@nilszeilon

Swedish software engineer. I like to tinker around with technology. Inventor of the smashed cevapi.

Joined August 2013
  • Tweets1.1K
  • Following428
  • Followers250
72Media
In a digital world software dependencies is the mother of all fragilities, limit your exposure as much as ~~humanly~~ agenticly possible.
20
Nils Zeilon retweeted
Nassim Nicholas Taleb
@nntaleb16h
Wake up. Every profession deemed white-collar, the middle class dream, is about to be decimated by AI, at a much faster pace than anticipated, with monstrous social consequences.
908
2,383
487
19,573
1,360,741
I would like to thank my procrastination that I don’t have to stay up all night stressing out about the internet exploding with vulnerabilities.
37
Nils Zeilon retweeted
Sashi
@sashikantsingh_Oct 21
Anthropic plans to give each kid in America portion of its $80B in losses.
246
1,193
156
22,877
539,470
I love tailscale
55
Nils Zeilon retweeted
dinos
@din0s_Oct 2
wake up babe the freaking pope got schmidhuber'd
55
316
37
3,899
177,698
Looks amazing! first order invariant jev-like model afik.
AVB
@neural_avbSep 28
Releasing bev-decider-0.4B: a lightweight causal decision model that is architecturally invariant of input choice order. Its built on the first 20 layers of Qwen. Processes the prompt and choices in one parallel forward pass. Runs superfast at ~50ms even on my 3-yr old mac (no speedups in the demo). This started as for a ed youtube video, but I was having too much fun, so I'm releasing everything. Dataset, code, weights are all open source. The video (full project, start to finish) drops this week! Bev is pip installable and comes with a Jev-compatible /v1/systemone api. You can find benchmark details in the HF model page. :) Github: github.com/avbiswas/bev-trai… Model: huggingface.co/avbiswas/bev-… Data: huggingface.co/datasets/avbi…
59
Why don’t you just put its sandbox in another sandbox, like an onion
73
I'm building an RLM harness for myself
1
54
My harness reviewing itself, its getting there!
30
Opus 5.5 is great, i really hope they don’t sandbag it a week from now
50
I got to say it, i really like the new opus model.
2
1
60
This is they way, focus on a cheaper and stronger frontier. Drop the end of the world nonsense.
13
Finally, i got high hopes
Claude
@claudeaiSep 22
Replying to @claudeai
Opus 5.5 communicates more naturally, addressing some of the most common feedback we heard on Opus 5. It puts the most important information up front and follows the writing rules you give it, which makes long sessions easier to follow.
68
What model did they distill to beat astra?
Salio
@Mr_SalioSep 19
🚨 Deepseek v5 Leak: Beats Astra > DeepSeek is reportedly preparing an imminent V5 launch > it could match or beat Fable 5.1 and Astra > Expected to deliver much stronger performance at a lower cost > DeepSeek is reportedly keeping the open-weight strategy Could DeepSeek V5 become the new king of open-weight AI?
1
104
Nils Zeilon retweeted
the tiny corp
@__tinygrad__Sep 19
Replying to @chamath
The cyber incidents weren't enough for them to get the regulatory capture they wanted. They need to make a bio incident next. I hope our government is on top of this, their ideology has no limits to what they will do for "good"
40
235
37
3,065
448,685
Made a little tree visualization of my custom RLM harness, today I will try out some jev ideas to improve the harness
66
Make everything faster.
48
system one models is such a good name, from thinking fast and slow by kahneman. Jev is your gut decisions and LLMs your slower more logical decisions.
45
unreal how effective ads on youtube are for shutting the app down
45
Load more