Bible Network Crypto DeFi Onchain RWA AI Agent Stablecoin Chain SAFU CryptoTax DeFAI AGI Claude Me Claude Skill Claude Design Claude Cowork
Independent Media
Not affiliated with any project
Learn Claude Skills. Do Everything Better.
claudeskill-me.com
LATEST
Claude Code Sandbox Will Not Start on Ubuntu 24.04: Checking and Fixing the AppArmor Block on bubblewrap User Namespaces  ·  Which Claude Code Effort Setting Wins: Precedence of CLAUDE_CODE_EFFORT_LEVEL, --effort, modelSettings, effortLevel and maxEffortLevel  ·  Claude Code v2.1.310 and v2.1.311: Compaction Retries a Missing Summary, ultrathink Runs at xhigh, DISABLE_COMPACT Becomes a One-Way Switch  ·  Watching Claude Code Compaction and Hooks Through OpenTelemetry: Reading the compaction, hook_execution_complete and precompute_reuse Fields  ·  Blind Spots in Bash Permission Rules: deny Misses /bin/rm, bash -c and git -C, and devbox run * Lets Any Inner Command Through  ·  CLAUDE.md Too Long? Use .claude/rules/ and paths So Rules Load Only When Claude Touches Matching Files
news

Claude Code v2.1.310 and v2.1.311: Compaction Retries a Missing Summary, ultrathink Runs at xhigh, DISABLE_COMPACT Becomes a One-Way Switch

30-Second Version · For the impatient
Neither version has one big feature; the theme is closing the edges of long and unattended sessions: compaction retries, stale files and the 200K stop behavior.

Full Explanation +
01 · Why did this happen?

Should I upgrade?

If you run long sessions or Scheduled Tasks, move to v2.1.311 or later: a missing Compaction summary is retried and a request over 200K tokens now compacts and continues instead of stopping. Short-session users have no urgent reason.

02 · What is the mechanism?

What does ultrathink do now?

Per the release notes, in the interactive terminal through Anthropic's API on Opus 5.5, Sonnet 5.5, Haiku 5.5 or Fable 5.1, a turn containing ultrathink runs at xhigh effort. The notes do not describe other environments such as Bedrock.

03 · How does it affect me?

I used a settings file to turn DISABLE_COMPACT on and off. What now?

The notes say that in settings files and in a mod's $.env.set it can now only turn Compaction off, not back on. After upgrading, check that compaction still behaves as you expect in your setup.

04 · What should I do?

I'm not an engineer. What is the one-line version?

It is like fixing the tires and the navigation on a car that makes long trips: no new engine, but fewer breakdowns halfway.

Full Content +

Claude Code v2.1.310 (2026-10-23) and v2.1.311 (2026-10-26) carry an unusual number of items about Compaction, effort and permissions. Everything below is from the official changelog, with the version marked on each item; the "what this means" section is my judgment and says so.

Compaction

  • v2.1.311: when the model's reply during compaction has no summary in it, Claude Code now usually asks for the summary once more before it replaces the conversation.
  • v2.1.311: fixed Claude seeing stale MEMORY.md and CLAUDE.md content right after automatic compaction when those files were edited earlier in the session.
  • v2.1.311: standard-context sessions that stopped with "Usage credits required for 1M context" when a request ran over 200K tokens now compact the conversation and continue.
  • v2.1.311: DISABLE_COMPACT in settings files and in a mod's $.env.set can now only turn compaction off, not back on; also fixed agent teammates compacting their history, and the rewind menu offering "Summarize from here", when DISABLE_COMPACT is set.
  • v2.1.311: /teleport and claude --teleport warn when a compacted cloud session loads from its last compaction, so the messages it summarized are not in the local copy.
  • v2.1.310: fixed compaction after a "Prompt is too long" error keeping too much of a conversation with dense text such as base64 or SVG, so the next request could fail again; memory use of non-interactive sessions (-p, Agent SDK) during compaction also improved when the transcript file is very large.
  • v2.1.311: fixed long freezes at startup and after reading or editing files when a CLAUDE.md holds a very long list item.

