THE FLAGSHIP · COMPUTER USE
Find valuable work. Develop a capability that can deliver it. Use frontier reasoning and Computer Use to improve strategies and discover a fundamentally different game to play.
Begin free with a mission worth pursuing. Computer execution uses your configured native executor and model; useful results need review.
Public software. Frontier research. Mathematical exploration. Start with a valuable outcome, then develop the capability to pursue it.
Reviewed deliverables, strategies and lessons. Keep and improve what proves useful.
ONE PROJECT. EXPLICIT CHOICES. RETAINED HISTORY.
Define the outcome, prepare bounded Computer Use work and review what was actually delivered. Keep strategies, attempts and lessons in the same project as you continue through Alpha, Jobs and Fleet.
Level 3 · Meta-strategic intelligence: discover, test, combine and improve strategies themselves. Level 4 · Transformational intelligence: recognize a restrictive formulation and investigate a fundamentally different game to play.
The browser prepares native missions and unfunded work, and preserves checked historical returns. Computer execution requires a configured model and disposable Chromium runner. Independent qualification and settlement remain separate requirements.
Maintained engineering examples are available separately. Explore local catalog tools ↗
CHOOSE THE PAYMENT ROUTE
AGIJobManager uses USDC. Alpha and AGI Jobs v0 use $AGIALPHA. Each keeps its own pricing, admission and settlement rules; no automatic conversion occurs.
Fleet receives an explicitly selected local task and returns execution and internal checking evidence. A returned file grants no execution permission.
ALPHA AGENT + AGI JOBS
Alpha helps plan and produce work. AGI Jobs provides rules for commissioning, reviewing and settling it. Explore how a carefully connected workflow can bring both strengths together.
ONE SHARED TOKEN · ETHEREUM MAINNET
0xA61a3B3a130a9c20768EEBF97E21515A6046a1fA
5 Alpha + 24 AGI Jobs instances in the 11–12 October 2026 public records.
Read the linked deployment evidence and readiness notes. A shared token does not create an automatic bridge or open live job intake.
NOVA-SEED STUDIO
Shape a project, check its exact work plan and protect it with encrypted recovery. Keep the same history as you connect execution, inspect results and consider the next cycle.
Local preparation needs no wallet or payment. New minting, funding and commitments are blocked on the archived Mainnet deployment. External execution, review and settlement have separate prerequisites.
A PROJECT YOU CAN KEEP
Download recovery before leaving Studio. Keep the capsule passphrase separately; never use your wallet recovery phrase.
Simple from the first step
Choose a result that matters. Improve the approach. Keep the work and what you learned.
Choose a consequential opportunity in software, research or mathematics. State who benefits and what a useful deliverable must achieve.
Level 3: test and combine strategies. Level 4: challenge an unnecessary restriction while preserving the required outcome. Review the real deliverable before accepting an attempt.
Save your workspace with its results, strategies, reviews and history. Reopen it, preserve what worked and prepare the next improvement.
A little reassurance
Open any question to read the answer.
Opportunity planning, mission preparation and workspace downloads are free, without an account or wallet. Native Computer Use execution requires an operator’s configured runner and model, with separate provider and operating costs. Maintained local engineering examples remain available.
Nothing to install for browser planning, the catalog example, personal lab or simple planner. Native Computer Use execution requires a configured model and container runner. Your saved workspace remains an HTML file you can open offline.
Browser planning and catalog processing stay in your browser. Native Computer Use sends the approved work surface and screenshots to the operator-selected model. Use public, non-personal inputs. Save your complete workspace before closing and after changes; exports include your supplied inputs.
A proposed capability, generated patch or mathematical argument is a candidate. Inspect its actual work, test the important claims and measure full costs before claiming advantage. Independent proof, Specialist status, income and operational permission require their own evidence.
Yes. Choose Larger text at the top or use your browser zoom. English and French are available on the homepage, in the Alpha flagship, in your personal lab and in the simple planner.
YOUR NEXT DISCOVERY STARTS HERE
Discover valuable work. Build a capability to deliver it. Improve the strategy. Change the game.
