150 wires and counting

$ follow Anthropic

Keep up with Anthropic in about 3 minutes: what actually shipped — the commits, pull requests, releases, and security advisories that matter.

or

fair warning: these emails are deeply technical. diffs, version numbers, CVEs, benchmark deltas. if that's not your idea of a good read, this isn't your newsletter.

Folds into your digest — weekly by default, monthly if you prefer. Unsubscribe in one click.

$ status

wire 2026-10-17
stories 52

© 2026 RepoJournal Home Showcase How it works Privacy

$ the-wire · showcase

sandbox-runtime fixes a quadratic glob parse, and five projects enroll in OSS Scanner

By RepoJournal · Filed · About Anthropic · Composed from the cited sources · methodology

The day's real engineering is in sandbox-runtime, where a 300 ms permission getter is cut back to release-speed numbers; the rest of the wire is enrolment paperwork and workshop reshuffling.

Brackets are read in linear time, and a read rule is compiled once anthropics/sandbox-runtime

by ronleizrowice-ant

globToRegex() split its pattern with /(\[[^\]]*\])/, which rescans for a closing bracket from every opening one, so unclosed brackets made the parse quadratic. That path accounted for 77% of getFsWriteConfig()'s time, and read rules, compiled per answer, are now compiled once: up to 300 ms on main against 4 ms in 0.0.79. The regression arrived with #687 and shipped in no release.

Release v0.0.80 (#664) anthropics/sandbox-runtime

by ronleizrowice-ant

v0.0.80 is cut, so if you pin sandbox-runtime you can stay on 0.0.79 or move to 0.0.80 without ever touching the slow bracket path.

feat(managed_agents): add agent security review cookbook anthropics/claude-cookbooks

by cj-ant

A new cookbook, managed_agents/CMA_agent_security_review.ipynb, reviews finished agent sessions after the fact: one AGENTS entry per agent declaring what it may do, transcript normalization from Managed Agents sessions plus Messages API and Claude Agent SDK logs, then 60 pattern rules with bounded regex search and seven structural rules to flag what looks off.

ship-your-first: three pages, multiagent, memory stores, file output (#28) anthropics/cwc-workshops

by Gagan Bhat

The SRE agent workshop is rebuilt in four parts instead of one, opening with three pages firing at once from two bad deploys and climbing to an incident commander with one investigator thread per page and a read-only memory store seeded from the repo. agent.py ships as stubs for a coding agent to fill while the participant makes the judgment calls.

Enrole diesel-rs/diesel anthropics/oss-scanner

by weiznich

Enrolments landed for diesel-rs (weiznich), along with DPDK, simplexmq, yopass and shopware, each from a core maintainer with tools/validate.py and tools/check run locally; also on the wire, cargo-nix-plugin reverted thinlto stable names.

Quick answers

What shipped in Anthropic on October 17, 2026?
The day's real engineering is in sandbox-runtime, where a 300 ms permission getter is cut back to release-speed numbers; the rest of the wire is enrolment paperwork and workshop reshuffling. In total, 27 commits, 24 pull requests, and 1 releases landed.
Who contributed to Anthropic on October 17, 2026?
9 developers shipped this update, including shemminger, shumvgolove, jhaals, shyim, weiznich, ronleizrowice-ant, CJ Avilla, and Jörg Thalheim, and 1 more.
What were the notable Anthropic updates?
Brackets are read in linear time, and a read rule is compiled once, Release v0.0.80 (#664), and feat(managed_agents): add agent security review cookbook.