This is a fantastic article that breaks down Loop Engineering in a way that’s accessible even for non-technical folks. It perfectly connects to how I already use features in Codex and Claude every day—automations, skills, worktrees, sub-agents, and persistent memory—and shows how Loop Engineering ties it all together into a smarter, more autonomous system.
Instead of constantly prompting the agents yourself, you design the ‘loop’ (a recursive goal + verification cycle) that handles discovery, execution, checking, and iteration on its own.
Addy Osmani explains the five core building blocks beautifully and why this shift—from manual prompting to system design—could define the next era of coding with AI.
Some key takeaways from Ryo’s sharing:
The shift: AI is collapsing the distance between ideas and reality, shortening the iteration loop from days to seconds.
The risk: When execution costs nothing, "slop becomes free." We risk moving from an era of slow craftsmanship to a world of infinite, soul-less software—interfaces that are correct but feel completely dead.
The human element: "Taste is not a prompt. Caring is not a parameter." The real danger isn't that machines become too creative, but that humans become too passive around them.
Just spent the entire night immersed in the greatest movie posters ever created, all to craft one event poster with genuine taste and intention.
Even built a custom AI skill on the spot so I could experiment with every compelling style I discovered.
This is AI’s true gift: it delivers mass quantity at scale, yet frees non-designers like me to deeply study the masters, sharpen my own taste, and pour in real human craft and care.
“AI raised the floor. Time to raise the ceiling.”
Give the model your complete personal context — goals, workflows, SOPs, past portfolio, preferences, and operating style — and AI evolves from advisor to amplifier.
It goes from “here’s a good strategy” to “I’ll help you ship it.”
The execution layer is where the real leverage compounds.
I think one underestimated thing when we look back on it was how useful it is to have your own personal brain and company brain in 2026 at the dawn of usable AGI
AGI gives you the intelligence
You still have to collect your personal context to get the real unlock
@petergyang@liu8in Thanks for sharing. My friend who has video a podcast recently used HyperFrame to do all the cutting in Codex, and it works excellently.
AI can literally do all the syncing, switching, and timeline under the video, as well as pick up the highlights and etc.
Agree.
Tools like Typeless (or any solid voice input) are genuinely game-changing here. I’ll regularly speak for 5+ minutes straight, dumping all the nuance, constraints, examples, background, and what I actually care about into the AI. It feels exactly like briefing someone in a real meeting.
A good rule of thumb for preventing AI slop (in writing, design, etc):
Is your input (the context) longer than the output?
I've found that for the AI to produce quality results, my input is often 3-5 times the length of the output.
If your input is much shorter than the output, it's almost certainly going to produce slop
There are two small Codex features I cannot live without:
1. The ability to add more information and send it to Codex while it is still processing my last input. It doesn't stop the current task but continues to integrate the new information I just sent, which makes it feel much more like a real, live conversation in a group chat.
2. Record and Replay. It can literally mimic every step I take for a complicated task and redo it for me. It really feels like having a human intern who has been taught by me step-by-step and is doing exactly what I want using the same SOP.
I've been using Artifacts in Claude Code for everything: visual explanations of tricky code, system diagrams, quick previews of a few animation options, data analyses and dashboards I share with the team. They are a game changer for how I work with Claude. Can't wait to hear what you think!
Same here. Codex crushes it with my work tree, plugins, and automations + much better cost efficiency. Claude still better for planning and design for me. Competition is winning for us.
I used to be a die-hard Claude Code user.
Codex has won me over because:
→ GPT-5.5 is excellent
→ Fast mode + generous limits = more reps
→ Little touches like steering, auto remote control on phone, etc
But most of all Codex's browser and computer use capabilities are simply goated. I built so many workflows relying on those two things alone instead of hunting for APIs.
I still use Claude Code too. The app seems to be getting better and the design and frontend capability of Opus is still much better than GPT. Whenever Fable comes back that's another reason to go back.
Honestly, I hope these two compete forever and other players (Cursor/Grok, Gemini, etc) all stay competitive.
This way the builder keeps winning 🙂
People should just use Codex as ChatGPT
Since I started using Codex/Claude Code daily, I rarely open ChatGPT these days (unless it’s a quick online search type of task, in which case I use Gemini)
The coding agent’s result is usually strictly better than a chatbot’s