SELECTED READING
Explore opportunities, compare improvements and retain useful discoveries.
Governed improvement
Connect candidate improvements, baseline comparisons and retained evidence to the longer-term successor vision.
English essay · LinkedIn
Explore the current guide: Governed improvementSustainable improvement
Ask whether a useful capability can cover its execution and checking costs before reinvesting in another improvement.
English essay · LinkedIn
Explore the current guide: Sustainable improvementDiscovery & invention
Explore a market that allocates problems to capable solvers, pays for discernment and remembers what the work establishes.
English essay · LinkedIn
Explore the current guide: Discovery & inventionEvidence & memory
Explore how a work history can keep the original request, delivered result and review discoverable together.
English essay · LinkedIn
Explore the current guide: Evidence & memoryEssays by Vincent Boucher · English originals on LinkedIn. Some describe earlier versions; current guides explain setup, payments and availability. LinkedIn may ask you to sign in.
Alpha and AGIJobsv0 share a $AGIALPHA ecosystem. AGIJobManager provides the separate USDC route. Each execution and settlement connection needs its own configuration and evidence.
Reading notes and version contextFor the curious
There is more here when you want it: planning tools, experiments and the original research.
Start with the Alpha flagship above, explore the personal lab or try the simple planner. The advanced tools and research below are in English; some offer their own language options and may need technical knowledge.
META-AGENTIC α–AGI · OPEN RESEARCH
A laboratory for the next frontier of intelligence. Discover opportunities. Orchestrate useful work. Follow the value through the entire loop.
New here? Follow a simple guided example in English or French. No account, wallet, payment or installation needed.
Identify · Give the mission a goal and bounded, explicit inputs.
BUILD · VERIFY · CARRY KNOWLEDGE FORWARD
Construct a bounded capability and inspect its evidence. Run a local rehearsal in English or French, then exchange requests and results with your native agent. Knowledge carries forward; permissions require fresh review.
NEW / THE COMPOUNDING LAB · v1.35.0
Learn once. Freeze the capability. Test a different future. Inspect real predictions, expose failure, count the costs, and take the complete evidence with you.
Begin the experiment ↗NOVA-SEEDS / PROOF BLOOM
Five guided experiences. One complete path from unsupported claim to executable jobs, reviewed evidence and the next invention.
Enter the Proof Bloom ↗Current Mainnet record · five verified contract addresses. Read-only evidence recorded on 18 October 2026. Operational approval remains a separate step.
THE WHITE PAPER, MADE EXPLORABLE
Take a real computational journey through the original vision. Each decision changes the plan. Every result can be inspected.
01 / ENERGY & RESILIENCE
Seven opportunities. One pilot budget. Find a portfolio that makes the most of both.
Launch the mission ↗02 / SCIENCE & DISCOVERY
Turn a crowded research backlog into a coordinated, reproducible experiment plan.
Launch the mission ↗03 / OPERATIONS & VALUE
Choose useful automation pilots. Make cost, risk and review effort visible.
Launch the mission ↗Editable scenario data · actual portfolio and scheduling algorithms · real encryption · clearly labeled economic simulations
Methods & evidence ↗NEW / THE INSIGHT ATLAS
Explore the $15 quadrillion illustrative envelope. Inspect each sector, challenge the assumptions and give the next opportunity a testable shape.
Explore the opportunity ↗Compare agent firms. Expose the bottlenecks. Search 18 architectures, then evaluate the result on a separate holdout.
Build an agent firm ↗03 / ALPHA UNDER TRIALFollow the sources, export a validation brief and promote only replayed model results into a recoverable Chronicle.
Inspect the evidence ↗Three connected experiences · editable inputs · exact calculations · replayable evidence · no account or API key
THE GOVERNANCE OBSERVATORY
Move the assumptions. Watch strategies compete. Test the risk budget and the eight-day governance gate.
Enter the observatory ↗01 / MAKE SOMETHING USEFUL
Try the example. Replace it with your own inputs.
Every
result comes with its method and limits.
Ready. The example uses sample planning units; edit them for your problem.
