SUCCESSOR Ω: preserve the mission while models change

AGI Jobs are bounded work orders with evidence and acceptance criteria. The SUCCESSOR Ω layer connects that work to a lasting mission: its objective, executable methods, comparisons, permissions and failure history.

Start with one recurring decision and a credible alternative. Develop several candidates, inspect counterexamples, freeze one exact system and examine it. Keep the finding even when no candidate is better. A proof result, payment, knowledge admission and permission to act remain separate decisions.

Open SUCCESSOR Ω Mission Studio → starts with your real mission, current approach, credible alternative, actual requirements and public evidence. Business transformation and frontier mathematics presets help define work for an owned Computer Use runtime. Develop and improve strategies, propose a different formulation, and retain experiments and evidence. See the Mission Studio guide for the working desk, exact native worker task and reusable record.

The Computer Use laboratory provides reproducible browser execution and bounded strategy-development checks. The source mission creates a traceable evidence brief from public text. The catalog laboratory remains a supporting reproducibility example. See the research-to-usefulness guide for source identities and limits.

Run the local rehearsal

From the repository root after installing its locked dependencies with the pinned Node toolchain:

npm run successor:demo
npm run successor:verify
npm run successor:test

The demo writes a new directory under reports/successor/ and reports its path. Inspect the invoice, world and resource results, world-examination.json, fixture trust record, proof rehearsal, Mission Pack, restored record, descendant and release-evidence JSON. The verification command checks the local journey's explicit invariants; it does not certify production.

For individual steps, run npm run successor -- help. For example:

npm run successor -- mission init --mission invoice --out reports/successor/mission.json
npm run successor -- jobs compile --file reports/successor/mission.json
npm run successor -- mission run --mission world
npm run successor -- pack verify --file reports/successor/RUN_DIRECTORY/mission-pack.json

Replace RUN_DIRECTORY with the path returned by the demo. --file selects input and --out creates output without overwriting an existing file. Use a new path on another run. Proof verification and admission require a separately supplied --trust file; no demo silently provisions production trust.

Choose the next step

What the first session means

The invoice fixture investigates a synthetic C$18,700 repair bill against a C$12,400 purchase order, possible duplicate billing, changed bank details, incomplete completion evidence and possible warranty coverage. Its recommendation is HOLD_AND_ESCALATE, not a fraud finding or payment instruction. A person receives the evidence and decides the next step.

The other workbenches examine an unsafe apparent improvement and a resource-constrained research decision. A stronger comparator or different outcome frequencies can remove a claimed advantage. These negative findings become reusable knowledge.

The Node WORLD examiner freezes all 16 components of the actual synthesized candidate, binds its incumbent and strongest comparator, and reexecutes their 60 public development cases. The strongest comparator ties the challenger, so the signed internal I0 finding is FAIL: no demonstrated advantage. The bundle preserves the exact programs, case commitments, conditional calculations and missing production assurances. Its fixture keys establish local integrity only; independent admission is denied. Development cases, authored cost tariffs and nominal sampling groups do not become protected real-world proof.

The first session is public synthetic research. It does not need a wallet or production connection, and it does not establish independent Specialist ASI, realized customer Alpha or live authority. The public website cannot keep confidential evidence or protected final cases. Use the matrix for current implementation and validation limits.

Beta is capability others can rent. Spread is a possible advantage before all costs and risks. Alpha is the residual mission advantage that survives a credible comparison. The Gym develops candidates. Proof records an examination. Admission is an accountable decision. Authority is the exact current permission to act. Chronicle preserves scoped knowledge and failures. Restoring Chronicle does not restore permission.

Parcours en français

Choisissez une décision récurrente, un résultat mesurable, un responsable et une solution de référence crédible. Développez des candidats, examinez les échecs, figez un système précis, puis comparez-le. Une conclusion « aucun avantage démontré » est un résultat utile.

La facture synthétique conduit à SUSPENDRE ET SOUMETTRE AU RESPONSABLE (HOLD_AND_ESCALATE). Elle ne démontre pas une fraude et ne déclenche aucun paiement. Les résultats de répétition locale ne constituent ni une preuve indépendante, ni une admission en production.

Les modèles peuvent changer. La mission, les méthodes autorisées, les preuves et les leçons peuvent être préservées. Une restauration ou un descendant commence sans autorité de production héritée.

Continue an existing Alpha dossier

The native project receiver preserves the exact original project and its failed or rejected work. It compiles unsigned, unfunded AGIALPHA work orders and executes the bounded public-data task locally. Its versioned project work envelope is separate from the Python Alpha market work-spec namespace. Fleet execution and internal checking, operational admission, independent review and on-chain settlement remain separate decisions.