AI Coding: My PLANNING System
AI Coding news: June 3, 2026.
Hey hey,
Big update on my AICodingDaily.com website: I’m introducing COURSES for Premium members! For now, only the first mini-course on Planning system that works for me, but already a few more coming.
With that in mind, I’m moving membership subscriptions from here to the website, as on Substack I have very limited options for publishing something like courses. Here’s an article with details: Moving Paid Substack Membership to AICodingDaily.com Website
And now, as usual, a lot of useful links to my videos and community tools/tips/news. Enjoy!
Hey hey,
Big update on my AICodingDaily.com website: I’m introducing COURSES for Premium members! For now, only the first mini-course on Planning system that works for me, but already a few more coming.
And then, as usual, a lot of useful links to my videos and community tools/tips/news. Enjoy!
From My Website
[Premium Mini-Course] AI Coding: My PLANNING System
Across 5 short videos in 37 minutes, I’ll show how I think through planning, use plan mode in Claude Code and Codex, and pressure-test ideas with the grill-me skill. Source code and prompts included.
Skill /handoff by Matt Pocock: Summarize Current Session for Another Agent
A very simple skill that you may also use as a prompt, when you’re running out of context/time in the current coding session, and you want to continue later, potentially with another agent.
My YouTube videos
I Tested NEW Minimax M3 on Four Projects
A new Minimax M3 model was released, and it’s pretty different quality from previous M2.7, let me show you.
I Tested NEW Opus 4.8 on Four Projects
The new Claude Opus 4.8 is out, and I ran my LLM benchmark to compare how better/different it is from 4.7 and other models.
I Tried NEW Gemini 3.5 Flash. It Has One Big Problem.
After a week of hesitating, I decided to test out the Gemini 3.5 Flash, after all, and here’s what I saw.
I Tested GPT low/mini/older Models: Price/Quality Difference
In the comments to other YouTube videos, you asked to test GPT-5.5 Low effort, GPT-5.4-mini, and GPT-5.3-Codex.
Coding LLM Prices Comparison: My 5 Takeaways
I have filled in the prompt cost for all 13 LLMs into my leaderboard, and let’s discuss those numbers.
AI Coding Community News
Shpigford/chops: Your AI agent skills, finally organized — github.com
A macOS app to browse, edit, and manage skills across Claude Code, Cursor, Codex, Windsurf, and Amp.
Agent Skills Cheat Sheet: Complete Reference — webfuse.com
How to create Skills, and how to work with them. SKILL.md spec, authoring patterns, scripts, evals, and client integration in one place.
A harness for every task: dynamic workflows in Claude Code — x.com
A very detailed article by Thariq from Claude Code team.
The Horror of First Day of GitHub Copilot new pricing — x.com
Sarcasm: “Day one of GitHub Copilot token-based billing and the customers LOVE IT! They’re all celebrating the power and value of generative AI!” (and this is just one of multiple tweets on the same topic)
Tibo on X: “We will be sunsetting GPT-5.2 and GPT-5.3-Codex in Codex on June 2nd” — x.com
To simplify our Codex compute fleet management, we will be sunsetting GPT-5.2 and GPT-5.3-Codex in Codex on June 2nd when logged in with your ChatGPT account. For free plans, GPT-5.5 will be the default frontier model to build and work with going forward. These models will remain available on our API.
More cache hits means lower cost and latency for your API requests.
How to Stop Claude Code From Reading Your API Keys, SSH Keys, and Passwords (Exact Deny Rules) — x.com
Big article on X by Rody.
Cat on X: “New Claude Code feature: dynamic workflows!” — x.com
Mention “workflow” in a prompt and Claude will dynamically create an orchestration plan that it strictly follows, allowing you to confidently trust that every stage happens in the right order even across 100s of agents.
OpenCode on X: “OpenCode x MiMo V2.5 - Free for a limited time” — x.com
1M context • reasoning • text • image
Reddit post with surprisingly sane and solid workflow for AI-assisted development.
That’s a wrap for this week. Keep building with AI!
Povilas Korop
AICodingDaily.com
Packing the grill-me skill and planning mode into a 37-minute mini-course saves builders hours of experimentation.
We analyze agent handoffs at theaifounder.substack.com, finding that jumping between Claude Code and Codex shatters the initial project context.
What specific guardrails force the secondary model to obey your original architecture spec?