GodelNumbering

6 ms·

1.
GoBench: Evaluating LLMs on 9×9 Go using KataGo opponents as Elo anchors(rolandgao.com)
2 points by GodelNumbering 2h ago|1 comments
2.
by GodelNumbering 5d ago
Some months ago I was evaluating command output compressors to integrate into Dirac[1] as that seemed like an easy win that would compliment and compound with Dirac's other mechanisms. I tested rtk among these and it was actually a net
3.
by GodelNumbering 6d ago
Tangential to the subject, but this is a bluesky post, containing a screenshot of an X post, which itself starts with "in a detailed Mastodon post"...
4.
by GodelNumbering 13d ago
> That's quite common with many models Such as? I can't think of any. Diminishing returns, yes. Occasionally flat, yes. Downright regression, no.
5.
by GodelNumbering 13d ago
The most interesting part, even more than ARC 3 score, to me is that this is the first model I recall seeing that scores lower on Max than High reasoning effort on some coding benchmarks: Terminal-Bench 4.0: High (57.9%), Max (56.7%) DeepSW
6.
by GodelNumbering 13d ago
I decided to front run and added support for it in Dirac (coding agent) a couple of hours ago, using best guess pricing: input/output/cache: $10/$50/$1.
7.
by GodelNumbering 14d ago
Do you have instructions on how to run a custom harness against this? There are none on the linked page. I want to run Dirac ( https://github.com/dirac-run/dirac ).
8.
by GodelNumbering 15d ago
Interesting, even if we were to ignore the cache-hits, reads and output, the reasoning cost (aka test time compute) per task should remain a fully comparable metric - it went from $1.25 (Fable5) to $1.48 (+18.4%) for an improvement signific
9.
by GodelNumbering 15d ago
The price reduction comes from the cache read pricing falling from $1/M to $0.25/M, which means that Fable 5.1 now costs half of Opus's cache read costs ($0.5/M). This gives a lot of credit to the theory that Anthropic d
10.
by GodelNumbering 22d ago
1.2 TB/s bandwidth of M5 Ultra comes from two dies of M5 Max (each 614 GB/s) connected together using 4.4 TB/s inter-die fabric. For a non-quantized Deepseek V4 flash on an ultra, I would estimate about 1000+ tokens per secon
11.
by GodelNumbering 24d ago
This is a plug, but relevant. I recently added a 'build native tools on the fly' functionality to Dirac ( https://github.com/dirac-run/dirac ) that works like: 1. You can use the '/new-tool' and
12.
by GodelNumbering 1mo ago
I agree in principle that the buck has to stop somewhere, but taking full responsibility for the actions of a fundamentally statistical system that you didn't build and have no interpretability of is uncomfortably risky for most partie
13.
If your agent commits a crime, who is responsible?(signalbloom.ai)
37 points by GodelNumbering 1mo ago|104 comments
14.
by GodelNumbering 1mo ago
I think introductory here means more or less permanent but they can't publicly admit there are no takers at a higher price. Anthropic for instance announced a couple of days ago that they are making Sonnet's 'introductory pri
15.
by GodelNumbering 1mo ago
The corresponding OpenAI post https://openai.com/index/previewing-ultrafast/ There is no pricing info, which could mean it's "if you have to ask..." territory or they are simply gauging interest bef
16.
by GodelNumbering 1mo ago
> introductory price They should call it 'face saving pricing after we realized just how terribly did we mis-price the flash 3.5' > since google betrayed us with those price hike, people already spent their time making their
17.
by GodelNumbering 1mo ago
https://xcancel.com/finkd/status/2086755195535413696 "... Soon we'll also release the weights for Muse Spark 1.2, our latest foundation model..." This is bigger news - good for self hosting enthusia
18.
by GodelNumbering 1mo ago
Am I missing something or the evals do not compare it to the baseline deepseek-v4-flash? Without a baseline comparison, it is hard to tell what works well and what doesn't
19.
by GodelNumbering 1mo ago
Apologies, I had no idea. I lookup host image and that site came up. I don't think the site itself is of NSFW nature.
20.
by GodelNumbering 1mo ago
If that was it, wouldn't Claude go directly to the canonical source?
21.
by GodelNumbering 1mo ago
I just checked Cloudflare for SignalBloom ( https://www.signalbloom.ai , which I own and operate). Over the last 72 hours, Claude-searchbot [1] alone fetched ~205,000 pages. Sent exactly 1 referral. There is a lot of free financia
22.
by GodelNumbering 1mo ago
You should play Blood on the clocktower
23.
by GodelNumbering 1mo ago
I think that's a fair offering tbh
24.
by GodelNumbering 1mo ago
If there were, do you believe it would be in their interest to answer this publicly?
25.
by GodelNumbering 1mo ago
Correct, but it was preview release. I was referring to GA in my comment.
26.
by GodelNumbering 1mo ago
This is one of the interesting aspects the 'AI job loss' community doesn't account for. As the technology unlocks things, more startups are created. And even at a lower nominal engineer-to-work ratio, overall demand for talen
27.
by GodelNumbering 1mo ago
So, in last several months, all the prominent names Google lost: Demis Hassabis (technically still with google but these things are usually presented with a spin), Jeff Dean, Sanjay Ghemawat, Oriol Vinyals, Quoc Le, Noam Shazeer, John Jumpe
28.
by GodelNumbering 1mo ago
I disagree to the maximum extent possible. I have lived in 6 countries. Americans to me are by far most friendly and supportive of new ideas. When you are visiting a place, you see it with rosy eyes. You are not forced to deal with the day
29.
by GodelNumbering 1mo ago
I have been writing a 'paper' [1] on an adjacent topic for months now. At some point, I decided to make it an empirical paper vs position paper. I am still chasing the experiments (when I get some free time waiting for agentic loo
30.
by GodelNumbering 1mo ago
Thanks, I will reach out. I have also posted for contributors on localllama https://www.reddit.com/r/LocalLLaMA/comments/1vg40w8/anyone_... > We can do a mix of general use (as in user stories) plus a

More ›