The "slow response" hint is calmer and more patient. When the model takes a while to start streaming, Claude Code used to flash "No response from API · Retrying in …" after 10 seconds — which read like an error even when nothing was wrong. v2.1.185 reframes it as "Waiting for API response · will retry in …" and only shows it after 20 seconds of silence. Net effect: a momentarily slow response no longer looks like a failure, and you get fewer false-alarm "retrying" flashes on a healthy connection. No config or behavior change beyond the wording and timing.
Last updated: 2026-06-21
2026-06-20
Claude Code
v2.1.185
2026-06-21
NemoClaw
commits (internal)
Changelog — June 21, 2026
A quiet day after yesterday's big Hermes v0.17.0 release. Claude Code shipped v2.1.185, a one-line UX fix: the hint you see when a response is slow now reads "Waiting for API response · will retry in …" instead of the more alarming "No response from API · Retrying in …", and it waits 20 seconds of silence before appearing instead of 10. NemoClaw shipped internal refactor, test-migration, and CI work only — still v0.0.65.
No new release (still v0.0.65). The last 24h on main were entirely internal: a fresh batch of live end-to-end tests migrated from shell scripts to typed Vitest scenarios (Kimi compatibility, Hermes inference switch, Brave search, device-auth health, stale-sandbox upgrade), a CI change that auto-retries flaky live E2E runs twice before failing (#5549), and more of the ongoing onboarding/registry nullability cleanup (#5566, #5563, #5562). One contributor-facing tweak: the PR Review Advisor now labels findings by urgency (required fix / resolve-or-justify / in-scope improvement) instead of softer language (#5567). No user-visible behavior change.
See all releases
Browse the full changelog index for the complete history across all platforms, or the daily one-liner for the most recent state of each agent.