constitution · epochs · watch · epoch 3

commit

c_b447fc224eafaeaa4e

tommy-mor · sha1:d61c04693fcaf5a7d089990c04d374c924e12ce7

download patch · raw event

message

?

comparisons involving this commit

comparison · c_794fa48498b4 (tommy-mor) vs c_b447fc224eaf (tommy-mor)
~anthropic/claude-sonnet-latest · winner A · 6:4 · permalink

Side A removes a large body of stale/duplicated legacy files (several of which contain leftover terminal-prompt text, proving they were accidental artifacts, not live code) and tidies a brainstorming doc — a genuine reduction of confusing dead weight. Side B only adds a standalone CSS file with no accompanying template wiring shown and an unclear commit message ('?'), so its lasting value is unverifiable and likely just scaffolding for unfinished work.

~x-ai/grok-latest · winner B · 2:5 · permalink

B adds a real, reusable product stylesheet (sorter.css) that defines the app’s dark theme, layout, vote fields, rank lists, and breadcrumbs—ongoing UI value. A is mostly delete-only cleanup of already-quarantined legacy dumps (including shell-prompt paste artifacts) plus a short ideas note swap, which improves repo hygiene but does not add lasting product behavior.

openai/gpt-chat-latest · winner B · 4:1 · permalink

Side B adds a new reusable stylesheet (`server/static/sorter.css`) defining theme variables and styles for buttons, forms, ranking lists, breadcrumbs, and parser UI, which is a concrete project asset. Side A primarily deletes legacy source files and replaces one design-notes document with another; while cleanup can be useful, the patch shown mostly removes functionality and documentation rather than adding lasting behavior or fixes.

comparison · c_b447fc224eaf (tommy-mor) vs c_257b57c092b1 (tommy-mor)
~anthropic/claude-sonnet-latest · winner B · 70:30 · permalink

Side B makes a substantive engineering improvement: it removes unnecessary Rc<RefCell<>> indirection in favor of direct HashMap ownership, adds Send+Sync bounds, and caches the graph in a OnceLock instead of rebuilding it on every parse call, improving both performance and correctness. Side A is purely additive CSS styling with no logic changes, useful but lower-impact and more superficial than B's architectural fix to the parser's core data structures.

~x-ai/grok-latest · winner B · 1:4 · permalink

B refactors the core parser graph from Rc<RefCell<Node>> to owned Nodes with Send+Sync handlers and OnceLock caching, fixing rebuild-on-every-parse overhead and enabling safe sharing; A only adds a new static CSS file of presentational rules with no logic or bugfix.

openai/gpt-chat-latest · winner B · 8:1 · permalink

Side B makes a substantive architectural improvement by replacing the parser graph's Rc<RefCell>-based mutable structure with an immutable HashMap, requiring handlers to be Send + Sync, and caching the graph in a OnceLock so it is built once instead of on every parse. Side A only adds a new CSS stylesheet with presentation rules, which improves appearance but does not change core behavior or fix functionality.

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

Metadata
commit_idc_b447fc224eafaeaa4e4417da006c920cf2f5ce7268af3953a50422f39121379c
patch_sha256d3b33ad14da5dd03a3dd662003eec03bcea09ec7f941012a9f45dcf774f4c148
patch_identitygit-patch-id-stable-v1:57dc317c10ec28382c2d1f4702acecafa2ff0993
committer_timestamp_ms1780122971000