RESULT / READY FOR YOUR REVIEW
Browser reports contain a SHA-256 integrity digest and your review, not a trusted agent signature. Use your local agent for signed identity, durable journals and verified $AGIALPHA receipts.
Only missions you explicitly save stay in this browser. Reload saved inputs to improve a decision. Export important work before clearing browser data.
02 / EXPLORE LOCAL INTELLIGENCE
Try actual GPT-2 text generation in this browser. The model downloads only when you ask, then generates locally. It is a text-completion research baseline; it can be wrong and is not a reasoning assistant.
Open the full Insight simulation studio ↗First use downloads about 128 MB of model weights plus the runtime. Prompts stay on this device.
Your model output will appear here.
03 / FOLLOW YOUR CURIOSITY
Eleven practical workspaces. Bring your constraints, calculate the trade-offs, and export a plan you can use.
No matching demos. Clear the search to see all entries.
01 / Build a business
Which investments fit a $1.2M budget and 500 staff days?
Investment slate · cash-flow model · downside sensitivityOpen workspace ↗
02 / Build a business
Which inventions justify the cost of developing and operating them?
Development portfolio · prerequisite map · verification jobsOpen workspace ↗
03 / Run operations
How should 60 batches be sourced when no supplier may exceed 45%?
Order allocation · landed costs · supplier-outage stress testOpen workspace ↗
04 / Run operations
Can seven customer orders share four resources and meet the delivery window?
Resource-by-hour schedule · lateness report · priority orderOpen workspace ↗
05 / Run operations
What delivery order fits a shared design, engineering and QA team?
Launch sequence · resource calendar · bottleneck analysisOpen workspace ↗
06 / Run operations
How much stock should be ordered for a two-week supplier lead time?
Holdout evaluation · replenishment quantity · demand stress testOpen workspace ↗
07 / Run operations
Which shifts need extra staff, what will they cost, and can the backlog be cleared?
Shift-by-shift staffing · coverage caps · backlog flow · incremental costOpen workspace ↗
08 / Run operations
When should a battery charge while serving a facility across peak tariffs?
24-hour dispatch · energy balance · tariff cost comparisonOpen workspace ↗
09 / Prove a claim
Which claims survive the evidence, and what exactly must be proved next?
Evidence docket · assigned proof backlog · acceptance criteriaOpen workspace ↗
10 / Prove a claim
Is a promising capability ready for an independently reviewed pilot?
Promotion hold reasons · proof jobs · attributable evidence docketOpen workspace ↗
11 / Prove a claim
Does redesigning the review pipeline improve outcomes after all stated costs?
Frozen selection · confusion matrices · costed held-out comparisonOpen workspace ↗THE ORIGINAL IDEAS, PUT TO WORK
Each original demo opens its practical workspace. The research implementation and guide remain linked inside. Choose a demo and run it locally.
Invention to operating value: Which inventions justify the cost of developing and operating them? Development portfolio · prerequisite map · verification jobs
Open practical workspace ↗
Invention to operating value: Which inventions justify the cost of developing and operating them? Development portfolio · prerequisite map · verification jobs
Open practical workspace ↗
Reproducible planning: Select a constrained enterprise portfolio, stress the downside and export reviewable Ascension jobs.
Open practical workspace ↗
Ω-Lattice capital committee: Which investments fit a $1.2M budget and 500 staff days? Investment slate · cash-flow model · downside sensitivity
Open practical workspace ↗
Local evidence review: Screen cross-sector hypotheses, allocate review time and export input-bound verification jobs and plaintext Nova-Seed drafts.
Open practical workspace ↗
Browser + simulation: Explores scenarios and Pareto search; optional browser GPT-2 performs real local text completion.
Open practical workspace ↗
Supplier resilience desk: How should 60 batches be sourced when no supplier may exceed 45%? Order allocation · landed costs · supplier-outage stress test
Open practical workspace ↗
Factory delivery control: Can seven customer orders share four resources and meet the delivery window? Resource-by-hour schedule · lateness report · priority order
Open practical workspace ↗
Factory delivery control: Can seven customer orders share four resources and meet the delivery window? Resource-by-hour schedule · lateness report · priority order
Open practical workspace ↗
Ω-Lattice capital committee: Which investments fit a $1.2M budget and 500 staff days? Investment slate · cash-flow model · downside sensitivity
Open practical workspace ↗
Offline learning lab: Learns a bounded policy from synthetic outcomes, tests it against a frozen baseline, and exports reviewable evidence.
