Journalist / shipped evidence

What changed.

Verified deployments, completed measurements and applied design changes. Curated every two hours from persisted records, with direct links to evidence.

Where we want to go → · JSON API

Published 18 Oct 2026 · 18:43 UTC

Measurement completed: OCRBench v1 / full serial reference

Qwen/Qwen2-VL-2B-Instruct completed 1000 examples (full split). Metric: accuracy. This records execution, not an independent validation or a global ranking.

Inspect evidence →

Event: 18 Oct 2026 · 18:33 UTC

Published 18 Oct 2026 · 18:43 UTC

Measurement completed: IAM lines / full 2,915-line test

microsoft/trocr-large-handwritten completed 2915 examples (full split). Metric: cer. This records execution, not an independent validation or a global ranking.

Inspect evidence →

Event: 18 Oct 2026 · 18:08 UTC

Published 18 Oct 2026 · 18:43 UTC

Site release published

The live CodeSOTA web application is serving release 20261018T182148094876Z. This deployment record does not establish which features changed.

Inspect evidence →

Event: 18 Oct 2026 · 18:21 UTC

Published 18 Oct 2026 · 16:44 UTC

Usability agent: first recorded completed cycle

Publishes page-usability findings as an open improvement backlog. A recorded completed cycle demonstrates execution, not the correctness of every recommendation.

Inspect evidence →

Event: 18 Oct 2026 · 13:17 UTC

Published 18 Oct 2026 · 16:44 UTC

Measurement completed: BoolQ (full validation, raw likelihood accuracy)

SlayerLab/Slayer149-Balanced-ARC-E-ft-v2 completed 3270 examples (full split). Metric: accuracy. This records execution, not an independent validation or a global ranking.

Inspect evidence →

Event: 17 Oct 2026 · 16:02 UTC

Published 18 Oct 2026 · 16:44 UTC

Measurement completed: HellaSwag (full validation, character-normalized accuracy)

Qwen/Qwen3.5-0.8B completed 10042 examples (full split). Metric: accuracy_char_normalized. This records execution, not an independent validation or a global ranking.

Inspect evidence →

Event: 17 Oct 2026 · 16:11 UTC

Published 18 Oct 2026 · 14:44 UTC

Measurement completed: BoolQ (full validation, raw likelihood accuracy)

Qwen/Qwen3.5-0.8B completed 3270 examples (full split). Metric: accuracy. This records execution, not an independent validation or a global ranking.

Inspect evidence →

Event: 17 Oct 2026 · 17:03 UTC

Published 18 Oct 2026 · 14:44 UTC

Designer agent: first recorded completed cycle

Reviews page feedback and can apply bounded visual settings. A recorded completed cycle demonstrates execution, not the correctness of every recommendation.

Inspect evidence →

Event: 18 Oct 2026 · 13:09 UTC

Published 18 Oct 2026 · 14:44 UTC

Site release published

The live CodeSOTA web application is serving release 20261018T133952568623Z. This deployment record does not establish which features changed.

Inspect evidence →

Event: 18 Oct 2026 · 13:39 UTC

Published 18 Oct 2026 · 13:17 UTC

Measurement completed: HellaSwag (full validation, character-normalized accuracy)

NamanAgnih0tri/AlphaRoute-0.8B-v1.5 completed 10042 examples (full split). Metric: accuracy_char_normalized. This records execution, not an independent validation or a global ranking.

Inspect evidence →

Event: 17 Oct 2026 · 16:59 UTC

Published 18 Oct 2026 · 13:17 UTC

Readability settings updated

Design version 1 applied bounded typography, table-spacing or desktop-width settings. Live settings are public; browser visual QA is not implied.

Inspect evidence →

Event: 18 Oct 2026 · 13:09 UTC

Published 18 Oct 2026 · 13:17 UTC

Site release published

The live CodeSOTA web application is serving release 20261018T131642771909Z. This deployment record does not establish which features changed.

Inspect evidence →

Event: 18 Oct 2026 · 13:16 UTC

Changelog entries are immutable server-provided facts, not agent-written claims. Selection does not validate scientific conclusions.

Feedback about this page

Saved privately with this page URL. Do not include passwords or personal data. Network identities used for spam limits are hashed; raw IP addresses are not stored with feedback.