Welcome back to Agentic Coding Weekly. Here are the updates on agentic coding tools, models, and workflows for the week of May 3-9, 2026.
Executive Summary:
Anthropic increased Claude Code five-hour limits for Pro, Max, Team, and seat-based Enterprise plans. Weekly limits are unchanged.
Codex CLI now has a built-in ralph loop. Set an objective with /goal and let it run autonomously until completion.
DeepSeek V4 Pro available at 75% off through API until end of May.
Worth reading: Simon Willison on the blurring line between vibe & agentic coding, and a good argument for HTML over Markdown as LLM output format.
Meanwhile, if you're looking for a new hobby, try staring at walls.
1. Tooling and Model Updates
Higher Claude Code Limits
Anthropic partnered with SpaceX to increase compute capacity and announced higher Claude usage limits. The five-hour rate limits for Claude Code on Pro, Max, Team, and seat-based Enterprise plans are now doubled. Weekly limits are unchanged, so this mostly helps with bursty sessions rather than total weekly usage.
API rate limits for Claude Opus models also increased quite a bit. Tier 2 accounts, for example, went from 450k/90k to 2M/200k max input/output tokens per minute.
Codex CLI Ralph Loop
Codex CLI now has a built-in goal loop, same principle as the ralph loop pattern. Instead of stopping after each turn, Codex keeps working toward a single objective until it hits a stopping condition. Can run for hours independently.
Enable it by running /experimental, then set your objective with /goal <objective>. Check current goal with /goal, control execution with /goal pause, /goal resume, or /goal clear.
Example goal loop prompt for building a prototype:
/goal Implement PLAN.md, creating tests for each milestone and verifying the output with playwright interactive.
Quick Updates
DeepSeek V4 Pro is available at 75% off through API until May 31st. After discount: $0.435/$0.87 per million input/output tokens.
antirez built a specialized native inference engine for DeepSeek V4 Flash optimized for Apple Silicon Metal. Clean, minimal codebase worth reading. Runs at 2-bit quants on 128GB Macs, 4-bit on 256GB.
2. Community Picks
Vibe Coding and Agentic Engineering Are Getting Closer Than I’d Like
Simon Willison on the uncomfortable reality that as coding agents get more reliable, he hass stopped reviewing every line they write, even for production code. Have found myself doing this for simpler stuff as well.
The problem is that as the coding agents get more reliable, I’m not reviewing every line of code that they write anymore, even for my production level stuff.
I know full well that if you ask Claude Code to build a JSON API endpoint that runs a SQL query and outputs the results as JSON, it’s just going to do it right. It’s not going to mess that up. You have it add automated tests, you have it add documentation, you know it’s going to be good.
But I’m not reviewing that code. And now I’ve got that feeling of guilt: if I haven’t reviewed the code, is it really responsible for me to use this in production?
The Unreasonable Effectiveness of HTML
An argument for using HTML as the output format with LLMs and AI agents over Markdown. Read the post.
Programming Still Sucks
Very well written. Just go ahead and read it.
That’s it for this week. I write this weekly on Mondays. If this was useful, subscribe below:
