manmal

13 ms·

331.
by manmal 10mo ago
I‘ve just tried the dxastgraphx one in pi with Opus 4.5. This was its response: I couldn't find a library called dxastgraphx in either pip (Python) or npm (JavaScript) package registries. This library doesn't appear to exist
332.
by manmal 10mo ago
Both the before and after are better done manually. What you are describing is fine for the heck of it (I‘ve vibe coded a whisper related rust port today without having any actual rust skills), but I’d never use fully vibed software in prod
333.
by manmal 10mo ago
A surgeon in our family got basically all his (private) clients from Google. Spend was multiple k per month. If you consider that one surgery brings in 7k in revenue, then those numbers actually make sense. He's retired now but did thi
334.
by manmal 10mo ago
Isn’t this design DOA though, with LFP and upcoming solid state batteries?
335.
by manmal 10mo ago
Why would someone do this to themselves? Such a trip doesn’t sound like a lot of fun. Wouldn’t long term sauna have similar effects and be way less straining on the cardiovascular system?
336.
by manmal 10mo ago
But it doesn’t sound standardized at all if so much customization work is needed.
337.
by manmal 10mo ago
Is the reason that it works for big tech that those can spare dozens of engineers to make it work?
338.
by manmal 10mo ago
Our kids‘ school received some old surface laptops for free through some initiative. I‘ve been tasked to take care of them. And of course, WiFi mysteriously stopped working for all of them, after ca 1 week of use. It turns out that a recent
339.
by manmal 10mo ago
PRs don’t really hurt autonomy if stacked branches are used routinely. Those do hurt speed, yes, but not autonomy. PRs are so important that I‘d never skip them within a team.
340.
by manmal 10mo ago
I totally agree with this part. Right now, we are in a transitioning phase, where parts of a team might reject the notion of using AI, while others might be using it wisely, and still others might be auto-creating PRs without checking the o
341.
by manmal 10mo ago
The one on Generate AI seems a bit outdated. This was before Claude Code was released.
342.
by manmal 10mo ago
Code reviews are not only about raising quality, but mainly about communicating changes to the wider team. Suggesting to eliminate code reviews when LLM use is so rampant is also quite uhm courageous.
343.
by manmal 10mo ago
Well one of those has to work well or you would have fainted by now.
344.
by manmal 10mo ago
Congrats, your metabolism (FAO, gluconeogenesis, insulin sensitivity, …) works well. Don’t assume that’s the case for everyone, because it’s well researched that this is highly individual.
345.
by manmal 10mo ago
It’s also accelerated muscle loss, even heart tissue. I‘m not saying it’s not worth it, but this should be kept on the radar for sure.
346.
by manmal 10mo ago
That is the reason they severely limited Claude Max subscriptions. Some users racked up 1k+ in API equivalent cost per day.
347.
by manmal 10mo ago
If you read my comment, I was describing the consolidation part.
348.
by manmal 10mo ago
I‘ve channeled my inner Larry David into a prompt to make fun of y‘all ;)
349.
by manmal 10mo ago
Keep track of test coverage and ask it to delete tests without lowering coverage by more than let’s say 0.01 percent points. If you have a script that gives it only the test coverage, and a file with all tests including line number ranges,
350.
by manmal 10mo ago
I get where you're coming from. But you'll likely get better results by starting fresh and letting it read key files or only just a summary of the project goals/spec. And then implement the next feature building up on the pre
351.
by manmal 10mo ago
I wouldn’t recommend that (after having tried it for a while). It is very easy to get the impression that the changes are sound, because on their own they look good. But often when looking at the final result in a git diff tool, the holes
352.
by manmal 10mo ago
Why don’t you reset context when working on something else?
353.
by manmal 10mo ago
Yeah a bridge has a plan that it’s built and verified against. It’s the picture book waterfall implementation. The software industry has moved away from this approach because software is not like bridges.
354.
by manmal 10mo ago
I have this mental model of LLMs and their capabilities, formed after months of way too much coding with CC and Codex, with 4 recursive problem categories: 1. Problems that have been solved before have their solution easily repeated (some w
355.
by manmal 10mo ago
I struggle to see where Pro is better than 5.x with Thinking. Actually prefer the latter.
356.
by manmal 10mo ago
It’s obviously an amalgamation that’s weighted in favor of your interests.
357.
by manmal 10mo ago
What LLM generate is an amalgamation of human content they have been trained on. I get that you want what actual humans think, but that’s also basically a weighted amalgamation. Real, actual insight, is incredibly rare and I doubt you see m
358.
by manmal 10mo ago
Swift has been OSS for 10y now, and still only runs usably on Mac and Linux (the only platforms Apple cares about - coincidence?).
359.
by manmal 10mo ago
Just because Claude ran the commit command, doesn’t mean it wrote the code. That’s just a nasty marketing hack from Anthropic.
360.
by manmal 11mo ago
Thanks for this idea!

More ›