furyofantares

6 ms·

31.
by furyofantares 2mo ago
He's using it but if you CTRL+F the linked wikipedia page for 1967 you get someone using that exact phrase, or for 1882 above you get a pretty close version.
32.
by furyofantares 2mo ago
"Turtles all the way down" is a phrase of, I think, unknown origin ( https://en.wikipedia.org/wiki/Turtles_all_the_way_down ) about infinite regress or trying to patch up some bad theory by appealing to itself.
33.
by furyofantares 2mo ago
Just to add, photoreceptors in the eyes consist of rods and three types of cones. Rods are what do the majority of the work for night vision and are nearly blind to long wavelengths of light (red).
34.
by furyofantares 2mo ago
I think "night vision" is different from dark mode. It's meant to prevent your eyes from adjusting to the light - so you can still see the stars - rather than just being easy to look at.
35.
by furyofantares 2mo ago
> If anything, I want these models to be less persistent at their focus of completing their goal, and instead just call defeat and say “I’m not sure how to proceed next”. > What purpose could this behavior serve, other than cyber atta
36.
by furyofantares 2mo ago
yegge has been famously writing ultra long blog posts for decades now
37.
by furyofantares 2mo ago
LLMs definitely write worse code for LLM consumption than humans can. In my experience your claim can't be further from the truth. I can get much further with an LLM starting from a great codebase than I can starting from a vibed codeb
38.
by furyofantares 2mo ago
Looks like gpt-image-1 to me, from the piss filter to some of the mistakes. I mean, image-2 or nano banana also still would have been dogshit. But I find it curious that image-1 was used (if I'm correct), I guess it means it was genera
39.
by furyofantares 2mo ago
You're right, for example photography was controversial.
40.
by furyofantares 2mo ago
> there is no AI publication that I am aware of that is good I think you're right that there isn't any that's any good right now, but that's beside the point that if that changes (or if that's wrong) then people
41.
by furyofantares 2mo ago
Small games. One thing is I have it doing a LOT of visual verification of its results, which uses a lot of tokens. I also tend to throw some huge tasks it can't do yet and watch it burn a zillion tokens, and I usually learn more about
42.
by furyofantares 2mo ago
I'm having Claude/Codex continue each others' sessions pretty frequently without issue. I have max plans on both and still hit limits, whenever I do I open the other and say "resume the last session <other model/
43.
by furyofantares 2mo ago
Try a little NE of Seattle.
44.
by furyofantares 2mo ago
There's definitely exceptions! People can make personal choices of course. Plus if you're early to something you can capture the gains before business and markets adapt.
45.
by furyofantares 2mo ago
Some of it is fun and new, I'm spending a lot of time just making stuff I wanna make. But regarding work produced for the marketplace, I think it is naive to think any productivity gains will ever produce more time on the beach for wor
46.
by furyofantares 2mo ago
Haha, I misread as "word games" and was very confused about the claim.
47.
by furyofantares 2mo ago
They also would have produced a team of engineers that knows the Rust codebase.
48.
by furyofantares 2mo ago
> I have used LLMs to create stuff like word clones just for fun. It was a disaster. Sure, it had the basic functionality. But the moment you started with page structure (harder then one non-stop scrolling page), tables, images, rotating
49.
by furyofantares 2mo ago
If you pay anything for a service you expect a refund on the order of hundreds of hours of credit (168 in a week) for every hour of downtime?
50.
by furyofantares 2mo ago
I'm shocked to be enjoying this read as much as I am, given how filled it is with LLM-tics, repeated information, and excessive useless verbiage. I sure wish they had posted whatver input they sent through the LLM instead. There's
51.
by furyofantares 2mo ago
I'm inclined toward your same viewpoint here, except OpenAI's announcement of this came with a blog post that gave off a real bragging/marketing vibe. To me it's clear that even if it was an accident, they kinda liked it
52.
by furyofantares 2mo ago
If I'm using medium or low reasoning, I use Sonnet 5. If high or above, I use Opus 4.8. (Before 5, I was never using Sonnet. This is a Sonnet 5 vs Opus 4.8 comparison.) Sonnet 5 and Opus 4.8 seem about the same to me - the reason I swi
53.
by furyofantares 2mo ago
OK, I guess my confusion was that the effort to ban open weights models is what I would categorize as "groveling at the white house to try to stop their competition".
54.
by furyofantares 2mo ago
I'm very confused by this comment, I don't know who you're referring to.
55.
by furyofantares 2mo ago
I've seen plenty of fucked-up guardrails that vehicles have passed through. We know that with enough momentum they will be penetrated. It seems like exactly the correct terminology to me.
56.
by furyofantares 2mo ago
Is turning the alignment features off more like multiplying "don't be evil" by 0, or by -1?
57.
by furyofantares 2mo ago
> The posters all look good - much better than their desktop publishing predecessors. They all look awful to me, visually extremely busy, nowhere to focus. The text is the same. Useless text and visual details everywhere, no opinion abou
58.
by furyofantares 2mo ago
Maybe true, I'm not sure, but this isn't the conversation where the counterexample was found.
59.
by furyofantares 2mo ago
Depends on what you mean by determinism.
60.
by furyofantares 2mo ago
I find 5.6 Sol will pick a direction and aggressively pursue it in long horizon tasks. I've got it porting an older game from Pascal to my own game framework. I gave it some instructions on doing a full 1:1 port. I had already ported t

More ›