svachalek

3 ms·

1.
by svachalek 5d ago
Yeah subscriptions are another big one. So many profits today are generated by bleeding off customers that just can't deal with the hoops they need to jump through to get unsubscribed. Tasking that to an agent would destroy the whole m
2.
by svachalek 5d ago
The voice is the weird part. The early Grok 4 models had a very distinct presentation unlike anything else out there. Then suddenly it made a big jump in coding ability and started sounding just like every other model.
3.
by svachalek 5d ago
Agreed. It's very capable for something carrying the "flash" label, super fast, and very clear to read.
4.
by svachalek 5d ago
The target market for frontier models is CEOs who want to lay off entire departments of their company. So the long autonomy benchmarks would seem to be sending exactly the right signal.
5.
by svachalek 5d ago
Look up the output-style setting, which is a bit stronger than putting it in the system prompt. The new "concise" setting is better than the default but in practice, Claude is a very stubborn model when it comes to these patterns
6.
by svachalek 5d ago
Embeddings just convert the tokens to a vector that represents the text in an abstract semantic space. JEV goes a step further and actually processes the instructions/meaning of those embeddings to produce output, just not the usual se
7.
by svachalek 8d ago
Good point, it's similar to music and other forms of art. The qualities that the people creating it care most about often have little to do with how well it is received.
8.
by svachalek 8d ago
The image helped for me at least.
9.
by svachalek 8d ago
I think the confounding issue is that by now, millions of sessions of Claude Code and Codex are now in the training set for these models. So they have been trained to work the way these harnesses are configured, and at least in the case of
10.
by svachalek 9d ago
Cool idea. I tried using it to port a little meeting fixer cron job I vibe coded, it seemed a natural fit as its essentially trying to satisfy invariants in my calendar. It basically succeeded but Claude (Opus 5) did have some complaints: &
11.
by svachalek 9d ago
Hmm I don't know if they've updated it, but I'm not smelling any of the Claudisms I know so well. No convoluted sentences, metaphors, therapy-speak, 'it's not' constructions, "load bearing" etc etc.
12.
by svachalek 10d ago
Wow! Did not expect this blast from the past this morning. I worked with Alberto and Bob at the same startup long ago. Hello to any other Agitators who found this today.
13.
by svachalek 10d ago
1100 at online speed chess or something, could be. I'm not that deep in the chess world but everyone I know that can make 1100 in official rating can name a dozen openings and most of the known tactics, and is pretty good at applying a
14.
by svachalek 12d ago
Excuse me, please show a little more respect for my profit margins.
15.
by svachalek 12d ago
Fable 5.1 is fairly pleasant to work with, the first in a while. Too bad it's so ridiculously overkill for most tasks. They need to reel in Opus and Sonnet.
16.
by svachalek 12d ago
That's true but Claude has so many more forms of this. It tries so hard not to be sycophantic that it inserts little negging sidecars into any sentence that agrees with you. "That's true but not how you're thinking"
17.
by svachalek 12d ago
I just started playing with building a game in Godot with Fable 5.1 last weekend and the pace is crazy. Obviously there's a world of difference between what looks playable on my machine and what a AAA studio with the best human artists
18.
by svachalek 15d ago
We're not talking about a 6k virus file though. More like 10 terabytes and it needs a server that can pack all that into VRAM.
19.
by svachalek 15d ago
DigitalOcean (at least through OpenRouter) is pretty reliably bad in my experience. Fireworks can be great, but it depends on the model and on the day.
20.
by svachalek 15d ago
I don't know why mathematicians got lumped in with physicists further down the thread. In my experience this is a well known stereotype of physicists and while of course not universal you don't have to look very hard to see where
21.
by svachalek 15d ago
Ugh. To be fair to the LLMs they're trained on industry blogs, and if you spent too much time reading them you'd also think every electronic stamp album needs to scale to a billion users with millisecond regional failover.
22.
by svachalek 15d ago
Yeah I've felt that gut punch all the time. Seems to happen with every technology. In the early days of GPS, people would ask for directions to the gas station then pull it up on their phone map and go "yeah, you're right!&qu
23.
by svachalek 16d ago
I tend to agree. Software development history is a long saga of things that are bad and still popular. Think of how popular J2EE, XML, and SOAP were at some point, and they're just miserable to use or maintain.
24.
by svachalek 17d ago
Wait, isn't it ChatGBD?
25.
by svachalek 22d ago
I'd say Occam's Razor leans easily to the former as well, given the history of projects that Snowden revealed and were never shut down, plus all of the cooperation with the federal government that's being touted in recent ann
26.
by svachalek 24d ago
Those hypothetical "small government conservatives" are running up the Federal deficit at a breathtaking, unprecedented pace. What's actually going on is much simpler to explain as corruption.
27.
by svachalek 24d ago
Ah yes, the great US institution of ad tech. Mass privacy intrusion in the service of selling clicks for a nickel.
28.
by svachalek 24d ago
I think this was the biggest shock of the original ChatGPT for me. Just how completely unrobotic its voice was compared to everything we'd ever imagined in sci fi. Even that early version was also way more adept at understanding things
29.
by svachalek 25d ago
The entire output of an LLM is also part of its input. Appending to long stretches of LLM-generated text, it will continue to get more and more robotic as the style of the input gets replicated and enhanced in the output. It's possible
30.
by svachalek 25d ago
Neither book stores nor app stores pay for quality, beyond some minimum low bar. Markets are optimized for getting product out quickly, sales and marketing, and follow on effects of succeeding with a hit. If you made a better app, game, boo

More ›