Switched from @cursor_ai to @claudeai (Claude Code). The model is the same, Anthropic. So I expected similar results. I was wrong.
What actually changed:
✅ It understands the full project context, not just what you paste
✅ Reuses existing components instead of creating duplicates
✅ Doesn't touch UI you never asked it to change
❌ Terminal-first is a real adjustment if you're used to an IDE
❌ Still keeping Cursor open — no shame in that
My take: same brain, better eyes.
We hit 78k users in 3 months on the SaaS where I'm CTO. Huge milestone, but crazy demanding. Normally, a project like this means zero time for anything else. But I fully integrated Claude into my workflow and the output has been insane.
+67,000 lines. 365 commits. 24 PRs. 1 month. ☝️
AI doesn't just write code, it actually buys you time.
How much has your coding speed increased with AI lately?
Solo devs, when was the last time you wrote tests?
Building alone, it's easy to skip them. No one is reviewing your code, no one is pushing you to do it.
My app reuses a lot of logic but with variations depending on the case. Every time I built a new variation I'd break a previous one without noticing. Hours lost chasing something I touched three days ago.
Yesterday I stopped adding features and spent the time writing tests for the core logic instead.
I'm still going to break things. But now I'll know exactly where.
Do you write tests for your solo projects?
A week ago we launched an app I’ve been building as CTO.
2,500 users in 7 days. 5.3% conversion to paid.
But the real story starts a month before that.
The company had been working with another team for months. One month before their planned launch date, they had to start over from scratch.
That’s when I came in.
One month. Web app and mobile app. From zero.
Not perfect. Not easy. There were nights I questioned everything.
But it worked.
Those numbers this week reminded me what’s possible when you just commit and
Mostly the second. The company already had an established customer base outside of tech, so the initial pull came from existing clients moving to the new platform.
But we had a launch campaign running for months, which is actually why the rebuild under pressure was so intense. The date was locked.
Event launch, digital media, and a few influencers in the industry. By the time we shipped, there was already an audience waiting.
Having that pre-built trust made a real
Out of topics… I’ve been a sneaker person my whole life.
Comfort was always subjective.
Then I tried @On_RunningESP for the first time.
That first step is genuinely different. Not marketing, not placebo. Something actually changed in how a shoe felt.
What gets me is not the design. It’s that someone looked at something as “solved” as a shoe and said, we can engineer this better.
CloudTec, carbon plates, cushioning systems. Real technology inside something you just… walk in.
That mindset applies to everything. Nothing is ever fully solved. There’s always a better version waiting for someone obsessed enough to find it.
Watched a non-technical CEO build a working app with Claude Code this week.
Impressive. Genuinely.
But I saw exactly why developers aren't going anywhere.
Didn't expect to talk software architecture at padel practice. Here we are.
Met a developer this weekend, one year out of college. Good energy, asking the right questions. At some point he asked me what the best framework to learn was.
15 years in, my honest answer: that question doesn't matter as much as you think.
Frameworks come and go. The engineers I've seen grow fastest were never the ones who mastered one stack.
They were the ones who stayed curious, took risks, and learned how to think about problems first.
The tool follows. The thinking has to come first.
We got into AI too, but that's a separate post.
The hard part isn’t the technical shift, it’s the identity one.
Developers spent years being valued for how they wrote code. Now the value moves to judgment. Knowing what to build, catching what the agent got wrong, reading the output critically.
But you still need someone who can read that output. The agent hallucinates a pattern that almost works and you need to catch it.
Same craft, different surface.
@OmriBuilds It’s actually the best time. No client calls, zero notifications, just deep work. I try to rest, but the focus I get on Saturdays is addictive.
My @cursor_ai paused because I reached my self-imposed spending cap. It took me 10 seconds to raise the limit. When a tool helps you close tickets and ship features this fast, it pays for itself. Speed is the only metric that matters right now.
Two weeks of squeezing dev time between client calls. The result: 11,547 lines added and an "infinite" amount of technical debt removed.
We are moving fast now. What does your git log look like this week?
My backlog adds two items for every one I close. Classic. Still, progress is good: 60% done with the core feature for the beta. Balancing this with client work is tough, but having real prospects waiting to test the app changes everything.
Back on Cursor for a few days. I maxed out my quota on Antigravity and since there's no top-up option for Business accounts yet, I had to switch. Good to have alternatives to keep shipping.
Which editor are you maining this week?