Prewire · case file · public copy of the record
FILE P364 PREWIRE receipts archive the live wire
CATCH #1364 DETECTED SEP 15, 2026 01:08 UTC
THE CATCHscore 7/10
EXHIBIT 1364 · URGENT
Sep 15, 2026 01:08 UTC binary @anthropic-ai/claude-code-darwin-arm64
Anthropic's Claude Code binary for Darwin ARM64 was updated, revealing a new model ID `claude-opus-4-5-20251101` and several internal feature flags related to worker behavior, beta features, and tool use.

READY TO POST:

anthropic's latest claude code binary has a new model id: `claude-opus-4-5-20251101`. could this be a future opus variant or just an internal codename?

interesting feature flags spotted in the new claude code binary, including 'dangerous-tool-use' beta and 'thinking-resumption'. seems like anthropic is testing some advanced internal capabilities.

EVIDENCE — the receipts
⚠️ FROM COMPILED BINARY — lower confidence, verify before posting.
(May include bundled-dependency strings, not just first-party changes.)

@anthropic-ai/claude-code-darwin-arm64 2.1.270 -> 2.1.272: +12991 strings, -10574 strings, 29 interesting

[model_id] (2)
  + /$bunfs/root/claude-code.d.ts-21b4def3.txt.zst
  + claude-opus-4-5-20251101

[feature_flag] (17)
  + ${e?Mt:"Every message you send is to the user."} Worker results and system notifications are internal signals, not conversation partners \u2014 never thank or acknowledge them. Summarize new information for the user as it arrives.
  + <${xW}>No completion record was found for them in the previous session. ${k} Task ids in this notification beginning with "${Ne}" are internal scan markers, not tasks.</${xW}>
  + @internal Emitted when a scheduled task (cron) fires. content is the render text. From internal SystemMessage 'scheduled_task_fire'.
  + @internal Present, and true, only when this worker uses the relay_marker member of a turn_handoff request; a client sends that member to no other worker.
  + @internal Success payload answering turn_handoff.
  + @internal The calls queued to run (status accepted).1
  + @internal Why org policy or managed settings lock `remoteControlAtStartup`; absent when allowed. Fails closed while the policy is unverified.
  + This build of @anthropic-ai/bun-internal has no Bun.ant.CellSegmenter; src/ink needs bun-internal >= the version pinned in package.json.
  + [reminders] the server rejected the clear_at beta header or field (HTTP 400); retrying without them. For the rest of this session earlier reminders are not re-sent, and requests that can use a Message Threads thread get no reminder.
  + [server-classifier] an unrecognised HTTP 400 arrived while the dangerous-tool-use beta was on the request; retrying once without it, and auto mode decides with the local classifier for the rest of this conversation
  + [server-classifier] the platform rejected a beta on this request (HT

caught by the 20-minute sweep · the live wire · all receipts

THE WIRE RECEIPTS ACCESS @VEDOLOS TERMS PRIVACY break the story before it breaks