Claude Code v2.1.288 Ships Resume Fixes, Ctrl+C Draft Recovery, and /code-review --max-findings
Claude Code v2.1.288 is a bug-fix release, but the fixes land on the parts of the CLI that hurt most: session resume, mid-turn timeouts, and plugin loading. If you've been burned by --resume losing context, this version is worth pulling.
Session resume actually works now
\nSeveral fixes target resume reliability, which had a cluster of overlapping failure modes:
\n- \n
--resumeno longer drops files and other context that a compaction had just restored. \n- A resumed session now saves the last response of a turn — previously the next
--resumeshowed the prompt unanswered. \n - Resume won't load a transcript cut short when the same session rewrote the file during the load. \n
- Resuming a conversation started on 2.1.286 or earlier no longer drops the model's earlier thinking. \n
Prompt and timeout behavior
\nIf you clear a prompt with Ctrl+C, pressing Up on the empty prompt now brings the draft back — including pasted text and images. Mid-response API timeouts no longer fail the turn: non-interactive sessions and subagents continue from the partial response, and thinking-only responses are retried.
\nA separate fix: long conversations that reported zero token usage on the last reply now auto-compact instead of failing with \"Prompt is too long\".
\n\nNew flags and keybindings
\n- \n
/code-review --max-findings <n>|all— report more or fewer findings than the usual limit. The choice is reused until you pass--max-findings default. \nCtrl+Ffinds a session by name;Alt+↑/↓jumps between groups in the agents view. Both, plus rename, can be rebound inkeybindings.json. \n- Mods get
$.ui.selection(), returning the last selected text in fullscreen mode and, when the selection lies within one transcript row, that row. \n
Plugins and cloud sessions
\nPlugin fixes are extensive: LSP servers no longer receive literal ${user_config.*} and ${CLAUDE_PLUGIN_ROOT} placeholders in initializationOptions and settings; git-subdir installs no longer fail or cache an incomplete plugin on git before 2.39 (e.g. Ubuntu 22.04's 2.34); and plugins loaded with --plugin-dir now show \"Configure options\" in /plugin.
For cloud sessions without the GitHub CLI, a built-in gh api is now available. Gateways that reject structured outputs (Mantle, some proxies) can set CLAUDE_CODE_DISABLE_STRUCTURED_OUTPUTS to turn them off — this fixes session titles, memory recall, and prompt hooks failing behind those gateways.
Permissions and sandboxing
\nSandboxed heredocs with an unquoted delimiter (python3 <<EOF) no longer ask for approval on every run under sandbox auto-allow when the body holds only plain text and simple $VAR references. The Bash permission check now prompts before a BASHPID assignment the shell would evaluate as arithmetic, instead of silently allowing it. Auto mode denial messages no longer point at a Bash permission rule when the blocked tool wasn't Bash.
Full release notes also cover MCP OAuth re-authentication prompts, screen reader announcements for permission mode changes, OpenTelemetry span fixes, and Cowork cloud session prompt suggestions.
\n\n📖 Read the full source: GitHub Claude-Code
👀 See Also

Yann LeCun's AMI raises $1B for AI world models, challenges LLM approach
Yann LeCun's startup AMI raised over $1 billion to develop AI world models that understand the physical world, arguing LLMs alone won't achieve human-level intelligence. The company will build systems with persistent memory, reasoning, and planning capabilities for manufacturing, biomedical, and robotics applications.

OpenClaw Codex OAuth returning billing errors despite valid account
OpenClaw Codex OAuth is returning a 429 error stating 'Your account is not active, please check your billing details' even though billing is confirmed valid and the exec command works. The issue persists across multiple OpenClaw versions.

Coding Agents Supersede Human Code Review: Paper Argues Traditional Review Is Dead
arXiv paper argues coding agents have crossed the threshold to replace human code review, offering lower cost and higher throughput.

NVIDIA Releases Nemotron-3-Ultra-550B: 55B Active Parameters, 1M Context, LatentMoE Hybrid
NVIDIA released Nemotron-3-Ultra-550B-A55B-BF16, a 550B parameter model with 55B active, 1M token context, hybrid LatentMoE architecture (Mamba-2 + MoE + Attention + MTP), and configurable reasoning.