# Public Gooo 0.6.14 and complete source-fill use — 2026-10-09

[Public release](https://github.com/kimjooyoon/meta-ontology-go/releases/tag/v0.6.14-dev)
uses clean source `2162809f3c49637f0db5d1003873db5e8189d16e`, Go 1.27.1 and
decision runtime v0.2.26-experimental. The real macOS arm64 archive was downloaded,
verified and installed. Its SHA-256 is
`f204cd7dd80d2282efdead2381019b19b0ae9c9c4cbb6a617d3756b8f8750a0e`.
The installed bytes and source identity match that public archive.

## What the installed file did

| Original program / mode | Program attempts | Original final cases | New model predictions |
| --- | ---: | ---: | ---: |
| Budget, two attempts | 2 | 1/4 | 0 |
| Budget, three attempts | 3 | 4/4 | 0 |
| Budget, saved replay | 3 | 4/4 | 0 |
| Mixed, fixed order | 24 | 4/4 | 0 |
| Mixed, own graph model | 3 | 4/4 | 1 |
| Mixed, saved replay | 3 | 4/4 | 0 |

Two related programs, four fresh constructions and two saved replays reuse the
original cases. Exact JSON numbers preserve values above 2^53. The partial result,
all attempted condition/value assignments and separate local holdout outcomes are
retained. Fixed/model order selects the same ordinary Gooo source. Both selected
ordinary sources also reproduce their four final outputs using `body-compose`,
with zero runtime predictions. This source reuse is additional execution of the
same programs, not independent model accuracy evidence.

The unchanged own graph model orders record choices; fills use deterministic
source order. Metadata SHA-256 is
`3c68205a660695103712e2115ce90accd298b406dced8ad5ad5c9b84bab0a202`, weights
`76f68845a03ed8c8bc261a57c919e96dcdd352d3c35bcdaeb3d9b51a88975c6f`.
No new weights or programs were trained for these release observations. Host CPU
utilization was not sampled and these runs are not a controlled speed benchmark.

With the installed public compiler, the workbench's Gooo feedback loop retains
one original counterexample and makes four rounds / seven program attempts.
Final holdout is 4/4, separate fill holdout 1/1, and final evaluation performs zero
new predictions. Original sources, inputs, expectations and counterexample rows
agree with the downloaded workbench CI outputs.

## Source, CI and publication

- [Compiler PR 1392](https://github.com/kimjooyoon/meta-ontology-go/pull/1392): tested
  `ce78a0ab5b113929a39a0e888a9b9a7f8a600bf0`, merged `2162809f...`, equal file trees.
  [Authoritative CI 37820995742](https://github.com/kimjooyoon/meta-ontology-go/actions/runs/37820995742)
  passed all six canonical jobs. Original proof artifact `11570890887` SHA-256
  `180ca28e66676e065d2038bd67b1283a7fe99013983654543eaab510741b578f` was checked,
  and the proof verifier plus live PR tuple passed before expected-head merge.
- [PR readiness 37820995379](https://github.com/kimjooyoon/meta-ontology-go/actions/runs/37820995379)
  and [merged readiness 37823798635](https://github.com/kimjooyoon/meta-ontology-go/actions/runs/37823798635)
  each passed Linux amd64, Windows amd64 and both macOS architectures, with 7/7
  eligibility cells closed. Twelve source-fill records per run reproduce the
  local candidate's exact assignments and partial/complete/replay outputs. Original
  ZIP digests were verified before extracting receipts and raw outputs.
- [Publication 37824796087](https://github.com/kimjooyoon/meta-ontology-go/actions/runs/37824796087)
  passed. The annotated tag targets the merged source. All nine asset identities
  and eight checksums agree; downloaded metadata and macOS bytes were checked.
- [Workbench PR 37](https://github.com/kimjooyoon/gooo-ecosystem-workbench/pull/37): tested
  `77d99133f61269d31ef76786112e3d0d63742f92`, merged
  `9b636ab047b9b60272d0a79b2dfa417b028ea2f3`, equal file trees.
  [CI 37823861494](https://github.com/kimjooyoon/gooo-ecosystem-workbench/actions/runs/37823861494)
  passed. Original artifact `11570607732` SHA-256
  `6101cf22f38fbcd8c0108cc2918b413b58b38b7ae87bf69d0b628306e0bf7173` was checked.
  Retained source-fill outputs were recounted against both prior observations and
  the installed 0.6.14 output. The prior verification's 0.6.13 reuse line refers to
  the earlier observation; `public-verification.txt` verifies the new installation.

## Retained records

`summary.json`, `build.json`, verification logs, original platform receipts/raw
outputs, proof, public metadata and actual installed observations accompany this
page. Executable archives and model weights are excluded. `published-observe.go`
recounts exact original values against the candidate record; `compare-platforms.go`
compares all twelve native records for a supplied source SHA. Run
`verify-workbench.go` from the workbench checkout with extracted CI and installed
observation roots. Source-fill inputs are the unchanged compiler/workbench fixtures.
The installed-observation archive contains the workbench loop under `workbench-loop`.

[Usage](https://github.com/kimjooyoon/meta-ontology-go/blob/v0.6.14-dev/docs/releases/0.6.14-dev.md)
· [frozen release plan and candidate observations](https://github.com/kimjooyoon/meta-ontology-go/tree/ce78a0ab5b113929a39a0e888a9b9a7f8a600bf0/docs/research/release-0614-20261009).

Workbench [post-merge main CI 37825386735](https://github.com/kimjooyoon/gooo-ecosystem-workbench/actions/runs/37825386735)
also passed on `9b636ab047b9b60272d0a79b2dfa417b028ea2f3`; its run metadata is retained.

Main promotion [PR 1393](https://github.com/kimjooyoon/meta-ontology-go/pull/1393) passed [CI 37823815993](https://github.com/kimjooyoon/meta-ontology-go/actions/runs/37823815993). Original proof artifact `11572220350`, SHA-256 `dd2e6a298a56c649379a97332ff993539d463126a645c5ca7d54aa90138d15a0`, and digest-bound promotion authorization passed verification. Live dev, main, candidate tree and parent were rechecked. GitHub accepted the permitted expected-head squash merge into main `27f81c5901cdbe0ce61263cfe5399eb112daf17d`; its tree equals released dev `2162809f3c49637f0db5d1003873db5e8189d16e`. The initially rejected merge-commit strategy is retained in the record.