constitution · epochs · watch · epoch 3

commit

c_efa306aa8040b0413f

tommy-mor · sha1:d1ad77cf7e2daa6979a5612b266e3b6ad03f07b2

download patch · raw event

message

Add transparent HTML evidence graph for epochs and rankings.

Persist verbatim commits, comparisons, attempts, and judgments in the ledger, serve them as linkable HTML indexes, and keep ranking resumable across restarts.

Co-authored-by: Cursor <cursoragent@cursor.com>

comparisons involving this commit

comparison · c_5bc77fbfad41 (tommy-mor) vs c_efa306aa8040 (tommy-mor)
~anthropic/claude-sonnet-latest · winner B · 9:1 · permalink

Side B introduces a substantial, well-tested feature: a content-addressed, hash-chained evidence ledger with resumable ranking, HTML index pages, credential-stripping, and comprehensive new tests (test_evidence.py, integration.clj updates), representing real architectural and durability improvements. Side A is a small, narrowly scoped UI tweak adding a view-count parameter to one page with a matching test, useful but far more limited in scope and lasting impact.

~x-ai/grok-latest · winner B · 1:25 · permalink

A only wires an existing view-count into the vote-compare layout (a few call-site lines plus assertion updates). B adds a content-addressed Evidence ledger, idempotent append/hash-chaining, resumable ranking that skips duplicate LLM calls, and a full crawlable HTML evidence graph (epochs/commits/comparisons/attempts/judgments) with substantial tests—core lasting transparency infrastructure versus a minor display tweak.

openai/gpt-chat-latest · winner B · 9:1 · permalink

Side B introduces a substantial evidence and audit infrastructure: it adds versioned append-only Evidence events, idempotent persistence with hash chaining, resumable LLM ranking, HTML evidence browsing, richer audit links, and extensive integration/unit tests covering replay, downloads, and legacy behavior. Side A is a useful but localized feature that threads existing view-count support into the vote comparison page and adds tests verifying the displayed count increments.

comparison · c_efa306aa8040 (tommy-mor) vs c_794fa48498b4 (tommy-mor)
~anthropic/claude-sonnet-latest · winner A · 9:1 · permalink

Side A adds a substantial, tested feature (content-addressed evidence ledger, resumable ranking, HTML evidence graph, integration/unit tests) that materially extends the project's transparency and durability guarantees. Side B is purely deletion of dead/legacy files and one idea-note replacement—housekeeping with no functional value and no tests.

~x-ai/grok-latest · winner A · 95:5 · permalink

Side A adds a durable, tested evidence ledger (content-addressed Evidence events, idempotent append, ranking resume, byte-faithful blobs) plus public HTML indexes for epochs/commits/comparisons/attempts/judgments—core product transparency. Side B only deletes unused legacy sources and swaps an ideas note, removing dead weight without improving the live system.

openai/gpt-chat-latest · winner A · 20:1 · permalink

Side A adds substantial, lasting functionality: a versioned append-only evidence ledger, idempotent persistence with resumable ranking, content-addressed evidence records, HTML evidence browsing endpoints, richer audit links, API enhancements, and extensive integration/unit tests. Side B is primarily repository cleanup, deleting legacy files and replacing one ideas document, with little evidence of new project functionality or bug fixes.

The full patch is loaded only by the download route: download patch

Metadata
commit_idc_efa306aa8040b0413fac070eaca49d0608bbe26260cb526bad7a6f02c4a3712d
patch_sha25623851ac944fec497c41286026a4c032b011567a756b2d7971336efb06f3950ac
patch_identitygit-patch-id-stable-v1:0b9750cbcf988bc380573ef46962681c11813521
committer_timestamp_ms1784419783000