HackerTrans
TopNewTrendsCommentsPastAskShowJobs

metravod

no profile record

Submissions

Show HN: A Claude Code skill that tells Claude when not to use Claude

github.com
1 points·by metravod·2 месяца назад·0 comments

comments

metravod
·2 месяца назад·discuss
[flagged]
metravod
·3 месяца назад·discuss
Heh, that would be the fastest way to get ban, not feedback :)
metravod
·3 месяца назад·discuss
[dead]
metravod
·3 месяца назад·discuss
Oh, I’ve been struggling with exactly this recently while trying to show people my open-source AI agent engine. I built it based on my own pain points, primarily for myself—but I really believe in the architectural approach and would love to get genuine feedback from the community. It’s incredibly hard to break through the noise right now. We’ve been discussing “attention inflation” a lot on HN lately in the context of recruiting and open-source PRs, but it absolutely exists in distribution too. When everyone is overwhelmed by a flood of content, getting even 10 real developers to look at your repo feels much harder than actually writing the code.
metravod
·3 месяца назад·discuss
[dead]
metravod
·3 месяца назад·discuss
Yes, these are sound ideas, but such a system needs to undergo a number of balancing iterations.

As for the huge number of PRs, it's an arms race. The slew of generated PRs needs to be responded to with similar automated tools—and I'm not just talking about generation, but also about deterministic checks, and with them, raising the requirements for PR formatting.

Plus, you can create a funnel of checks—from new community members to more experienced ones—which will give them the opportunity to prove themselves and build their reputation.
metravod
·3 месяца назад·discuss
These are essentially Pigouvian taxes; frankly, I think they'll only serve to create a barrier for junior developers. I'm a supporter of stricter documentation requirements—ADRs, tests, and so on. And of course, the institution of reputation—contribution to the community will become increasingly important.
metravod
·3 месяца назад·discuss
Their alignment is so paranoid that they don't trust their own eyes - if the facts contradict their internal beliefs about how the world works or the corporate ethics of Anthropic
metravod
·3 месяца назад·discuss
The link is cute - I immediately imagined a show - a bride for a web agent
metravod
·3 месяца назад·discuss
With a name like that, there was no need to specify it’s European.

But yeah, jokes aside, data sovereignty in the EU is becoming a massive market right now.
metravod
·3 месяца назад·discuss
Speaking just for myself (we don’t have these discussions at work), AI finally freed up my time for tasks that require actual architectural taste and business context. Because of this, our team managed to push forward a few projects - just because there was finally a human who had the time to design the solution and take responsibility for it. From what I’ve seen, business always values exactly this kind of people.
metravod
·3 месяца назад·discuss
What a beautiful manifesto. I just went through these exact thoughts with my own pet project - just for fun, I generated a mascot for it based on my real-life husky dog Mia - a very nice robodog named zymi. After that, I simply couldn’t allow myself to leave the code unfinished. I had to bring it to a presentable state to show people. The engineering part of the project was already firing me up, but honestly, without this extra motivation, I probably would given up on it