Signing you in...

Please wait while we verify your authentication

Article · Thursday, October 1, 2026

Agentic Coding

Top tech stories today across software, hardware, AI, and product launches. Senior engineer audience — skip rumour churn and pre-announcement leaks. Lead with shipping-now stories and what changed for builders.

By Marius BongartsTech120 editions
← See today's latest
Editions
7 / 120
Generated by AI overnight from public sources, refreshed daily.
Agentic Coding
Thursday, October 1, 2026
AI Agents - Agentic Coding

Less is more for Claude Code, Opus gets a million-token window, Codex brings AGENTS.md to GitHub reviews

1 min read

Claude Code Setup Audit

Your advanced Claude Code habits might be slowing you down.

A workflow comparison between an experienced user and a beginner revealed that many power-user tweaks backfire. Keeping MCP servers minimal prevents tool-selection confusion and context bloat. Limiting memory layers avoids stale context creeping in. Anthropic actually recommends keeping your CLAUDE.md short rather than stuffing it with every preference you've ever had [Source: daily.dev]. Plan mode? Reserve it for genuinely complex tasks—two-minute edits don't need upfront planning.

Strip the clutter periodically; your setup accumulates cruft faster than you notice.

Opus 5.5 Context Window

You can now load your entire codebase into one session.

Claude Opus 5.5 ships with a 1M token context window, and it's now the default model in Claude Code v2.1.280 for Pro, Max, and higher tiers [Source: ClaudeLog]. The practical guidance: use Sonnet for 80% of your work, reserve Opus for complex architectural planning, and drop to Haiku for simple operations. Watch your context usage—performance drops noticeably in the final 20% of your window. Use /compact strategically and batch related changes into focused sessions.

Max users can now run Opus all day without limits—hybrid workflows finally make sense.

Codex GitHub Reviews

AGENTS.md just landed in GitHub pull request reviews.

OpenAI's Codex now integrates directly with GitHub for automated code review, and you can configure repository-specific guidance using AGENTS.md files [Source: ChatGPT Learn]. Trigger reviews with @codex review or @codex security review, or enable automatic reviews on every PR. Custom review rules live in your AGENTS.md, so the same patterns you're building for Claude Code transfer to your review workflow.

Your context files now work both sides of the development loop.

Sources
I compared my Claude Code workflow to a beginner's, and...
I compared my Claude Code workflow to a beginner's, and...
9 hours ago ... A regular Claude Code user audits their own workflow against a beginner's leaner setup, discovering that many 'advanced' habits were counterproductive....
daily.dev
AI Summary

The user maintains an efficient Claude Code workflow by regularly auditing their setup to remove counterproductive habits. Key findings include keeping MCP servers minimal to avoid tool-selection confusion and context bloat, limiting memory layers to prevent stale context issues, keeping CLAUDE.md files short per Anthropic's guidance rather than stuffing them with every preference, and reserving Plan mode for complex tasks where upfront planning adds value rather than using it for simple two-minute edits. The core lesson is that advanced tweaks don't automatically improve results, and periodically stripping accumulated clutter from a Claude Code configuration helps maintain productivity.

Visit source
Claude Code Limits | ClaudeLog
Claude Code Limits | ClaudeLog
6 hours ago ... Workflow Optimization​. Task Segmentation: Breaking complex features into manageable phases prevents overwhelming context windows and enables better progress ...
claudelog.com
AI Summary

Claude Opus 5.5 now features a 1M token context window, enabling developers to load entire codebases for extended development sessions with minimal optimization overhead. The latest Claude Code v2.1.280 release made Opus the default model across Pro, Max, and higher tiers. For workflow optimization, strategic model selection follows a hierarchy—use Sonnet for 80% of development work, reserve Opus for complex architectural planning and analysis, and use Haiku for simple operations. Context management is critical: avoid using the final 20% of your context window for complex tasks as performance degrades significantly, use the /compact command to strategically reduce context size, and batch related changes into focused, single-purpose sessions to prevent wasteful context consumption. Max users can now use Opus freely throughout all-day coding sessions without hitting limits, enabling hybrid workflows where Opus handles initial architecture and planning while Sonnet executes implementation work.

Visit source
Review GitHub pull requests with Codex - ChatGPT Learn
Review GitHub pull requests with Codex - ChatGPT Learn
10 hours ago ... A GitHub repository connected to Codex. For legacy integration setup, see Codex Cloud (Legacy). Access to Codex code review settings. An AGENTS.md file if you ...
learn.chatgpt.com
AI Summary

This content is about GitHub integration with Codex for automated code review workflows, including AGENTS.md file configuration for repository-specific guidance. While it mentions AGENTS.md files which relates to the user's interest in efficient agents and Skills, the content does not address Claude Code workflow optimization, Cursor integration, agentic coding memory management, or AI pair programming IDE integrations that the user is specifically seeking.

Visit source
Compiled overnight by MorningMail.aiDelivered at 04:55 AM