Repository navigation
Rerun the five-harness screen on the 0.7.0 daemon - #461
Merged
Merged
Conversation
The README's comparison table, EVIDENCE.md and HARNESS_MEASUREMENTS.md now come from a 2026-10-30 rerun at main 2182b4a, after the night's daemon and performance changes, with every harness rerun on the same host kind and the same pinned versions. Agent's top case is unchanged within drift (28.9 MiB, 0.65 s CPU, 599 ms turn p99 against 28.7 MiB, 0.69 s and 608 ms on 10-28). The other engines' binaries did not change, yet moved by up to about 20% between a trial run and the final one an hour and a half apart, so the docs now say to compare harnesses within one screen. The 10-28 table is kept as the previous screen. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_013jkdGF8c2soAtTDhmz5uha
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
Merged
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Requested by George · project thread
Docs only. The README's "How fast is it?" table, the EVIDENCE.md bullet and HARNESS_MEASUREMENTS.md now come from a rerun of the five-harness screen at main
2182b4a, after tonight's daemon and performance PRs (#425, #434, #436, #447, #450, #458, #460 and the rest). Every harness was rerun on the same 4-vCPU Linux VM kind and the same pinned versions as the 2026-10-28 screen (Pi 0.85.1, Codex 0.153.1, opencode 1.18.32, Claude Code 2.1.267).32 agents, 64 KiB per turn, medians of three runs:
* below the configured concurrency (Codex 8–10, Claude Code 26–31 of 32).
759c0db90 minutes earlier had Pi 155, Codex 247 and Claude Code 6,577 MiB. That is host drift of up to about 20%, so the README and HARNESS_MEASUREMENTS.md now say to compare within one screen, not across screens. The 10-28 table is kept in HARNESS_MEASUREMENTS.md as the previous screen.comparerefuses an engine below its concurrency (this time Codex at 8 agents with 64 KiB), so the c32 cases were run by hand with the matrix's ownbench runflags and engine order, as the 10-28 screen did for its last case.Not rerun: the September live fleet check (real providers) and the Pi Durable lifecycle comparison.
Generated by Claude Code