Muhammad Ali Hassan

Ever since I started using agentic CLIs I've been staring at the bottom 20% of my screen all day. My posture has never been worse.

Claude Opus is the sibling that moves fast, breaks things, and needs constant supervision. Codex is the one that quietly gets it right the first time. You love both but you trust one with the important stuff.

Rumor is Opus 4.7 is dropping this week and it might include a design tool. Also, has anyone else noticed Opus 4.6 has been performing way better this past week? Feels like the first week after launch again. Something definitely changed.

Agentic coding is literally just reading and reviewing markdown files all day. Nobody told me this is what the future of software engineering looks like.

Took me a while to figure out why Claude Code would randomly start writing nonsense. Turns out once context hits 60-70%, it basically loses the plot. Now I clear context before every new task and never let it go above 50%. Night and day difference.

Writing tests isn’t difficult. Knowing what to test is.

Will LLMs keep getting better forever or just plateau at some point? 🤔

Have you built a side project for yourself that unexpectedly gained traction?

Do you use sub-agents in Claude or Cursor as part of your workflow? If so, how?

AI is already good enough to write clean, working features. If it keeps producing bad code in your project, the problem is often the existing architecture or context you give, not the model.

Is AI helping you get more done, or slowing you down in the long run?

What’s your go-to agentic tool for actually building software?

Is it just me, or does Opus 4.5 feel… worse lately? Feels like something changed.

When you’re stuck on a problem, what’s your go-to move to break out of it?

How many of you have broken a perfectly working app after a “quick refactor”? 😭🙋‍♂️

If you find yourself consistently sacrificing the quality of your code just to meet deadlines, it might be time to reassess whether that’s the right job for you.

Every time I use GPT, I have to remind it not to just agree with me. Feels like I’m arguing with a very polite friend.

What’s a software tool you’ve spent money on that was worth every cent?

Have any of you tried Augment Code in vscode? Fore it's the best agentic AI. Every request has 200k+ token context and almost never hallucinates. But it costs 50$/month 🥲

What’s the most expensive AI tool or subscription you’re currently paying for?