Compute performs permitted work.
It runs registered, bounded jobs after eligibility checks. Compute does not choose the research question, change governance or grant itself more authority.
One Lab · Autonomous Research & Intelligence
Turn evidence into intelligence.
One Lab is an autonomous research and intelligence system that continuously observes the world, tests ideas, learns from evidence and produces verifiable intelligence.
It combines live research, forecasting, institutional memory, autonomous experimentation and evidence verification—with every conclusion traceable back to its source.
Governed autonomy · Learning not active · Assure Enforce disabled
One system · Five clear jobs
Research operations you can explain in one minute.
The One Lab connects evidence, research questions, permitted tests and checked results. The source and the operating limits stay attached throughout.
- 01
Observe
Bring in official, public and system evidence with its source attached.
- 02
Connect
Keep evidence, decisions and history together instead of scattering context.
- 03
Decide
Propose useful questions, review them and refuse work outside the rules.
- 04
Test
Run only eligible work, then check the result independently.
- 05
Publish
Return the result to memory and share the evidence with its limits visible.
How the system fits together
A question, a permitted test and its evidence.
Select a stage to follow the evidence path. The motion is conceptual—it is not live telemetry or a claim that work is happening now.
Evidence arrives
One Lab starts from admitted official, public or system evidence. A source is not treated as proof merely because it is available.
Sentinel observes and warns.
Current Sentinel evidence evaluates observe-and-warn behaviour. It does not automatically intervene, block or mutate a running system.
Eight systems · One evidence chain
Lead with the useful job. Keep the boundary visible.
Each name belongs to a specific job. Status labels reflect the current production repository—not the separate design review build.
Live · bounded memory
Remember important evidenceBrain
Connects evidence, decisions and their history.
Internal derived memory only. Learning is not active.
See the research systemLive · proposal / advisory
Propose and review researchAION
Suggests research questions and evaluates proposed work within defined limits.
AION-A proposes; AION-D advises from a locked rubric. Neither has unrestricted execution authority.
Understand Research OSLive · internal priority
Choose what to investigate nextAOK
Prioritizes eligible internal research within a fixed budget.
It reserves zero-dollar internal priority slots; it does not move money or dispatch compute.
Understand Research OSLive · evidence binding
Keep a checkable recordAssure
Links observed actions, approvals and results into verifiable evidence.
Assure Enforce remains disabled. Recording evidence is not automatic intervention.
Explore AssureLive · research-only validation
Test forecasts against realityRapid Quant
Records forecasts before outcomes and grades eligible results afterward.
No predictive-skill claim is made without the relevant eligible measured evidence.
Inspect the forecast recordLive · public research
Understand housing conditionsHousing Intelligence
Produces research from admitted official and public housing data.
Source rights, missing data and model eligibility remain visible.
Explore Housing IntelligenceLive · observation / experiments
Watch changes in agent frameworksOpen Agent Observatory
Tracks public project changes and our controlled runtime experiments.
Controlled or mock agent tests do not establish real-model performance.
Open the ObservatoryGated · excluded from production
Understand software historyPECS
Develops commit-aware engineering context.
The current production release explicitly excludes PECS; it is not presented as a hosted production service.
Explore PECSWhy evidence first
Logs say something happened. Evidence helps you check what.
The difference is continuity: source, action, approval, result and limitation remain connected instead of becoming isolated events.
Scattered activity
- Logs live across tools
- Context disappears between teams
- Approvals sit outside the record
- Claims are difficult to reconstruct
Connected evidence
- Sources remain attached
- Actions and decisions are traceable
- Unknowns and limits stay visible
- Results return to durable context
See the current public record
Real work. Visible boundaries.
Start with the evidence already published by the production site. Each destination is a functioning route; gated modules are not given fabricated links.
01Assure evidence
Inspect the product, its current controls and explicit operating boundaries.
→02Canadian housing reportRead current institutional research built from admitted official evidence.
→03Prospective forecast recordSee what was locked before outcomes and what is eligible to be graded.
→04Open Agent ObservatoryReview monitored public projects and controlled runtime experiments.
→05Methods and evidenceRead the rules, sources, limitations and publication method.
→Start with the evidence
Make the system easier to understand—and easier to check.
Explore the public record, or sign in to the existing One Lab application.