Source Intelligence
Follow Discord

DisclaimerUnofficial, and not affiliated with Anthropic. Nearly all of this is read straight out of what ships: npm bundles, captured prompts, published docs. Anthropic's own notes go in verbatim, marked as theirs. The rest is my reading, and every entry carries the strings behind it. If one looks wrong, vote it down and say why.

← All releases Home ←olderv2.1.98 v2.1.101newer→

Claude Code v2.1.100

6 entries read diff v2.1.98 → v2.1.100 Markdown

This release adds reassuring status messages during extended thinking periods, relaxes the stall detection thresholds so the spinner no longer shows a "stalled" state prematurely, and tightens several system prompt sections to produce more concise responses. Opus 4.6 users with a server-side flag may also see stricter word-count limits on responses.

Find
Pick an entry · j / k steps through
2 entries

New Featuresopen

Extended Thinking Status Messages#

What

During long thinking periods, Claude Code now displays timed reassurance messages so you know it's still working.

Details
  • After 30 seconds of thinking: "Thinking a bit longer… still working on it…"
  • After 90 seconds: "This is a harder one… it might take a few more minutes…"
  • After 270 seconds (~4.5 minutes): "Hang tight… really working through this one…"
  • Messages appear as dimmed text below the spinner and disappear when thinking ends
  • Only displays when the model is in the "thinking" state (extended thinking / chain-of-thought)
Evidence

Thinking status message array with afterMs timings (search for "Thinking a bit longer")

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

Related

Other releases about the same thing. Found by shared names or similar wording; neither means one caused the other.

Numeric Length Anchors for Opus 4.6 [Gradual Rollout]#

What

A new system prompt section that enforces strict word-count limits on Claude's text output when using Opus 4.6.

Details
  • Text between tool calls is limited to ≤25 words
  • Final responses are limited to ≤100 words unless the task requires more detail
  • Only active when using the Opus 4.6 model AND the server-side quiet_salted_ember flag is set to "true"
  • Injected as a numeric_length_anchors prompt section in the system prompt
Evidence

Length anchor prompt section (search for "numeric_length_anchors" and "≤25 words")

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

4 entries

Improvementsopen

Relaxed Stall Detection During Thinking#

The stall detection threshold was increased from 3 seconds to 10 seconds, and the intensity ramp-up period was extended from 2 seconds to 10 seconds. Additionally, the thinking state is now explicitly treated as an active state for stall detection purposes. Together, these changes mean the spinner no longer incorrectly indicates a "stalled" connection during legitimately long thinking periods.

Evidence

Stall threshold change from 3000 / 2000 to 1e4 / 1e4 (search for j > 1e4 in the stall detection function k87), and added q === "thinking" condition in the calling code

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

More Concise End-of-Turn Summaries#

The system prompt guidance for end-of-turn summaries was simplified from "state what changed and what's next. That's it — no recapping the journey, no restating the problem, no listing everything you considered" to the more direct "one or two sentences. What changed and what's next. Nothing else."

Evidence

Updated summary instruction (search for "End-of-turn summary: one or two sentences")

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

Shorter Exploratory Question Responses#

When the user asks an open-ended or exploratory question, Claude is now instructed to respond in 2–3 sentences with a recommendation and the main tradeoff, rather than providing extended analysis with multiple options and tradeoffs. This makes exploratory exchanges snappier while still waiting for user agreement before implementing.

Evidence

Updated exploratory question prompt (search for "respond in 2-3 sentences with a recommendation")

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

Removed Verbose Communication / Output Efficiency Prompt Section#

The standalone "# Communicating with the user" prompt section (for Opus 4.6 users) and the "# Output efficiency" prompt section (for other users) were both removed from the system prompt assembly. Their guidance is now covered by the tighter communication style instructions and the new numeric length anchors.

Evidence

Both "Communicating with the user" and "Output efficiency" strings are absent from v2.1.100 (confirmed via search)

Strings lifted out of the shipped bundle, so the claim above can be checked against them.

Verbatim
No official entry

Not individually listed upstream

v2.1.100 does not have its own entry in Anthropic’s official changelog, because prerelease builds are often folded into a neighbouring release. Browse the full changelog for the closest official notes. Everything else on this page came out of the bundle instead, which is why the two lists don't match.

Read it on GitHub →
System prompt

The system prompt was not captured for this release, so this page cannot say whether it moved.

Which releases were captured →