Open practical workspace ↗
Reproducible paper research: Inspect next-open fills, cash, costs and risk decisions; compare strategies and replay every reported result.
Open practical workspace ↗
Proof Debt → AGI Jobs: Which claims survive the evidence, and what exactly must be proved next? Evidence docket · assigned proof backlog · acceptance criteria
Open practical workspace ↗
Demand & replenishment desk: How much stock should be ordered for a two-week supplier lead time? Holdout evaluation · replenishment quantity · demand stress test
Open practical workspace ↗
Second-order agency trial: Does redesigning the review pipeline improve outcomes after all stated costs? Frozen selection · confusion matrices · costed held-out comparison
Open practical workspace ↗
Second-order agency trial: Does redesigning the review pipeline improve outcomes after all stated costs? Frozen selection · confusion matrices · costed held-out comparison
Open practical workspace ↗
Reproducible program induction: Generate tasks, evolve bounded solver configurations and review a frozen candidate on separate tasks.
Open practical workspace ↗
Offline search lab: Explore competing workflow rewrites, resource schedules, held-out review gates and an exhaustive benchmark.
Open practical workspace ↗
Research planning: Train a small neural model and compare learned-model search with measured baselines.
Open practical workspace ↗
Research training: Retrieve cited evidence, compare optional local-model advice with trained search, and measure simulator outcomes.
Open practical workspace ↗
Microgrid dispatch console: When should a battery charge while serving a facility across peak tariffs? 24-hour dispatch · energy balance · tariff cost comparison
Open practical workspace ↗
Nova-Seeds pilot gate: Is a promising capability ready for an independently reviewed pilot? Promotion hold reasons · proof jobs · attributable evidence docket
Open practical workspace ↗
Product launch room: What delivery order fits a shared design, engineering and QA team? Launch sequence · resource calendar · bottleneck analysis
Open practical workspace ↗
Offline governance review: Inspect nine proposal gates and export reproducible evidence and validator jobs.
Open practical workspace ↗
Signed, review-gated workflow: Choose a feasible portfolio, schedule approved work, and produce an attributable evidence brief.
Open practical workspace ↗
Proof Debt → AGI Jobs: Which claims survive the evidence, and what exactly must be proved next? Evidence docket · assigned proof backlog · acceptance criteria
Open practical workspace ↗04 / TAKE THE NEXT STEP
Keep browser exploration light.
Move reviewed
work to a persistent, controlled runtime.
Download the wheel, installer, locked dependencies and checksums from the same release. Python 3.11–3.13 is supported.
Get the release ↗Import your downloaded mission in the console. Execute it, inspect evidence and approve a signed result. Configure models, isolated coding and wallet identity deliberately.
$AGIALPHA receipts connect approved work to recorded value. The agent checks payments and records reinvestment earmarks. Spending stays under your control.
Identity & payment setup ↗Check an exported agent result here, without uploading it. Obtain the agent’s public key through a trusted channel; copying a key from an unknown file does not establish identity.
Supports v1.4.0 exports. Verification does not establish the truth of sources or mainnet settlement.
THE ORIGINAL VISION, KEPT IN VIEW
Identify → Learn → Think → Design → Strategise → Execute. The flywheel connects evidence, useful work, human review and memory. The original diagrams, research, examples and historical implementations remain available.
This release provides bounded tools toward that vision. AGI, autonomous profit and open-ended self-improvement remain aspirations, not demonstrated capabilities.
Read the original vision & flywheels ↗CANONICAL $AGIALPHA · ETHEREUM · 18 DECIMALS
0xa61a3b3a130a9c20768eebf97e21515a6046a1fa
Alpha and AGIJobsv0 share the $AGIALPHA ecosystem. AGIJobManager provides the separate USDC route.