@thsottiaux Gave it an api key to something I don’t care about and it just refused refused refused to use it. Put it in env, nope. Cmon man fix auth already.
@nuke7official@KrispKiwi@v0xium Bro it’s not that hard. Just tell it to refactor your project to be more all the things. Have it come up with ways to do that, evaluate them, then have it implement one or more of those things. Based on risk tolerance, test suite coverage, expected future state. Just like before
@valyala@oliversgriff@v0xium I mean I’m in private repos all day and do some dedicated cadence of refactoring into something maintainable and extensible. Depends on the project. But for a greenfield I’ll slop it up for a week or two, then put agents on refactoring. It can produce high confidence with tests.
@ClassicGamerTWR@eleganttap Constant refactoring, comprehensive test suites (unit, functional, e2e) and helpful docs has allowed me to solo build our to 500k LoC in under 60 days. Adding features with Astra is a dream. Fewer regressions hit prod than the pre clanker era.
Just canceled @claudeai max 20x plan, have been a max subscriber since the launch of Claude code. Astra, ChatGPT desktop, remote control, it’s all just so much better than Claude now. Also @DarioAmodei is a lunatic. Not aligned with @AnthropicAI anymore. Ciao
@601dawg@SwissBitcoinPay@SoloSatoshi The entire game theory of software security has just flipped 180 degrees in the past 60 days. Expect these things to keep happening.
@anthdm Sounds like a prompting issue. Or your SKILL.md files. I had to do some tweaks on my setup but it’s really flying now. For me it’s hitting the sweet spot. I’ve been a professional software dev for 20 years, not just “vibing”. I don’t do one shot development.