KTibow

7 ms·

31.
by KTibow 1y ago
I'm not much of an ML engineer but I can point you to the original chain of thought paper [0] and Anthropic's docs on how to enable their official thinking scratchpad [1]. [0] https://arxiv.org/pdf/2201.11903
32.
by KTibow 1y ago
> With o3 you get something that feels like a human-written bug report, condensed to just present the findings, whereas with Sonnet 3.7 you get something like a stream of thought, or a work log. This is likely because the author didn&#x2
33.
by KTibow 1y ago
Some newer ones like calc-size are also like this.
34.
by KTibow 1y ago
It sounds like Kagi asked for something simple but creative, and the author overengineered it.
35.
by KTibow 1y ago
`post` automatically releases the connection. `with` only makes sense when you use a `requests.Session()`.
36.
by KTibow 1y ago
FYI - using service workers would automatically make it work offline and possibly be easier than trying to bundle it into one HTML file.
37.
by KTibow 1y ago
Vertex is the enterprise platform. It also happens to have much higher rate limits, even for free models.
38.
by KTibow 1y ago
It might be a weighted random.
39.
by KTibow 1y ago
It's worth noting that this is a gzip bomb (acts just like a normal compressed webpage), not a classical zip file that uses nested zips to knock out antiviruses.
40.
by KTibow 1y ago
Other companies have done this (see Qwen Coder). It doesn't scale past a few disciplines like math and code though, and using mixtures of experts give you most of the same benefits.
41.
by KTibow 1y ago
Uncensored and RP tuned are somewhat different.
42.
by KTibow 1y ago
Were you using it with search enabled?
43.
by KTibow 1y ago
Archived now: https://archive.is/BxyNA
44.
by KTibow 1y ago
I'm a bit skeptical of this until it's proven that they're getting the right answers in the right ways. It could be that base models are just more random and when given 200 guesses out of 1000 possible answers tend to distrib
45.
by KTibow 1y ago
Good correction - while the SDK used has supported changing the API through environment variables for a long time, Codex only recently added Chat Completions support recently.
46.
by KTibow 1y ago
If your favorite LLM inference program can run a Chat Completions API.
47.
by KTibow 1y ago
When AI Overview was called Search Generative Experience, you could do that. You can do that again now if you have access to AI Mode.
48.
by KTibow 1y ago
Without any changes, you can already use Codex with a remote or local API by setting base URL and key environment variables.
49.
by KTibow 1y ago
If it's GPL-like its usage would be mostly confined to open source projects.
50.
by KTibow 1y ago
VS Code is tightly integrated into Copilot even if you don't install the extension these days. It even has its own button in the title bar.
51.
by KTibow 1y ago
I think this is just a quirk of how Google does live captions.
52.
by KTibow 1y ago
I wonder if using thinking models would work better here. They generally have less variance and consider more options, which could achieve the same goal.
53.
by KTibow 1y ago
Slowed down: https://youtu.be/tRUngxK2yXY
54.
by KTibow 1y ago
No.
55.
by KTibow 1y ago
Claude is better at design but isn't as strong a reasoner. I'd expect the Max plan to add a Deep Research like feature too soon.
56.
by KTibow 1y ago
I've seen some speculate that o3 is already using visual reasoning and that's what made it a breakthrough model.
57.
by KTibow 2y ago
I think they confused Chrome and Googlebot.
58.
by KTibow 2y ago
Those points are only true if you're comparing to a React app that uses bloated frameworks and inefficient list manipulation (as other commenters pointed out, `list.push(...items)` causes problems)
59.
by KTibow 2y ago
Ideally web browsers would use better defaults so you wouldn't have to do this. Firefox is good with this in my opinion: scrollbars are very subtle, don't affect content width, and adapt to the `color-scheme`.
60.
by KTibow 2y ago
I think there's a difference between code you don't understand and code that's "unimportant" / you don't care about.

More ›