Skip to main content

Read the result in its research context.

Follow the published construction into its registered claims, then explore reviewed research questions, attempts and the connections that remain open.

Established in scope

Source-indexed local AI search corrects a computed false prediction

In the question: Automatic joint origin hypotheses and source responses

Executed software path and static inventory coverage. Not renewed proof audit of707 modules,arbitrary automatic physics judging,billion-candidate run or complete origin source. Four curated questions;whole hypotheses remain unverified and are not imposed as constraints. R103 extends the live driver; historical run metadata retain their original code hashes. Original six checker mappings were reviewed and retained by targeted regressions; no fresh R101/R102 model replay is claimed.

Read the exact result, assumptions and evidence

Result

All707 registered original Python sources statically parsed with707 docstrings,7597 check sites,zero missing/parse errors;788 ledger rows and718 Atlas findings retained. Existing Ollama client supports optional response schema. Actual two-round qwen3.6:35b run first predicted higher n allowed under primitive projection; exact checker contradicted it; next round corrected. Prior invalid/false proposals remain saved. Adaptive question selection rechecks both existing full processes and2245 moments;ten affine-correlated questions have rank1. Seven targeted regressions passed.

Required assumptions

  • AI proposals with static original-source inventory and curated controls

    Current registry/ledger/Atlas and statically read original scripts serve retrieval. The installed local Ollama client proposes structured hypotheses and explicit predictions for fixed curated checkers. Only named consequences are computed and fed back. Known source IDs, schema validity and passed controls do not prove a whole hypothesis or physical origin.

  • Do not identify source types

    Observable algebra, OS reflection, state, collar operator, time evolution, carrier connection, family holonomy, and readout are distinct data until an original map relates them.

Original evidence

  • GENESIS-SYNTHESIS-20261102.txt lines 1–177

    Source checked at this snapshot

    Original location and fingerprint

    experiments/theory-contracts/seam-positive-kernel-search-20261015/GENESIS-SYNTHESIS-20261102.txt

    5b86ff944370aa4fbf65ff712b92b581dfa1a1b80cccd6c928217e63057961b0

  • origin_hypothesis_search.py lines 200–396

    Source checked at this snapshot

    Original location and fingerprint

    experiments/theory-contracts/seam-positive-kernel-search-20261015/origin_hypothesis_search.py

    9ca8da3ea6388229581a548ca5dafa129fd1df1a2f1c673b8299d6c331e9f11e

  • registry_ast_inventory.py lines 1–131

    Source checked at this snapshot

    Original location and fingerprint

    experiments/theory-contracts/seam-positive-kernel-search-20261015/registry_ast_inventory.py

    01d4a058b02a4d5c0f4b070519c5b62e7a5045aace12a32123d1c23a58a4d904

  • adaptive_discriminator.py lines 1–132

    Source checked at this snapshot

    Original location and fingerprint

    experiments/theory-contracts/seam-positive-kernel-search-20261015/adaptive_discriminator.py

    3fe77c3a771d51b43ac9b129f0e88dae938c706275680b2134389cc10022028a

finding:r101-source-indexed-ai-loop-uses-executed-counterexample-feedback