-user MCP operation"],"confidence":0.91,"previous_snapshot_id":null,"mapping_mode":"contemporaneous","reconstructed_at":null,"created_at":"2026-10-04T21:57:54.258Z","updated_at":"2026-10-04T21:57:54.258Z"},{"id":"state-agents-20261008-persistent-principals","layer_id":"layer-agents","as_of":"2026-10-08 19:29:00.000","summary":"Persistent agents have moved beyond supporting state and application infrastructure toward an explicit organizational-worker architecture. Google documents cloud-persistent memory and context across surfaces, headless access, long-running delegated work, multi-agent orchestration, model choice separated from the persistent agent, and coworker agents with persistent roles and dedicated identities and resources. This is product-architecture evidence, not proof of reliable unsupervised multi-day performance or broad adoption.","dependencies":["layer-models","layer-compute","layer-software","layer-identity","layer-permissions","layer-human"],"constraints":["long-horizon reliability","rollout and adoption breadth","human oversight","cross-provider portability","cost and operational governance"],"emerging_capabilities":["persistent cross-device agent context","headless agent access","long-running delegated tasks","dynamic sub-agent orchestration","model-decoupled agent substrate","persistent coworker-agent roles"],"underdeveloped_components":["independent long-horizon task telemetry","portable agent state standards","cross-provider identity and delegation semantics"],"uncertainties":["independent reliability of hours-or-days execution","availability scope of all coworker-agent capabilities","degree of human intervention in real deployments"],"confidence":0.95,"previous_snapshot_id":"state-agents-20261004-state-infra","mapping_mode":"contemporaneous","reconstructed_at":null,"created_at":"2026-10-08T20:03:28.741Z","updated_at":"2026-10-08T20:03:28.741Z"}]}