nl

19 ms·

451.
by nl 5mo ago
Kilo (the open source coding agent) tested Deepseek v4 Pro and Flash vs Opus 4.7 and Kimi K2[1]. It did ok, but scored substantially less than Opus. It also cost nearly as much, even with the current launch promo pricing for Deepseek. That
452.
by nl 5mo ago
Quote: > “The two AI superpowers are going to start talking. We’re going to set up a protocol in terms of how do we go forward with best practices for AI to make sure nonstate actors don’t get a hold of these models,” Bessent told Joe Ke
453.
We Tested DeepSeek V4 Pro and Flash Against Claude Opus 4.7 and Kimi K2.6(blog.kilo.ai)
1 points by nl 5mo ago|0 comments
454.
US/China talks to make sure non-state actors don't get a hold of these AI models(cnbc.com)
3 points by nl 5mo ago|1 comments
455.
by nl 5mo ago
Claude is available on Azure and GCP already. https://azure.microsoft.com/en-us/blog/introducing-anthropic... https://docs.cloud.google.com/gemini-enterprise-agent-platfo... Claude has been availa
456.
by nl 5mo ago
> OpenAI and Anthropic both project huge losses for the next couple years, it's not hard to find. No. Anthropic at least expects to be profitable this year: > Anthropic expects its gross profit margin, which measures how much rev
457.
by nl 5mo ago
To quote Elon Musk in court: > OpenAI’s counsel asked Musk whether xAI has ever “distilled” technology from OpenAI. > Musk: “Generally AI companies distill other AI companies.” > “Is that a yes?” Savitt asked. > Musk: “Partly.”
458.
by nl 5mo ago
> None of these companies, be it Anthropic, OpenAI, xAI, Google, Meta, Microsoft, are profitable in the AI department, Citation needed. All reporting is that they are profitable on the inference side and all the VC money is going to buil
459.
by nl 5mo ago
Do you have any examples or data on the discriminatory power of the model for tool use? The examples are things like "What is the weather in San Francisco", where you are only passed a tool like tools='[{"name&quot
460.
by nl 5mo ago
I dug some numbers out on this: https://news.ycombinator.com/item?id=48116241 $2000/year on average seems a good estimate, and $3000/year is something that could be possible. This is higher than I expected.
461.
by nl 5mo ago
I went down a whole rabbit hole trying to find the numbers for this. If you Google there are lots of different numbers reported. According to [1], the average American household spend $682 on airfares in 2024, plus an additional $199 on &
462.
by nl 5mo ago
> The very first thing they show this new machine doing is helping people shop for clothes using AI. > No one is doing that, these people don't exist I don't know what world you live in but I personally know at least 4 peopl
463.
by nl 5mo ago
Don't worry about being nitpicky! I'm going to out-nitpick you.... Actually.... I write and publish my own benchmark for this stuff. It's an agentic SQL benchmark which isn't in the training data yet and I've foun
464.
by nl 5mo ago
I think knowledge is power. I think that the more people who try local models (especially the larger ones) the better. I sometimes get the impression that many people claiming that local models are as good as frontier models work in "t
465.
by nl 5mo ago
I think it's useful to be realistic about what you can do with a local model, especially something as small as the 9B the author is using. A 9B model is around the level of Sonnet 3.6 - it can do autocomplete and small functions but it
466.
by nl 5mo ago
Sonnet and Opus are from Elon Musk (given the people he's hired it seems likely it is approximately true). Mythos is quite widely spoken about.
467.
by nl 5mo ago
But they don't? Mythos is a 10T model. Opus is a 5T model. That's not an exponentially growing amount of compute but it is achieving exponential improvements (eg from Mozilla: https://blog.mozilla.org/en/priva
468.
by nl 5mo ago
> It ceases to become tacit as soon as it is collected. I'm not sure. It it is collected via preferences then it isn't necessarily something that can be communicated (except in the LLM's latent space). That still feels tac
469.
by nl 5mo ago
Depends when it happens. I'm sure they'll take up whatever capacity is available. At the moment computer providers are charging more for outdated H100 capacity now than when the H100s were new. That capacity is going to the smal
470.
by nl 5mo ago
The labs are spending hundreds of millions of dollars hiring people doing many fairly random (but economically valuable) tasks to collect this tacit knowledge for RL. It works really well.
471.
by nl 6mo ago
True. No one "needs" the internet or computers, right?
472.
by nl 6mo ago
> It wasn't like today, where most software is slop Did you ever use Window 3.1? Or any windows software? Are you forgetting the multiple crashes per day? Or the incessant playing with system.cfg to get MSDOS drivers to work? And
473.
by nl 6mo ago
> Scrambling? Yes. To quote: > Anthropic CEO Dario Amodei said his company tried to plan for 10-fold growth. But revenue and usage increased 80-fold in the first quarter on an annualized basis, which he says explains why it’s been so
474.
by nl 6mo ago
The point is that Anthropic is already a decent way into eating through all that capacity, and it's based on real revenue.
475.
by nl 6mo ago
Say what you like about Sam Altman, but given how Anthropic is scrambling to sign capacity deals for compute we can sure say he was right about the capcity build out needed.
476.
by nl 6mo ago
I was there. It's mostly rose tinted glasses. There were some amazing feats. But it was slow and frustrating. Like you wouldn't believe how long things took. In the 90s most technical documentation was in actual physical books
477.
by nl 6mo ago
That makes the title of another of his posts very ironic then: "Automatic programming" https://antirez.com/news/159
478.
by nl 6mo ago
Antirez wrote Redis. That is "production-code with critical concurrency" To quote another of his posts: > I fixed transient failures in the Redis test. This is very annoying work, timing related issues, TCP deadlock conditions,
479.
by nl 6mo ago
The author said "You know what was the biggest realization of all that?" > For high quality system programming tasks you have to still be fully involved, but I ventured to a level of complexity that I would have otherwise ski
480.
by nl 6mo ago
I think this is a bad framing. Javascript running on a page can use a feature that requires a model to be downloaded. I have pages that use it, or other LLM models via LiteRT or HuggingFace transformers.js. I try to warn the user, but that

More ›