43 wires and counting

$ follow OpenClaw

Keep up with OpenClaw in about 3 minutes: what actually shipped — the commits, pull requests, releases, and security advisories that matter.

or

fair warning: these emails are deeply technical. diffs, version numbers, CVEs, benchmark deltas. if that's not your idea of a good read, this isn't your newsletter.

Folds into your digest — weekly by default, monthly if you prefer. Unsubscribe in one click.

$ status

wire 2026-05-28
stories 57

© 2026 RepoJournal Home Showcase Explore How it works Privacy

$ the-wire · showcase

Preserving signed thinking payloads across providers

By RepoJournal · Filed · About OpenClaw

OpenClaw now preserves provider-owned signed thinking text byte-for-byte, fixing the regression where replay altered Anthropic thought blocks.

The most consequential change is a fix to thinking payload handling in agent transports [1]. Previously, signed Anthropic thinking text and Bedrock reasoningText were not preserved byte-for-byte on transport ingest and replay, which could corrupt replay fidelity. Now those provider-owned payloads are preserved, while unsigned fallback text and synthetic reasoning_content blocks keep their sanitation. This closes issue #87459 and affects any flow that streams or replays messages through Anthropic or Bedrock transports. If you depend on exact replay of reasoning, this is the fix you were waiting for.

A refactor centralizes the OpenAI Codex OAuth flow by removing duplicated browser/token code, keeping the existing OpenAI plugin OAuth runtime as the single implementation [2]. No user-visible OAuth behavior change is intended, and legacy src/llm login/refresh behavior is preserved, including abort propagation and callback handling. If you maintain forks or custom OAuth wrappers, you should verify nothing depends on the old duplicated paths.

CI gains phase-level timing instrumentation in scripts/build-all.mjs so `pnpm build:ci-artifacts` logs each build phase duration with a slowest-first summary [3]. A duplicate `Build Control UI` step was removed. This is instrumentation only; optimization is deliberately deferred until the timing data points to the real bottleneck. Expect clearer CI logs, not faster builds yet.

Across docs and packaging, the day is routine: five docs sync commits mirror upstream changes [4][5][6][7][8], and CrabBox released 0.21.0 and opened the 0.21.1 changelog [9][10].

Action items

References

  1. [1] fix(agents): preserve signed thinking payloads ↗ openclaw/openclaw
  2. [2] refactor(openai): centralize Codex OAuth flow ↗ openclaw/openclaw
  3. [3] perf(ci): instrument build artifacts phases - add timing ↗ openclaw/openclaw
  4. [4] chore(sync): mirror docs from openclaw/openclaw@d1577a2ff2830bb151cb820af5ebbb9323ef685e ↗ openclaw/docs
  5. [5] chore(sync): mirror docs from openclaw/openclaw@7c7fb7df678b1cb7c3b76564a591ad2b3ba0de23 ↗ openclaw/docs
  6. [6] chore(sync): mirror docs from openclaw/openclaw@7299c5695317f74ddbe4a1a68efafd723a38c6c1 ↗ openclaw/docs
  7. [7] chore(sync): mirror docs from openclaw/openclaw@bb752c2b4701f6742550c1e8d56303d9b04e87ab ↗ openclaw/docs
  8. [8] chore(sync): mirror docs from openclaw/openclaw@a7d2d9c6df002d718badc5500cf52f516d97d248 ↗ openclaw/docs
  9. [9] chore: release 0.21.0 ↗ openclaw/crabbox
  10. [10] chore: start 0.21.1 changelog ↗ openclaw/crabbox

Quick answers

What shipped in OpenClaw on May 28, 2026?
OpenClaw now preserves provider-owned signed thinking text byte-for-byte, fixing the regression where replay altered Anthropic thought blocks. In total, 53 commits and 4 pull requests landed.
Who contributed to OpenClaw on May 28, 2026?
4 developers shipped this update, including RomneyDa, joshavant, openclaw-docs-sync[bot], and Peter Steinberger.
What were the notable OpenClaw updates?
fix(agents): preserve signed thinking payloads, refactor(openai): centralize Codex OAuth flow, and perf(ci): instrument build artifacts phases - add timing.