ultrathink and effort

  • v2.1.311: ultrathink in the interactive terminal now runs its turn at xhigh effort on Opus 5.5, Sonnet 5.5, Haiku 5.5 and Fable 5.1 through Anthropic's API; typed while Claude is working, it raises the rest of that turn.
  • v2.1.311: the ultrathink rainbow appears only where the keyword can raise effort, not when the word is quoted or part of a file path.
  • v2.1.311: /effort now warns about a prompt cache miss in plan mode when the plan model differs from the session model.
  • v2.1.310: fixed requests failing behind some gateways and compatible servers that reject the effort-2025-11-24 beta value or mid-conversation system messages; Claude Code now retries without them.

Permissions, Sandbox and hooks

  • v2.1.310: on Windows, PreToolUse and PermissionRequest hooks let a tool call through when Git Bash is missing and the hook's command is only quoted strings, such as a quoted path. Fixed.
  • v2.1.310: file read permission prompts now show where a symbolic link leads.
  • v2.1.310: on Linux, /sandbox and the sandbox startup warning now name Ubuntu's AppArmor restriction on user namespaces, and how to lift it, when it keeps bubblewrap from starting the sandbox.
  • v2.1.310: Auto Mode's prompt before the first read outside the working directories now says when blocking those reads makes shell commands ask for permission.
  • v2.1.311: SubagentStop hooks ran about twice a minute for each running Subagent; they now run once when it finishes.
  • v2.1.311: fixed subagents being unable to use MCP tools from servers that finish connecting after the subagent starts.

Other behavior changes worth knowing

  • v2.1.311: Claude Code on Bedrock, Mantle, Vertex and Foundry loads model settings from a signed file on downloads.claude.ai; set CLAUDE_CODE_MODEL_CATALOG=off to turn this off.
  • v2.1.311: fixed claude -p and SDK sessions saving transcript messages out of order or keeping messages that were taken back.
  • v2.1.311: fixed Claude Code retrying requests the API marked not to retry, which could hold a turn for minutes or, in unattended retry mode (CLAUDE_CODE_RETRY_WATCHDOG), indefinitely.
  • v2.1.310: the Monitor tool and Plugin monitors now also work on Bedrock, Vertex and Foundry, and when telemetry or nonessential traffic is disabled.

What this means (my judgment)

Neither version has one headline feature; the theme is closing the edges of long and unattended sessions: retrying a failed compaction summary, reading stale files after compaction, and the stop behavior past 200K tokens. If you run long sessions or Scheduled Tasks these are worth upgrading for. DISABLE_COMPACT becoming a one-way switch is good news for teams that use it to force compaction off, but anyone who relied on a settings file to turn it back on will find that no longer works; the notes do not say why it changed. CLAUDE_CODE_MODEL_CATALOG gives Claude Code on cloud providers one more startup dependency on downloads.claude.ai, which matters if your environment restricts outbound traffic; the notes say nothing about the file's contents or about behavior when the fetch fails, and I found no further official explanation.

Sources: Claude Code changelog - Claude Code Docs, Claude Code changelog - gradually.ai
Ask a Question
Please enter at least 10 characters
Related Articles
Claude Code Sandbox Will Not Start on Ubuntu 24.04: Checking and Fixing the AppArmor Block on bubblewrap User Namespaces
practice · Oct 27
Which Claude Code Effort Setting Wins: Precedence of CLAUDE_CODE_EFFORT_LEVEL, --effort, modelSettings, effortLevel and maxEffortLevel
advanced · Oct 27
Watching Claude Code Compaction and Hooks Through OpenTelemetry: Reading the compaction, hook_execution_complete and precompute_reuse Fields
advanced · Oct 23
Blind Spots in Bash Permission Rules: deny Misses /bin/rm, bash -c and git -C, and devbox run * Lets Any Inner Command Through
practice · Oct 23
Related News
More Related Topics