manmal

6 ms·

1.
by manmal 3d ago
GPT doesn’t do all of that all that much when it itself is the implementer. RL has made implementation and reviewing two different behavior sets.
2.
by manmal 4d ago
The thing that changed with 5G is that my iPhone 13 Pro‘s battery has been draining 30% faster. Enable the hotspot and it gets burning hot.
3.
by manmal 4d ago
The critical IP that isn’t in the weights yet, is often a few hundred to a few thousand lines of code. All the rest, surrounding code, is infrastructure that is indeed in the weights.
4.
by manmal 5d ago
Even benchmark improvements over Sol are not that great (and they certainly tried). If we exclude the weird ARC-AGI situation.
5.
by manmal 5d ago
Unless you’re making 3D models of apartments and oneshotting trivial games, I‘m not convinced Astra writes better code than Sol. I‘m way more often disappointed about Astra forgetting some aspect that Sol just nails.
6.
by manmal 5d ago
I‘m using Sol all day every day. It’s on average doing better than Astra at what I need it to do. Astra is like an absent minded professor - gives good direct responses, but too inconsistent and forgetful for my codebase.
7.
by manmal 5d ago
It is a convenient time, just as coding capability seems to reach the upper end of the S curve.
8.
by manmal 5d ago
The kind of HIIT I'm doing is only 10-20s bursts, to stay within ATP-PCr energy metabolism (little lactic acid production). A short sprint followed by 2 minutes rest. That's hard to achieve with a treadmill. https://en.
9.
by manmal 5d ago
Can you really do HIIT on a treadmill? It doesn't sound safe to sprint on a machine. Re your observations. Are you referring to elderly people? I'm not sure this study is about them. My mental model of this mechanism is that relat
10.
by manmal 5d ago
I have to disagree on both points. Long races don't do this for me - that's also the point of the paper. And I have had more close calls of spraining an ankle when running uphill. The problem with hills also is that you might want
11.
by manmal 6d ago
I like sprinting a couple times up our hill, ca 10s each, 4-6 times. It’s full out, and when I‘m untrained (like now), I‘m quite exhausted afterwards. But the dopamine boost I get from this is immense. It lasts 24-48h. I‘m in a general good
12.
by manmal 6d ago
When it takes longer than 36h, just revoke and submit again. It's probably just stuck in some queue.
13.
by manmal 7d ago
Exactly my thoughts today. They have to make it cheaper after demoing what’s possible initially.
14.
by manmal 12d ago
Depends what actors you trust. They are equally bad but personally I‘m way more comfortable with the Discord story.
15.
by manmal 13d ago
If you came at it from the perspective of imitating what the human brain does, we now have a very very powerful speech center and short term memory, and vision catching up. The other parts are missing. I‘m sure that’s being heavily research
16.
by manmal 13d ago
I‘m using mostly the browser automations for things like that now. Same for research - ChatGPT is banned from reading many pages, but Codex can read anything I can. But isn’t it funny that Cloudflare is blocking AI on their pages, but on th
17.
by manmal 13d ago
Have you? It’s a dumb model that gets a lot wrong. Codex voice is IMO only good for when I cannot dictate into the good models. The delay introduced by letting it work with a good model kills it for me.
18.
by manmal 14d ago
I guess `caffeinate` doesn't go that far because it can easily lead to overheating.
19.
by manmal 14d ago
I think I'd use Discord instead. Clankers are happy to set it all up for you.
20.
by manmal 14d ago
If author runs the local model for privacy reasons, then I don't understand why they give Telegram access to all their conversations. It's well known that Telegram doesn't end-to-end encrypt bot accounts.
21.
by manmal 16d ago
The Amphetamine app from the App Store keeps a MacBook awake even with the lid closed.
22.
by manmal 17d ago
They are better at many things I‘m a novice at. At least at the operational level. I can’t rely on the semantics being fully correct because they tend to get subtle things wrong or just don’t ask. Like tax forms - not a good idea to put the
23.
by manmal 17d ago
Codex with computer use, yes. GP mentioned it. > Codex requires you run the desktop app and additional authentication requirements. The result of this has been codex is almost always relegated to fleet worker rather than orchestrator. I’
24.
by manmal 17d ago
One of the Codex employees on Twitter promised (threatened?) that the next gen of Codex will be cloud focused, and less local. I wonder now if that means it will be a more powerful GPT Work.
25.
by manmal 17d ago
I got a 404 when heading to the examples.
26.
by manmal 17d ago
I haven’t used Haskell much. I thought to try it with agentic help, but then read this post that its compilation times might make it unviable. But what’s your experience?
27.
by manmal 17d ago
They have exactly one guy as the expert in this video and he contributes all the facts. That’s bias.
28.
by manmal 17d ago
That video is quite biased. Just tell your clanker to critique it. It will point out that those numbers are quite cherry picked, and leaving out important details. E.g. the „GHG would reduce by only 2.6%“ study assumes that all crop product
29.
by manmal 18d ago
My problem with memory is that it goes stale, and updates to facts are often not changing all locations of that fact. Such a system should make it easier to maintain a single source of truth, and versioning, no?
30.
by manmal 18d ago
I guess you also get very high bandwidth that way? I‘m not sure that would come for free though.

More ›