Assistants on Tokenise

A PreToolUse hook is the guardrail that survives bypass mode

Mon, 05 Oct 2026 22:30:00 +0000

Permission prompts are only as strong as the mode you’re running in. Switch to bypass mode, or launch with --dangerously-skip-permissions, and the prompts are gone. A PreToolUse hook is different: according to the Claude Code hooks guide , it fires before any permission-mode check, and a deny from it blocks the tool even in bypassPermissions mode.

A separate evaluator agent is what makes long agent runs worth the money

Mon, 05 Oct 2026 17:00:00 +0000

Ask one agent to build a whole application and two things tend to go wrong. It loses the thread as its context fills, and it grades its own work generously. Anthropic’s engineering team wrote up a harness design for long-running application development that attacks both, and the numbers are honest about what it costs.

Cursor Projects puts a coordinator agent in charge of the work

Mon, 05 Oct 2026 07:00:00 +0000

Cursor’s September 10 release of Projects, in beta and rolling out to all users per its changelog , is built on a specific idea: the agent in charge shouldn’t be the one writing the code.

What 16 parallel agents building a C compiler teach about running agents at scale

Sun, 04 Oct 2026 12:00:00 +0000

Anthropic’s engineering team handed 16 Claude instances a goal, a shared repo and a test suite, and let them write a C compiler in Rust. The result, described in an engineering post , is a roughly 100,000-line compiler that can build Linux 6.9 on x86, ARM and RISC-V, plus projects like SQLite, postgres and redis. It took nearly two weeks, close to 2,000 Claude Code sessions, 2 billion input tokens and 140 million output tokens, for just under $20,000.

Let the agent interview you, then write the spec, then start fresh

Sat, 03 Oct 2026 17:00:00 +0000

Big features fail in a predictable way when you hand them to an agent in one prompt. The agent fills every gap in your description with a guess, builds confidently on the guesses, and you find out about the wrong ones at review. A better move is to make it ask first.

Claude Code's auto mode catches most risky actions. Anthropic's numbers show what it misses

Sat, 03 Oct 2026 07:00:00 +0000

Anthropic published how Claude Code’s auto mode works, with measured error rates. That’s unusual for a security feature, and it makes the engineering post worth reading before you let an agent run unattended. The full write-up is on Anthropic’s engineering blog.

Run parallel Claude Code sessions without them stepping on each other

Fri, 02 Oct 2026 07:00:00 +0000

Two Claude Code sessions in the same checkout will eventually edit the same file. Git worktrees are the standard answer: each session gets its own working directory and branch, backed by the same repository history. Claude Code has a flag for it, and the worktree docs cover more edge cases than the one-liner suggests.

Your CLAUDE.md is probably too long. Here's how to cut it

Thu, 01 Oct 2026 07:00:00 +0000

A CLAUDE.md file starts life as five useful lines. Six months later it’s two hundred, half of which describe things Claude could work out by reading your code, and the instruction you actually care about is somewhere around line 140.

How Codex reads AGENTS.md, and how to layer it so the right rules win

Wed, 30 Sep 2026 17:00:00 +0000

Codex reads instructions from AGENTS.md files before it starts work, and it doesn’t read just one. It builds a chain of them, in a documented order, and the order decides which rule wins when two disagree. If you’ve ever had Codex ignore a rule you were sure you’d written down, the layering is the first place to look.

Stop telling your AI agent to 'fix the bug'. Make it reproduce it first

Wed, 30 Sep 2026 12:00:00 +0000

“Fix the login bug” is a terrible prompt for a coding agent, and a frustrating one to debug when it goes wrong. The agent will read some code, find something plausible, change it, and tell you it’s fixed. Sometimes it is. Sometimes it has patched a bug you didn’t have.

Make Claude Code run your tests before it's allowed to say it's done

Wed, 30 Sep 2026 07:00:00 +0000

Telling Claude Code to run the tests in CLAUDE.md is a request. It usually works, and sometimes it doesn’t. If you want it to work every time, Claude Code has a mechanism for that: a Stop hook, a script that runs when Claude thinks it’s finished and can refuse to let it stop.

Cursor's Rollouts bot watches your deploys after the merge

Tue, 29 Sep 2026 12:00:00 +0000

Cursor has added a bot called Rollouts that watches a change after it merges and reports whether it’s healthy in each environment. It was announced in the Cursor changelog on September 23 and is available on Teams and Enterprise plans.