integrations at @HeyMiloAI evals | world models | mechanistic interpretability
Joined August 2026
- Tweets620
- Following6.6K
- Followers1.4K
- Likes32K
Pinned Tweet
For about $5 and 5½ minutes you can strip almost all the guardrails from a 12B Gemma 4.
Model: huggingface.co/pasangimhana/…
Report: pasangimhana.github.io/resea…
pasan retweeted
i forgot about this until @ScottWu46 reminded me, but back in january 2017 when we were 19, scott and i had lunch in SF
at lunch, we talked about how our IMO/IOI/math competition gang could be the next paypal mafia. i knew the competition kids were geniuses by any standard, and felt confident in time we could take the tech world by storm
fast forward nearly 10 years later, and many of our friends from that time are some of the most exciting people in AI and technology (@ScottWu46, @akshat_b, @chameleon_jeff, Jerry Ma, @randomjohnnyh, @0xShitTrader, @stevenkplus1, @thejessezhang, @vickie_ye_, @alexwei_, @demi_guo_, @d_m_ziegler, @hughbzhang, @_mohansolo, @MarkSellke, and many others i'm sure i'm forgetting)
i'm now at that age where i reminisce about everyone i grew up with, and i've known most of these people since middle school (via MATHCOUNTS!). even though we were scattered around the country, we would gchat about math problems and hang out on AoPS, then we did math/algo camps together, and then we did internships together.
very crazy to think about the degree of sheer brilliance of this group when we were just teenagers, and so special to see so many go on to leave their mark on the world.
note to parents: if your kid is on gchat all day solving math problems with some anonymous group of cross-country asian kids, (s)he may just turn out alright.
pasan retweeted
Introducing GPT‑6 Cyber: a major advance in security research that pairs stronger capabilities with better model alignment.
The model powered the Linux kernel research we announced yesterday, and security teams at companies like Mozilla, Palo Alto Networks, Calif, and Cisco—alongside our own at OpenAI—are already putting it to work.
GPT-6 Cyber is available to select Daybreak Red customers and we are expanding access over time.
openai.com/index/introducing…
pasan retweeted
Omarchy 4.5 is finally here! Traditional desktop mode, M1/M2/M3/Neo compatibility, 50 languages, 7-second installs, and a million other features, fixes, and improvements. omarchy.org/news/2026/11/oma…
upcoming cyber model 😁
Our security researchers built an autonomous cyber vulnerability discovery, exploitation, and patching pipeline.
In a single run with an upcoming cyber model, it validated and generated patches for 948 local privilege escalation vulnerabilities in the Linux kernel at a total API cost of $1.2 million.
5
543
these latest set of system prompt level guardrails are the reason Opus is kinda cooked now
An update on building games with Claude Code:
Last week some of you ran into Claude Code pushing back on game projects it should have just built. We'd added a short rule to the Opus 5.5 prompt to steer Claude away from recreating specific existing games, but it was also kicking in when you referenced one just to describe the mechanics you wanted.
We've updated that rule in 2.1.317. You can name a game to describe what you want and Claude should now build an original game in that genre. If you ask for a copy of the original, it'll offer to build something similar instead.
We want Claude to be as helpful as possible when you're building something original, but also to protect your work if someone else tries to copy it. We'll keep improving this, so let us know if Claude turns down work that's yours. Thanks to everyone who shared feedback!
Overhyped clickbait.
"Natively reading and writing weights" only works on tiny NNs.
Given only Qwen's weights, Astra made no edit.
What worked was standard interp, done autonomously.
Looks like no one has read the blog and just going off of the Tweet and the hook
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
pasan retweeted
The future is wild
Yeah we should add this. @bot kick off a job on beebox with opus 5.5 to build this. Use my full stop skill
pasan retweeted
BENCHMARKED: LFM3-1.7B by @liquidai.
Liquid AI published 1.6 GB for the KV cache at 128K context. I ran it on my Mac mini and measured 1.6 GB.
The whole model at 128K uses 3.15 GB of GPU memory, under 5% of the 64GB. It still generates 80 tok/s.
Only 6 of its 24 layers keep a full KV cache. The other 18 use short convolutions or sliding-window attention.
llama.cpp (Metal), Q4_K_M, M5 Pro 64GB, median of 3 runs
- 512 ctx: 184 tok/s, 0.09s to first token
- 4K ctx: 178 tok/s, 0.72s to first token
- 32K ctx: 141 tok/s
- 128K ctx: 80 tok/s
Q8_0 does 125 tok/s at 512, so 4-bit is 47% faster.
pasan retweeted
So one of our ambassadors @divagr1925 has been working on something cool
He found a way to use Qwen's Gated DeltaNet state as memory, and built a 32-bit RISC-V CPU directly into the model's weights. So instead of relying on predicting the answer to a calculation, the model can execute instructions internally. It can also run C and MicroPython programs.
