Skip to content
Tech readingTech reading
Menu
LibraryTrends
Daily issues
Sunday 8 November 2026Saturday 7 November 2026Friday 6 November 2026Wednesday 4 November 2026Tuesday 3 November 2026Monday 2 November 2026Sunday 1 November 2026
Weekly voices
Library
Library

2026-11-08

Daily reading: Sunday 8 November 2026

Today’s theme: trust in the tools you run agents with. It’s a light digest today, and several items are repeats. The one that stands out is a security report that touches the Claude tools on your Mac. The others are about choosing between those tools and the organisation around them.

Read first (about 25 min)

Section titled “Read first (about 25 min)”
  1. “Anthropic hacked to deliver remote access trojan to Claude developer laptops” (Kevin Beaumont, DoublePulsar, 4 min) This is a claim in the article’s headline. I haven’t verified it, and I only have the blurb. Beaumont is a well-known security researcher, so it’s worth reading first, since you run Claude desktop and Claude Code. Check Anthropic’s own status or security page before acting on it. Relates to: Monday’s “Building the harness around AI coding tools” (permissions and sandboxing) and Wednesday’s “Claude Bug Hunter”. This is why the sandbox layer matters.
  2. Claude Code Desktop vs CLI vs VS Code: 8 features only one has (Automation Labs, 9 min) It compares features after the October 15–16 releases. Relates to: your Electron vs native thread (all three surfaces run inside Electron apps or a terminal) and the Claude desktop / Agent View entry in your Coding agents landscape.
  3. Your developers aren’t slow. Your organization is. (Jekaterina Jegoscenko, 12 min) Relates to: yesterday’s “AI reviewing Linux” (the kernel got better but the release process got harder) and Sunday’s “Anthropic ditched Git”. Once agents speed up the coding, the bottleneck moves to the process around it.

Coding agents

Section titled “Coding agents”
  • Rex: Ghostty’s creator welds tmux and AI into the Mac terminal (Chimin, 5 min). It came up again today and is still in your to-read list from Tuesday. It’s a terminal-first take on running several agents side by side, alongside Orca.

If you have time

Section titled “If you have time”
  1. Deep research agent that remembers (Agent Native, 11 min, from a writer you follow) This is its third time in your digest. Relates to: Friday’s “Cache engineering” and yesterday’s “Understand any codebase with 99.7% fewer tokens”. It’s about memory as another way to stop agents redoing work.

Skipped

Section titled “Skipped”
  • “Did Stanford just kill ChatGPT?” (hype), “8 ‘Claude 2.0’ features” (the piece itself says there’s no Claude 2.0), “5 MCP servers for DevOps” (a listicle), Anatomy again, the AI history series episode, the Anthropic IPO breakdown and the Rubik’s cube guide. The knowledge-graph, harness and codebase pieces were repeats, already on your list.

Note: links between pieces are based on digest blurbs only.