Tags
Ten people ran Claude Code and Cowork for 30 days, and two sessions they reopened every day for over a week ran up 32% of the whole bill. Every call in those two sessions paid for ~500k tokens of cached context.
Close your laptop and a Claude Code session started with claude --cloud keeps running on an Anthropic-managed VM. There's no separate compute charge for it, but it draws on the same rate limits as all your other Claude usage.
Switching off telemetry in Claude Code also quietly disabled a feature that only reads a local file. AGENTS.md support waits on a remote flag that falls back to false, and nothing tells you the file was skipped.
Anthropic's own advice for Opus 5.5 is to delete "think carefully" from your prompts. When Anthropic cut that line in its chat testing, replies started sooner, with no clear drop in quality.
Claude Code users on Pro, Max and Team just got a spare limit reset they can spend on any bad day before Oct 22, and it comes on top of 5-hour session limits that went up 20% today.
Anyone budgeting next week's sprint around Claude Code's current weekly cap loses 17% of it on September 14, when the temporary 50% boost expires and a permanent 25% raise over the old baseline takes its place.
No CLAUDE.md in the repo? Claude Code now falls back to AGENTS.md, the instruction file rival coding agents already read, so one set of project rules covers all of them. CLAUDE.md still takes priority when both files sit in the tree.
Nearly two thousand files of Claude Code's original TypeScript shipped inside the npm source maps, and someone read all of them. The teardown runs 18 chapters, written by 36 AI agents in about 6 hours.
One goal, a thread per repo Set a goal to retire a deprecated v1 endpoint and Claude opens a thread per repo, runs tests, and says which PRs to merge first. Each thread is a full cloud session, so projects hit limits faster.
Researchers escaped Claude Code's sandbox with one message, and the sandbox itself is why no permission prompt appeared. Anthropic hardened every git call to fix it in 2.1.247.
Anthropic put a timeline on function hooks in Claude Code, "on the scale of weeks in lieu of days or months", and will ship the feature under the name Claude Mods. Anyone can test it today with CLAUDE_CODE_ENABLE_FUNCTION_HOOKS=1.
A high eval score tells you nothing on its own, so claude plugin eval runs every case a second time with the plugin switched off. The gap between the two scores, Δ, is what your plugin contributed.
Leave "verify twice" in a prompt after migrating to Opus 5 and the model takes it literally, duplicating lookups on every refund. Anthropic's /claude-api prompt-audit strips them: cost fell 14.6% and accuracy rose 5.3% on average.
"It increased from 45% to 55% while I performed no work," a U.K. consultant told TechCrunch about his Claude Max account. Anthropic traced it to a stolen session key minting unauthorized Claude Code OAuth tokens.
"Talking more doesn't make you smarter, Claude." In imec's aistack harness benchmark, Claude Code pushed 480M input tokens to Codex's 192M on the same 64 tasks, and its sweep cost $45 against $22.8.
Claude Code's MCP docs promise that tokens are "stored securely". On Linux that means a plaintext JSON file at mode 0600, and only Anthropic can change that, since the public repo leaves out the core CLI.
Security researchers found Git configuration ran commands before approval prompts in Claude Code and four other agents. A shared .zip with .git can deliver the configuration.
Context7's own benchmark puts fresh input tokens roughly 99% below Claude Code's built-in web search on documentation queries. Cost fell just 34.56% on average, softened by web search's caching and Haiku handling auxiliary calls.
Start a Claude session in the terminal, then pick it up in the Claude Code desktop app. Testing Catalog says a new /resume command carries the session over, and reads it as groundwork for Hub.
A break in Claude Code's Auto Mode, reported on August 24, is still open, according to a write-up on the security blog Embrace The Red. Anthropic made Auto Mode the default for new sessions on August 14.