He calls it Qwenputer, love the name btw. The results so far are pretty impressive. On one arithmetic benchmark, accuracy jumped from 48.6% to 100%.
Pretty cool to see developers experimenting with Qwen's architecture. Great work, my friend!
pasan retweeted
Introducing 𝙻𝙵𝙼𝟹: our new generation of Liquid Foundation Models. With a new tri-hybrid architecture and new multimodal encoders, the LFM3 family is the ideal model suite for specialized intelligence.
> 6 model instances, across text, audio, vision, and omni modalities, generative and decision models.
> 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱: text, 256k context window.
> 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱: vision-language model with video-input support for spatial understanding
> 𝙻𝙵𝙼𝟹-𝙰𝚞𝚍𝚒𝚘-𝟸𝙱: tool calls, voice chat, reasoning in 18 languages.
> 𝙻𝙵𝙼𝟹-𝙾𝚖𝚗𝚒-𝟸𝙱: an omni model combines vision, audio, and language.
> 𝚍𝟷-𝟸𝙱 and 𝚍𝟷-𝙾𝚖𝚗𝚒-𝟸𝙱: decision models with typed outputs and calibrated probabilities in a single pass. Omni adds audio input.
> fine-tune on 1 GPU.
> a new tri-hybrid architecture
> architecture optimized for 14 processor classes across GPUs, CPUs, and NPUs.
> 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱 is a beast. It performs competitively with Gemma 4 12B (6x larger), and Gemini 3.5 Flash Lite, and GPT-5.6 Luna, with 1.9B parameters.
> 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱 is #1 on the Pareto frontier of device intelligence 𝚙𝚒𝚙𝚎𝚝𝚝𝚎.𝚕𝚒𝚚𝚞𝚒𝚍.𝚊𝚒
Every modality for specialized intelligence. Pick the one that fits your use case, customize it, and run it on-device.
1/ 🧵
very cool stuff coming soon from @HeyMiloAI
1/ Building HeyMilo has given us a unique window into the world of work.
We’ve helped interview millions of individuals across a wide range of roles, industries, and backgrounds. Along the way, we’ve learned a lot about the work to be done in these wide range of roles.
It’s led us to a bigger question. 🧵
6
613
search and rescue obviously
Purdue University's IDEAS Lab built an autonomous drone that tracks runners through dense forests completely without GPS or cloud networks.
By running 3D LiDAR mapping and obstacle-avoidance AI entirely on-chip, it can weave under tree canopies and predict target movement through blind spots in real time.
Massive implications for next-gen search & rescue and edge AI.
pasan retweeted
google wrote a paper, published it for anyone to consume, then watched all eight authors leave & two outside labs capture most of the frontier model equity.
the most valuable research paper in history was effectively a gift from google's balance sheet to its competitors.
re mark able.
pasan retweeted
today with @Thom_Wolf, we introduce the alignment hub, a central place for misalignment incidents designed with three goals:
1. set a standard for minimal transparency that is easy for labs to follow
2. improve clarity by centralizing information and sourcing
3. track the progression of both the incidents and how they are reported
we believe there is a minimal level of transparency required to properly understand incidents and their impact. we want this to be as simple as possible for labs to disclose, and we noticed that even things that seem obvious like reporting the incident or detection date are not standard today. we believe that the public should know the answers to these basic questions: was there monitoring (CoT? network?), did it happen during training? eval (general? cyber? alignment?), deployment? and more.. we also think that labs should be more transparent than this, and we think that this minimal level of transparency should evolve as the field advances
having a central place also helps a lot to connect incidents and build understanding. a good example is the irregular incidents which are all linked and caused by a human error of allowing internet during a CTF eval. we also track progression over time, things like the % of incidents disclosed by labs vs third parties, the time it takes for labs to disclose, and a tentative transparency score.
this is a field moving very fast and we will do our best to make sure this hub evolves with the challenges ahead. this is still in "beta" (lots of vibecoded components even if we want this to be agent first!), but we want to publish it early so that people iterate on it. we welcome feedback and contributions, we see this as a community effort!
our hope is that this makes it easier for researchers, labs and the general public to report, learn, understand and analyze safety incidents
huggingface.co/spaces/witnes…
