Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
Bolwin
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
7 ms
·
121.
▲
by
Bolwin
8mo ago
Llms generally seem trained with the assumption that if you mention it, you want it. I don't think the instruction following benches test for this much and I don't know how you'd measure it well
122.
▲
by
Bolwin
8mo ago
You're entirely correct in that it's a different model with every message, every token. There's no past memory for it to reference. That said it can still be useful because you have a some weird behavior and 199k tokens of co
123.
▲
by
Bolwin
8mo ago
Well written. My whole life I've always insisted that everything I make, everything with my name on it have my marks on it. My decisions, my idiosyncrasies, my preferences. Whatever it was, essays, presentations, code, drawings. LLMs
124.
▲
by
Bolwin
8mo ago
That feels like a problem with your harness. What are you using it with?
125.
▲
by
Bolwin
8mo ago
You forgot to mention your tool does worse for 8/16 LLMs compared to replace? Problem is, replace has been around for so long, most LLMs are tuned for it now
126.
▲
by
Bolwin
8mo ago
This is gorgeous. I see a install on pwa, but it would be even nicer if this could be an android widget. Anyone know if this is possible?
127.
▲
by
Bolwin
8mo ago
Arc agi was never a good benchmark that tested spatial understanding more than reasoning. I'm glad it's no longer popular
128.
▲
by
Bolwin
8mo ago
Pro isn't even a model. If they actually used a model name I'd think they were just into LLMs. Chatgpt pro is a specific paid service.
129.
▲
by
Bolwin
8mo ago
Great explanation for why AI slop is so distracting. It's actively hurting my ability to find high quality things, because it's a strong signal for low effort, and yet so many good projects are starting to use it on readmes. I hav
130.
▲
by
Bolwin
9mo ago
Compared to?
131.
▲
by
Bolwin
9mo ago
Vivaldi still has it as an option. That said, I keep them combined and can't remember the last time it guessed wrong.
132.
▲
by
Bolwin
9mo ago
Odd you should say that, as that's exactly what they've been discussing
133.
▲
by
Bolwin
9mo ago
Not all. In fact a small model that has none of them but loads them on demand might be the most efficient thing
134.
▲
by
Bolwin
9mo ago
Which Kimi K2 model did you use? There's three. Also, you give models a separate "thinking" space outside their reasoning? That may not work as intended
135.
▲
by
Bolwin
9mo ago
My God those stabilized responses are sickening. If anthropic implements this, they'll kill their models' dominance in writing and roleplay. Opus 4.5 was already a step down in trying to play any character that didn't match i
136.
▲
by
Bolwin
9mo ago
People are definitely going to be sending you AI generated birthday invite posters soon. Oh and yeah, AI has already been shown to be more persuasive than the average human. It's only a matter of time before someone's paying to de
137.
▲
by
Bolwin
9mo ago
I'm assuming the climate impact of meat
138.
▲
by
Bolwin
9mo ago
They do if you're both the sender and intended recipient
139.
▲
by
Bolwin
9mo ago
Theres a heavier max plan
140.
▲
by
Bolwin
9mo ago
I have plenty more meaningful reasons to boycott companies and a limited capacity for boycotting, so no thanks
141.
▲
by
Bolwin
10mo ago
Have they improved tooling? I've yet to get any lsp working on windows
142.
▲
by
Bolwin
10mo ago
Studies have shown at least 30% are ghost jobs, which means in practice probably higher. Add the fact that they stick around longer and I can comfortably say 50%+ of the jobs people are applying to are fake.
143.
▲
by
Bolwin
10mo ago
No one's paying you to fill up gas either. Do you always go inside and ask someone to do it for you?
144.
▲
by
Bolwin
10mo ago
Scroll up to what? The whole list fits on screen for me
145.
▲
by
Bolwin
10mo ago
Mind you, you can get all that and more in a browser like vivaldi. And that market is.. small. Vivaldi doesn't have to develop a browser engine
146.
▲
by
Bolwin
10mo ago
atproto is literally one server
147.
▲
by
Bolwin
10mo ago
In their AMA moonshot said it was mainly finetuning
148.
▲
by
Bolwin
10mo ago
Is this.. actually performant? It sounds like one of those stupid but fun side projects, but there seem to be real apps using it.
149.
▲
by
Bolwin
10mo ago
MCP is overly complicated. I'd rather use something like https://utcp.io/
150.
▲
by
Bolwin
10mo ago
Wow I was just looking for finding like this, but.. can't be recharged? It would be one thing if it had like 500 hours of recording, but this has 12-15.
More ›