New binary strings suggest Anthropic is testing a `claude-opus-5-5` model, along with deta

EXHIBIT 1544 · FLASH

Sep 22, 2026 18:40 UTC binary @anthropic-ai/claude-code-darwin-arm64

New binary strings suggest Anthropic is testing a `claude-opus-5-5` model, along with detailed internal tooling for Claude in Chrome, sandbox, and skill management.

✓ VERIFIED — BEAT THE ANNOUNCEMENT BY 20.7H · AUTO-VERIFIED AGAINST ANTHROPIC NEWS

READY TO POST:

anthropic's latest binary shows 'claude-opus-5-5' model ids popping up. could this be a new frontier model in testing? 👀

looks like anthropic is building out some serious internal dev tools, with strings hinting at deep integration for claude in chrome, sandbox controls, and skill management. very interesting.

⚠️ FROM COMPILED BINARY — lower confidence, verify before posting. (May include bundled-dependency strings, not just first-party changes.) @anthropic-ai/claude-code-darwin-arm64 2.1.278 -> 2.1.280: +20113 strings, -16494 strings, 73 interesting [model_id] (5) + /$bunfs/root/claude-code.d.ts-3df8328a.txt.zst + anthropic.claude-opus-5-5 + claude-opus-5-5 + claude-opus-5-5[1m] + us.anthropic.claude-opus-5-5 [feature_flag] (48) + ${e?A:"Every message you send is to the user."} Worker results and system notifications are internal signals, not conversation partners \u2014 never thank or acknowledge them. Summarize new information for the user as it arrives. + <${m1}>No completion record was found for them in the previous session. ${h} Task ids in this notification beginning with "${Ue}" are internal scan markers, not tasks.</${m1}> + @internal How a self-hosted runner's session process ended, on the result the runner synthesizes for it.$ + @internal How long the worker waits for an announcement; a client that receives the request later than this after the session service stored it does nothing. + @internal Requests the browsers connected to the session's Claude in Chrome server, as the terminal's /chrome browser picker lists them. Fails when the server is not connected.3 + @internal Requests the conversation as the plain text the terminal's /export writes, with the file name that command would offer, for a host that copies or saves it with its own widgets. + @internal Requests the rows of the terminal's /skills menu, for a host that draws the menu with its own widgets: each skill with its source, its token estimate, its effective state, and what locks it. Read-only; a host writes a state through the CLI's edit-skill-overrides subcommand.J + @internal Requests the rows of the terminal's /status screen (version, session, account, provider, model, servers, settings), every value already rendered as text, for a host that shows them with its own wi