고정된 트윗
₭คցε
184 posts


@libriscent Remember when people were like "random acts of kindness" and now they're like "kindness is cringe?"
Hmmmm.....
English

@SergioRocks AI has solved coding, not software engineering. Your agents need to follow a reliable software engineering methodology in order to succeed.
English

Most people use AI tools like this:
- Describe what you want
- Try a few prompts
- Fix what breaks
It works for getting started. But it doesn’t build production-ready products.
The better approach is simple.
First, define the spec:
- The PRD for each product feature
- The engineering standards
- The data flows
- The UI screens
- How to handle errors and edge cases
Then use tools like Copilot, Cursor or Claude Code to build your product, feature by feature.
That way, these tools are not guessing. They are following a clear path. And that’s what makes it reliable to build software in the age of AI.
AI speeds up execution. A clear process makes that execution consistent.
English

Not only simple bugs, but also categorical failures - repeated evasions or compromises made by inherently lazy LLMs. The first pass at act101.ai produced very subtle fakes, and I had to audit the code for this specific failure mode in order to correct it. Different models make different category errors. Sometimes it feels more like sculpture than engineering.
English

Something vibecoders cannot understand:
you need hundreds or thousands of prompts to be able to build a production-grade app, even if you use Claude code
and then, even after that you will still have bugs
so thats why is important to know what decisions you take while developing, because AI doesn't know all
AI knows so much information that it often makes the wrong decision
So the chances of a non-developer vibecoding a production-grade app are so slim
better learn all these stuff from courses or youtube first, then try to vibecode your own production app
English

Cancelled both my Claude Code Pro and ChatGPT Pro for this.
Kimi K2.6 is just as good for my side projects as Opus or GPT 5.4 were. The price for this is crazy low, and there are a bunch of models I can try (like DeepSeek).
Bonus: I'm moving away from building everything on Claude Code - now that both @opencode and @cursor_ai have their SDKs open, I feel I can rebuild the agentic workflows I built for Claude Code in a more platform-independent manner.
Lotto@LottoLabs
Update on Opencode Go It’s great value for $5/month, there’s really no reason not to do the first month. At $10/month it’s still good value and gets you access to all sota OS models. You can’t daily drive it without hitting limits on the big models but w/ Kimi x3 you won’t hit limits unless you’re insane. Overall highly recommend the first month, then make your own decision.
English

@bas_fijneman Sometimes you don't know what you can build until you've already started. act101.ai started as a simple tool until I caught the vision of a universal tool supporting 163 languages and full codebase tools.

English

@yashmp2004 Adding features to act101.ai, the ultimate token-saving tool for agentic coding: one native executable with tree-based code navigation, refactoring and analysis tools. Free for personal/OSS projects.

English


























