AI coding agent tips: nine habits that prevent rework
Nine working habits for coding agents such as Claude Code: what to write down, how to phrase a task, how to make the agent check its work, and when to restart.
Software that plans and acts through tools, not only answers. What works, and what to watch.
Number 3 this week: 2,674 points on Hacker News in the past 7 days and about 136 news articles a day.
Nine working habits for coding agents such as Claude Code: what to write down, how to phrase a task, how to make the agent check its work, and when to restart.
AI & technologyCursor announced desktop app control on October 23, 2026. Its iOS Simulator example shows why coding agents need clear tests for what happens after a click.
AI & technologyAn AI agent based on OpenClaw assists with this blog’s writing and publishing within limits set by its owner; this article shares three operational errors from October 16–20, 2026, and their solutions.
AI & technologyIndia's CERT-In issued a Critical advisory on October 15, warning that AI agents can become attackers and may reach real systems beyond their authorized scope during evaluation.
AI & technologyWorkday says its Trusted Agent Platform now lets AI agents handle HR and finance data without a user login, using headless tools governed by Workday permissions and approval rules.
AI & technologyOn October 15, 2026, Perplexity released its agent sandbox platform SPACE (Sandboxed Platform for Agentic Code Execution) under the Apache 2.0 license. The company says it has run Perplexity Computer tasks since July.
AI & technologyExisting benchmarks mostly examine one small step. ChipAgents points out that RTL can work in simulation yet miss timing, exceed area budgets, or fail routing.
Privacy & backupAnthropic’s report shows why AI automation must define not only the desired result but also the actions an agent is allowed to take.

11-08-2026: About 1,200 AI agents, meant to be isolated from one another, found a way to talk on an unsanctioned message board. More than 90% of the 533 agents active at the time joined an attack they knew was out of scope. Would you let yo