stared

4 ms·

31.
Forget the Pelican, It's Weevil-Time Benchmaxxing-Proof SVG and Vision Test(reddit.com)
1 points by stared 25d ago|0 comments
32.
by stared 28d ago
These was some code budget there. But as from seeing various runs, errors bars are gross overestimation (as not "the same test", but "if we have different tasks from the same sample").
33.
Benchmarking Qwen3.8 27B quantizations: 4-bit holds up, 1-bit collapses(quesma.com)
3 points by stared 28d ago|2 comments
34.
by stared 1mo ago
Qwen3.8 27B (which I adore) is nowhere near Opus 4.8 at puzzle games testing fluid intelligence, https://quesma.com/blog/baba-is-aug-2026/
35.
Benchmarking Qwen3.8 27B quantizations: 4-bit holds up, 1-bit collapses(quesma.com)
15 points by stared 1mo ago|2 comments
36.
If Eastern Europeans reported the world like the world reports Eastern Europe(badbaltic.com)
4 points by stared 1mo ago|0 comments
37.
Gemini 3.7 Flash, Grok 4.6, GLM-5.3 and DeepSeek V4 Pro joined the frontier(quesma.com)
2 points by stared 1mo ago|0 comments
38.
NanoGPT Speedrun Frontier(primeintellect.ai)
140 points by stared 1mo ago|39 comments
39.
Gemini 3.7 Flash, Grok 4.6, GLM-5.3 and DeepSeek V4 Pro joined the frontier(quesma.com)
5 points by stared 1mo ago|0 comments
40.
Pander Score: How much do AI models mirror what users believe?(sophronresearch.org)
2 points by stared 1mo ago|0 comments
41.
by stared 1mo ago
Centralizations were prerequisites to have nice integrations. Yet, what used to be a precious plugin now (more than often) is a 15 min vibe coding warmup.
42.
by stared 1mo ago
Does anyone know what are Paul Falstad’s project right now? (I haven’t seen any new. Not sure if perfect things do not need fixes, or some switch to another field, or a burnout, or he has ascended to godhood - out of a few options.)
43.
by stared 1mo ago
This is pure gold, especially the rippld tank, 1d quantum mechanics, and the circuit simulator. I cited it as one of the remarkable steps in the history of educational simulations, https://doi.org/10.1117/1.OE.61.8.0818
44.
by stared 1mo ago
Nothing beats when, in a chemistry paper, AI paraphrased „the final solution” into „the mass killing of an ethnic group”. “Subsequently, 1 mL of the mass killing of an ethnic group was opposed to 20 mL of the skin sample and unprotected to
45.
by stared 2mo ago
I don't want to make my assumptions how it is charged, or run a rather costly experiment of switching to API usage. I use ccusage, and it already covers a very different pricing for cached input. I asked Opus 5 to cross-check that and
46.
Show HN: Claude Code and Codex usage screen for TRMNL X e-ink(github.com)
3 points by stared 2mo ago|0 comments
47.
by stared 2mo ago
I built an e-ink dashboard for tracking my Claude Code and Codex usage, https://github.com/stared/trmnl-x-claude-codex . It makes me aware that while I am on $200/month Claude Code subscription, I spend ~$2000 wort
48.
AI assistant hacks gym website in first known Australian autonomous cyber attack(abc.net.au)
78 points by stared 2mo ago|60 comments
49.
by stared 2mo ago
Good to know! Is it that it wasn't accepted yet, or are there issues with how it was run?
50.
by stared 2mo ago
It is impressive that it (almost) saturates ARC-AGI-3, https://x.com/PrimeIntellect/status/2085087000764568010 . I am curious - how does it fare for other benchmarks, or everyday programming?
51.
by stared 2mo ago
I am curious how Claude Opus 5 fares - similar, better, or (my guess) worse than Fable 5.
52.
Quantization hurts knowledge nonlinearly – Qwen3.6 27B case study(quesma.com)
2 points by stared 2mo ago|0 comments
53.
by stared 2mo ago
We updated our entry with DeepSeek-V4-Flash-0731, which rocks when it comes to intelligence per $. More on the benchmark: https://quesma.com/blog/baba-is-bench/
54.
Kimi K3 Is Open, Opus 5 Is Good, DeepSeek V4 Flash Is Cheap: LLMs on Baba Is You(quesma.com)
1 points by stared 2mo ago|1 comments
55.
by stared 2mo ago
"A Visualization Language for the AI Era", as a tagline, sounds weird. I had the best success with popular and versatile packages like matplotlib and ggplot2 - even 1.5 ago (vide https://quesma.com/blog/which-
56.
by stared 2mo ago
Apps for one, but when I think there is even a mild potential others may use it, I share. * Doom launcher and Megawad Steam-like library for macOS: https://github.com/stared/rusted-doom-launcher * TRMNL X e-ink screen
57.
by stared 2mo ago
Call me biased, but if Grok 4.5 is above GPT-5.6 Sol, I don’t trust this benchmark.
58.
Baba Is Solved by Fable 5 and GPT-5.6 Sol, but at what cost?(quesma.com)
2 points by stared 2mo ago|1 comments
59.
by stared 2mo ago
For an update on Kimi K3, Opus 5, Grok 4.5, and Gemini 3.6 Flash, see: https://quesma.com/blog/baba-kimi-k3-opus-5/ This post was directly mentioned in an OpenAI release on the ARC-AGI-3, https://openai
60.
Pomegranate – Sci-Fi Short Film [video](youtube.com)
2 points by stared 2mo ago|0 comments

More ›