zaptrem

6 ms·

31.
by zaptrem 10mo ago
This sounds like the same basic voice systems we’ve had for 15 years. Idk if that counts as modern “AI”
32.
by zaptrem 11mo ago
Not sure where that math is coming from. Assuming it's true, you're ignoring that some users (me) already pay 10X that. Btw according Meta's SEC filings: https://s21.q4cdn.com/399680738/files/doc_fin
33.
by zaptrem 11mo ago
I've fed thousands of dollars to Anthropic/OAI/etc for their coding models over the past year despite never having paid for dev tools before in my life. Seems commercially viable to me.
34.
by zaptrem 11mo ago
Latency may be better, but throughput (the thing companies care about) may be the same or worse, since every step the entire diffusion window has to be passed through the model. With AR models only the most recent token goes through, which
35.
by zaptrem 11mo ago
iPhone X’s removal of the home button was a pretty big change, made navigation more fluid and felt very futuristic.
36.
by zaptrem 11mo ago
I read the daily breeze article and can’t find where they claim the released water was cleaner than it was before?
37.
by zaptrem 11mo ago
They don't have prefix caching? Claude and Codex have this.
38.
by zaptrem 1y ago
The problem with that is we want the model to learn to deal with its own mistakes. With continuous diffusion mistakes mostly look like noise, but with what you’re proposing mistakes are just incorrect words that are semantically pretty simi
39.
by zaptrem 1y ago
When text diffusion models started popping up I thought the same thing as this guy (“wait, this is just MLM”) though I was thinking more MaskGIT. The only thing I could think of that would make it “diffusion” is if the model had to learn to
40.
by zaptrem 1y ago
> There are about 936 tokens with very low L2 norm, centered at about 2. This likely means that they did not occur in the training process of GPT-oss and were thus depressed by some form of weight decay. Afaik embedding and norm params a
41.
by zaptrem 1y ago
> perhaps Apple will deign to make their ridiculously overpowered SOCs usable for general purpose computing They've been doing exactly this since the first M1 MacBooks came out in 2020.
42.
by zaptrem 1y ago
The bad scrapers would get blocked by the wall I mentioned. The ones intelligent enough to break the wall would simply take the easier way out and download the alternative data source.
43.
by zaptrem 1y ago
If they don’t have the intelligence to go after the more efficient data collection method then they likely won’t have the intelligence or willpower to work around the second part I mentioned (keeping something like Anubis). The only problem
44.
by zaptrem 1y ago
While that’s a reasonable opinion to have, it’s a fight they can’t really win. It’s like putting up a poster in a public square then running up to random people and shouting “no, this poster isn’t for you because I don’t like you, no lookin
45.
by zaptrem 1y ago
If people are truly concerned about the crawlers hammering their 128mb raspberry pi website then a better solution would be to provide an alternative way for scrapers to access the data (e.g., voluntarily contribute a copy of their public s
46.
by zaptrem 1y ago
Are there still servers running games? Not that it's really necessary since CS2 is basically CSGO with better smoke effects/lighting.
47.
by zaptrem 1y ago
It's useful for hours-long long-context debugging sessions in Claude Code, etc.
48.
by zaptrem 1y ago
The image model (GPT-Image-1) hasn’t changed
49.
by zaptrem 1y ago
A few versions of that overview were not incorrect, there actually was another Dave Barry who did die at the time mentioned. Why does this Dave Barry believe he has more of a right to be the one pointed to for the query "What happened
50.
by zaptrem 1y ago
The “payment” can come in the form of lower rates, with penalties if you stop.
51.
by zaptrem 1y ago
Does HN auto re-write headlines using an LLM or something?
52.
by zaptrem 1y ago
I find it helps me hit these moments faster since I can watch it go and cut it off when I realize the issue.
53.
by zaptrem 1y ago
Claude Code converted me from paying $0 for LLMs to $200 per month. Any co that wants a chance at getting that $200 ($300 is fine too) from me needs a Claude Code equivalent and a model where the equivalent's tools were part of its RL
54.
by zaptrem 1y ago
Using this to bypass scraping protections would be genius.
55.
by zaptrem 1y ago
I hope there's still some sciencing left we can do better than the AI because I start to lose it after playing games/watching tv/doing nothing productive for >1 week.
56.
by zaptrem 1y ago
Alternatively: -- In such a future, people will have minimal income (possibly some UBI) and therefore there will be few who can afford the products and services generated by AI -- Corporate profits drop (or growth slows) and there is demand
57.
by zaptrem 1y ago
Don’t quote me on this but I think the A series also lacks support for virtualization.
58.
by zaptrem 1y ago
Why would they be liable for off-label uses of their product?
59.
by zaptrem 1y ago
Sorry, “”Midtown West””
60.
by zaptrem 1y ago
Google Maps zoomable map I found: https://www.google.com/maps/d/u/0/viewer?hl=en&ie=UTF8&msa=0... I wonder why it seems to circumvent Hells Kitchen?

More ›