Atuin 18.24 adds Threads to search, sync and resume agent sessions across machines
Top stories
Atuin 18.24 adds Threads to search, sync and resume agent sessions across machines
- Atuin announces Threads in version 18.24, letting developers search agent sessions across machines, resume them anywhere, or fork them into a different agent harness.
- It supports Claude Code, Codex, Pi, Omp and OpenCode; sync is end-to-end encrypted, capture is opt-in, and a server can be hosted, self-run or skipped.
- Agents can search past sessions through an MCP server installed with atuin mcp install, and the team is building opt-in session sharing.
AWS releases AWS Emulator, a free offline copy of 200+ services built by agents
- AWS launched AWS Emulator, a free, offline local copy of AWS for coding agents, released under CC0, with code written and tested entirely by agents.
- It is built from more than 400 API models covering over 200 services, checked by fidelity tests against real AWS and a merge gate with security agents.
- Swami Sivasubramanian says engineers' work shifts to specs, tests and merge gates; with a spec and test environment, a rebuild of Alf's replication engine took about two hours.
Read more at allthingsdistributed.com
In brief
- Devon Estes describes team practices once seen as bad ideas now working with Claude. Devon Estes says his team stopped peer PR review, with Claude opening PRs and the ticket owner reviewing after CI and Greptile pass, and finds other former bad ideas working too. (devonestes.com)
- Thomas Neumann says GenDB-V's 22× to 61× speedups reflect benchmark overfitting. Thomas Neumann argues GenDB-V's claimed 22× to 61× speedups over Umbra and DuckDB are infeasible and stem from per-query precomputation and hard-coded TPC-H knowledge. (databasearchitects.blogspot.com)
- Christian Bird talk asks why AI has not made software teams ship faster. Christian Bird of Dryden Technology Group gives a UC Irvine seminar talk on October 30, 2026, asking why AI has not sped up shipping despite higher activity; its references include an NBER working paper and an in-submission paper based on 30 interviews. (cabird.com)
- Chrome releases Modern Web Guidance 1.0 for coding agents, with Firefox and Edge joining. Chrome says Modern Web Guidance 1.0 adds a modernization workflow, a larger guide database, expert review and public evals, with Firefox and Edge joining the effort. (developer.chrome.com)
- Google DeepMind formally verifies parts of Signal protocol with AlphaProof Code. Ars Technica reports DeepMind used AlphaProof Code to verify several Signal protocol guarantees in four weeks, finding no new vulnerabilities but flagging known fixed flaws in older versions. (arstechnica.com)
- Darling adopts AI contribution policy and adds initial ARM64 support. Darling's Q4 2026 update covers rules for AI-assisted contributions, initial ARM64 support driven by the VibeDarling fork, and plans to merge forks and stabilize AppKit. (blog.darlinghq.org)
- Anthropic publishes prompting guide for Claude Fable 5.2 and Mythos 5.2. Anthropic's guide recommends starting at high effort and covers finishing whole tasks, code ownership, subagent use and refusals, noting merged repeated code in about half of tries with the ownership sentence. (platform.claude.com)
Chosen by software from the day's news, written by AI from the pages it links to, and checked against them by a second model. Published by Rapid Review, Inc.