redox99

8 ms·

151.
by redox99 2mo ago
Do not, my friends, become addicted to sleep. It will take hold of you, and you will resent its absence!
152.
by redox99 2mo ago
I think you overestimate the disadvantages of sleep. It's very species dependent, for some it might be a disadvantage because you're more vulnerable to predators, for others it's an advantage because you spend less energy. If
153.
by redox99 2mo ago
It wouldn't matter if it was stored only on their servers. As long as they offer the feature to downgrade a chat to a dumber model that can be jailbroken (and the downgrade keeps the reasoning), this trick works.
154.
by redox99 2mo ago
I think verifiable tasks like math will soon be like Chess. Machines way beyond human intelligence.
155.
by redox99 2mo ago
Yeah but open source is often a few guys working on their free time, or a large company open sourcing libraries that they wouldn't be able to sell anyway, or donations from corporations that are at best in the order of a million dollar
156.
by redox99 2mo ago
> This result is some evidence that AI will not solve RH soon. If there were any easy solution hiding in plain sight then it probably would have found it. There's no way you can conclude that. Yes, "Fable 2" or whatever th
157.
by redox99 2mo ago
There's nothing fast about GPT 4.1. It's ~50 tps AFAIK. Of course it doesn't use reasoning, but you can run modern models without thinking as well. GPT 5.6 Sol without reasoning should destroy it in intelligence.
158.
by redox99 2mo ago
Pro: It's web chat only, I haven't tried it so idk. I think it's useful for Math and that kind of stuff, but I only do programming. Ultra: Might be useful, the one time I tried it, it was extremely wasteful in terms of tokens
159.
by redox99 2mo ago
Yeah, it's hard, there's so many permutations. But most are bad so it narrows it down. Luna: Can be useful if price sensitive, always use with AT LEAST high effort. But codex plans are very generous, so just ignore it. Terra: Forg
160.
by redox99 2mo ago
It's not just about it "being smart enough". It's about there being actual user demand when it needs to compete with the shiny new model. A 10 year old iPhone is probably good enough, but is there demand for it? In a vac
161.
by redox99 2mo ago
Yeah I had to double check, o3 feels like it was ages ago. But GPT 5 came out Aug 7, so it's only one day off from my 1 year ago cutoff!
162.
by redox99 2mo ago
Is there any LLM from exactly one year ago that would be worth running? In Aug 2025 you had - OpenAI o3 - Opus 4.1 - Gemini 2.5 Pro - Grok 4 Even if those were almost free to run, you'd be way better off with Deepseek flash 0731 or GPT
163.
by redox99 2mo ago
Yes. You're better off running Luna Max or Sol Medium and never touching Terra.
164.
by redox99 2mo ago
So instead of just getting an answer, I have to wait until it asks me, and I have to type back a response? Extremely annoying.
165.
by redox99 2mo ago
Yes
166.
by redox99 2mo ago
Terra is awful and not cheap enough to make up for it. I'd suggest using luna or sol.
167.
by redox99 2mo ago
Because the model can't read your mind and know if you want a quick answer, or an hour long deep dive.
168.
by redox99 2mo ago
I'm not sure I agree 1) Back then, even as a free user you'd be able to use the strongest model (even if with tight limits). Now, you need to pay to use Sol, and you need to pay to use Opus or Fable. It does seem fairly premiu
169.
by redox99 2mo ago
Yeah it's definitely better in that regard, but it has the other disadvantages: oversized for single passengers, and also personally it feels like public transport, it's the opposite of luxurious.
170.
by redox99 2mo ago
But why include Opus then?
171.
by redox99 2mo ago
Other than attention optimizations and other minor changes, the top Chinese models (which are way better than gemini) have basically the same architecture as GPT2. Of course RL is key for agentic workloads, but I'd say it's correc
172.
by redox99 2mo ago
The cybercab would be almost perfect if it weren't for it's low height. It's easier for the elder to get into and out of taller vehicles.
173.
by redox99 2mo ago
I watched a bit of it and it's mostly stating their own conjectures as facts. There's nothing objective about it, it's just pushing their own personal agenda.
174.
by redox99 2mo ago
I think the SOTA is fable max. It's up to you whether that's good enough or not. Almost all other models do make some egregious mistakes with the bike. https://static.simonwillison.net/static/2026/fable-m
175.
by redox99 2mo ago
The UI is the equivalent of "You're absolutely right!". It's a big turnoff at least to me. That toast is such an AI thing.
176.
by redox99 2mo ago
I tried the codex extension many months ago but I didn't like it. I use mostly the Codex App, or sometimes the Codex TUI inside the IDE terminal.
177.
by redox99 2mo ago
Cursor has two benefits 1) It's still an IDE. Because of pricing I mostly use Codex, but I always have a VSCode/Cursor IDE open, thus have to juggle between the two. Working directly in the IDE is more comfortable. For full on vib
178.
by redox99 3mo ago
They basically don't exist in the currently most profitable LLM market (coding). Yes, subs like codex are heavily subsidized. But API billing has massive margins and that's what enterprises pay.
179.
by redox99 3mo ago
It was the agents job to pick the model.
180.
by redox99 3mo ago
Running above the display refresh rate is only decent when you are wayyyy above it, at least double. If you have a 120hz display and run something variable around 150fps, its much worse. But 500fps does help, you're basically seeing mo

More ›