girvo

18 ms·

241.
by girvo 6mo ago
Like me: that’s what I do. Either Opus 4.7 or GLM 5.1 for planning, write it out to a markdown file, then farm it out to Qwen 3.6 27B on my DGX Spark-alike using Pi. Works amusingly well all things considered.
242.
by girvo 6mo ago
Except most of the TUIs I’m seeing are god awful with horrible input latency because they’ve reimplemented everything from scratch in python or whatever. Multiple hundreds of ms per keystroke: it sucks.
243.
by girvo 6mo ago
The counterpoint to that is that the exact same tools that are allowing this personal software creation at massive scale are also excellent at black box vulnerability analysis…
244.
by girvo 6mo ago
Climate change due to CO2 is going to be far more catastrophic.
245.
by girvo 6mo ago
We’ve trained models on JSON schemas for “tool calls”, and then built software to interpret and run those calls for the LLMs
246.
by girvo 6mo ago
People don’t actually track wall clock time, I’ve noticed.
247.
by girvo 6mo ago
Yeah people aren’t aware that we don’t see the actual traces anymore lol
248.
by girvo 6mo ago
Not to mention it being very obviously written by AI.
249.
by girvo 6mo ago
That it made it so easy to whip up UIs and access Windows APIs, having never done it before as a 10-11 year old playing with VB6... ...which I think promptly used as a younger teen to write a custom keylogger for a friend to try and steal R
250.
by girvo 6mo ago
You’re absolutely right! …LLM-isms are like nails on chalkboard I swear. Instant turn off the moment I read them. Even if they’re maybe not lol, doesn’t matter my visceral reaction is negative.
251.
by girvo 6mo ago
Stuff it in an AGENTS.md to pretend you're AI-native ;)
252.
by girvo 6mo ago
Amusingly, my Cupra Born has all its connectivity disabled... because Cupra Australia just didn't want to bring it to this country. Not a bad thing really, aside from the annoying red notification dot telling me I have no signal!
253.
by girvo 6mo ago
We’re definitely not escaping the permanent goblin underclass with this one.
254.
by girvo 6mo ago
And yet I’ve seen in production the (ab)use of HTTP where fcgi would’ve been a much better fit.
255.
by girvo 6mo ago
Ha, HN is exactly where I'd expect to see 2 digits personally
256.
by girvo 6mo ago
They appear to be backing out (for a little while) of consumer cards, but datacentre/workstation/laptop GPUs are still their focus.
257.
by girvo 6mo ago
Oh wow, I really would've expected higher memory bandwidth. That's only ~2-3x the little DGX Spark-alike I have to play with. Would've expected more.
258.
by girvo 6mo ago
> Idk, I'd like to see AI design a schematic and lay out a board. It's being worked on, and as the other commenter said, it's a data problem not a fundamental limitation. That's what's scary to me. That said the
259.
by girvo 6mo ago
Oh I'm well aware. It's lagging behind "pure" software, but is catching up. The only thing I can see that acts as a bulwark is liability, bascially. The FW work I was doing requires a human and a large amount of careful
260.
by girvo 6mo ago
> The recent maturation of AI revealed how many people in software seemingly loathe their own profession. I always had an inkling this was the case, but man it's been depressing to see it laid so bare. So many proudly screaming &quo
261.
by girvo 6mo ago
If OpenAI had won the enterprise race, then maybe?
262.
by girvo 6mo ago
> On the other hand, they were so uptight that I’m not sure “opinions are my own” would have prevented it. In my experience it didn't matter at all, they considered "you work for us, its known you work for us, therefore your op
263.
by girvo 6mo ago
Very kiki, not very bouba, as Aphyr rightfully stated.
264.
by girvo 6mo ago
It's quite well done really, but the cadence... No x. No y. No z. Just abc. Its like nails on a chalkboard...
265.
by girvo 6mo ago
Yeah I know. Which is why I wish we’d (the royal we) all stop pretending and lying about it haha.
266.
by girvo 6mo ago
Please don’t post AI generated comments :( Just write it yourself. I promise it’s worth it
267.
by girvo 6mo ago
Same way everyone gives lip service to reviewing output. I know for a fact that at work most don't, not deeply/properly. You basically can't and hit the volume that's been demanded.
268.
by girvo 6mo ago
> But the depth and rigor I think may be missing, as the article suggests. Nearly certainly. Just turns out that depth and rigour matters a lot less than I would've hoped. Depressing, really.
269.
by girvo 6mo ago
Qwen 3.6 27B is shockingly good, just to add to your point.
270.
by girvo 6mo ago
> Some people are even comparing them to compilers. A lot of people are using them as such too: the amount of people talking about "my fleets of agents working on 4 different projects": they aren't reviewing that output.

More ›