Dare Obasanjo

I realize one of the reasons I’ve been historically skeptical of AI productivity gains is that I primarily used ChatGPT and the idea that could make anyone 2x more productive was laughable.

Switching to Claude Code completely switched my perspective.

The models and tool use matters.

Mar 21 at 2:45pmWeb
Show threadMastokarl 🇺🇦1d ago

@carnage4life Yes, Claude Code is mindblowing. But honestly, the combination of Cursor and GPT5.4 is similarly mindblowing.

Something happened in December. The new models - Claude 4.5, 4.6, GPT 5.3 codex, 5.4, together with an agentic IDE that uses them well, is significantly stronger than before.

Before, I found it a fun toy. Now more and more colleagues are reporting that it works well on LARGE codebases, for tasks they didn't expect it could handle.

Show threadMisuse Case1d ago
@carnage4life I do think coding is one place where AI can potentially be very useful, and even eliminate some common bugs and vulnerabilities in code (like buffer overruns). But programming and building systems is about more than coding, and I don’t see AI getting good at those aspects of the work any time soon.