Claude Code w/Sonnet 3.7 is the closes thing to AGI I've seen. It's magic. It built this Connect 4 app from scratch that 1. supports mobile 2. a minimax-tree AI 3. frontend, backend, DBs, git, and deploy! ~5000 lines of code, 30mins, $10. Why use Cursor anymore? FULL REVIEW:

Link: Git: Claude Code is an "agent" you access through your terminal in a repo and it just does things you ask it to. Writes code into files, runs them, fixes errors, and repeats. You just run ‘claude’. It’s in research preview but the experience is so good: 1. The UX is fabulous. It's a premium grade command-line tool. Very easy to follow, clean colors and the awesome "Vibing.." style animation 2. It will warn you when you're overspending. 3. It auto-fixes errors and tries sensible next steps, even tricky ones. In one situation, it found that Docker wasn't deploying because it had added Dockerfile to the gitignore! 4. If you're running up against the context window, it will ask you to press /compact and compress everything it needs to remember. 5. Fully capable of deploying to the cloud. In my case, I deployed Node with Google Cloud Run. 6. Will write your README for you. 7. Will keep track of every change in Git with detailed commit messages you'll never write under the Claude Code alias. 8. It's capable of finding non-trivial issues. 9. Doesn't get stuck running long-running commands. 10. Saves the context state so you can come back where you left off. One minor annoyance was that it would get tripped up about which folder it thought it was in. It really only tripped up 3 times during the entire session, and it was always able to fix it.

AppKit is the full-stack toolkit to build onchain app UX 🪄 ✅ Social, Email, and Wallet Login ✅ Embedded Wallets ✅ Crypto Swaps ✅ On-ramp Integrate with just 20 lines of code across 10+ languages for all EVM chains and Solana. Onboard millions of users for free today.

Sure, coding Connect 4 is definitely AGI

Cursor composer still struggles to autopilot, especially TDD, this is crazy. Write code, write test cases, iterate. But $10 for just 5k lines of code, as I dev I’d rather buy cursor + replit + coderabbit.

Imo the reverse might happen. cursor might merge with Anthropic

I had to downgrade from 3.7 to 3.5 because it’s incredibly chatty and committed to making mistakes, whereas 3.5 is more concerned about committing to anything. My existing codebase seems better off with 3.5 than 3.7. ~ 1200 engineers in my org, all downgraded to 3.5 by demand.

I really dislike the low rate limits tbh.

I built a connect 4 game -> this is artificial general intelligence is wild

Yeah, tried 3.7 Sonnet. Still sucks for anything serious. Useless but intricate (aka coding test apps) it’s amazing at. For real world engineering it’s still incredibly painful. “Ah, now I see the problem” Do you? Do you really?

Looks like an ad. Try Grok 3.

Does it allow you to manually edit the code?