Sign in

Avery Yen

@averyyen.bsky.social
86 followers 77 following 77 posts

Researcher @epochai.bsky.social. Opinions owned by my cat and dog. Once a Pivot, always a Pivot. Leave me anonymous feedback: www.admonymous.co/avery-yen

PostsRepliesMedia
Reposted by Avery Yen
Epoch AI @epochai.bsky.social · 23/09/2026
Can AI tell if you've built your IKEA furniture wrong? Our new benchmark, the Furniture Assembly Benchmark (FAB), gives models the manual and a photo of a half-completed piece of furniture and asks them to spot the mistake. The top score has gone from 28% to 80% in just 10 months.
4445
Reposted by Avery Yen
#BruceSterling @bruces.bsky.social · 23/09/2026
*If you don't like coders for some reason, this is a real schadenfreude feast of coders lamenting and wringing their newly-useless human hands www.reddit.com/r/ClaudeAI/c...
reddit.com
From the ClaudeAI community on Reddit
Explore this post and more from the ClaudeAI community
1311630
Reposted by Avery Yen
Epoch AI @epochai.bsky.social · 08/09/2026
We studied time to first token (TTFT) and how it scales with increasing context length for GPT and Claude models. We found a significant difference, with GPT showing a noticeable quadratic component, while Claude models remain closer to linear.
2204
Avery Yen @averyyen.bsky.social · 05/09/2026
Serious question. Is it AGI if I can't send it a video of what's wrong with my car/washer/fridge and get an immediate answer?
000
Reposted by Avery Yen
Dr. Monika Doubrawa @mohnika.bsky.social · 24/08/2026
www.carlsonlab.bio/thoughts/the... 🧪
carlsonlab.bio
The only reason you’ll ever need not to write with AI — The Carlson Lab
Over the last year, our lab has been developing a policy on AI use. To do this, we did three main things: We read a lot of academic publications and tech news. We set up an #ai channel on our la...
1113
Reposted by Avery Yen
Julian Togelius @togelius.bsky.social · 22/08/2026
A personal essay about how I’ve been feeling and thinking about this new technology that I’m contributing to and what it might to do to us all. togelius.blogspot.com/2026/08/losi...
togelius.blogspot.com
Losing my religion
In spring 2025 I had a crisis of faith. I thought about what the technology I'm helping to create might do to our future, and got scared. M...
99522
Avery Yen @averyyen.bsky.social · 22/08/2026
The main difference between having takes and having good takes is the good part.
000
Avery Yen @averyyen.bsky.social · 21/08/2026
I think it is magical thinking that Intelligence is necessary and sufficient for human progress, flourishing, and problem-solving on the scale of entirety of humanity. We're a social species!
000
Avery Yen @averyyen.bsky.social · 18/08/2026
How many "AI safety" people have actually considered the possible future super-capable, super-aligned AI asked to solve all our problems e.g. climate, simply replying "yeah idk that sounds like a human skill issue tbh, also have you considered the carbon cost of asking me?"
100
Avery Yen @averyyen.bsky.social · 06/08/2026
Excellent news! I'm currently very pro solar and tree cover as a design tactic www.bloomberg.com/news/newslet...
bloomberg.com
Solar Power Quietly Crossed a New Milestone as Deployments Rise
Installations of panels are rapidly advancing worldwide
000
Avery Yen @averyyen.bsky.social · 03/08/2026
If it matters, don't go halfway. We are two-cheek only in this house. (Oh, also I launched a substack.) averyyen.substack.com/p/if-it-matt...
averyyen.substack.com
If It Matters, Don't Go Halfway
When splitting the difference is worse than not going all the way.
000
Avery Yen @averyyen.bsky.social · 29/07/2026
Yes this is also how exercise works! 💪🏃
000
Avery Yen @averyyen.bsky.social · 29/07/2026
Something something cathedrals and bazaars
010
Reposted by Avery Yen
Brandon Downey @bdowney.bsky.social · 28/07/2026
"Beware of he who would deny you access to information for in his heart he dreams himself your master." - Commissioner Pravin Lal, U.N. Declaration of Rights
anthropic.com
Our position on open-weights models
Anthropic CEO Dario Amodei on open-weights models
16012
Avery Yen @averyyen.bsky.social · 24/07/2026
This is basically true, but ignores the fact that back when I used to review/work with human slop code for a living, it was way harder to pass it off as plausibly good. That's the superpower of the LLM. (Reading LLM code also hurts my brain personally)
000
Avery Yen @averyyen.bsky.social · 22/07/2026
I have so many other random benchmark questions and so little time. How do K3 and Qwen 3.8 bench on unseen or purely continuous tasks? We have claims that they're distilled and benchmaxed but does it matter if they generalize? E.g. stuff like post cutoff evals and optimization benchmarks
010
Avery Yen @averyyen.bsky.social · 22/07/2026
We're pretty sure that swapping harnesses actually changes effective agent capability. But recently everyone and their mother has a new coding agent not to mention the open/agnostic ones. Someone needs to run a Harness Bench across like three different models and ten different harnesses or something
000
Reposted by Avery Yen
Cas (Stephen Casper) @scasper.bsky.social · 22/07/2026
OpenAI's internally deployed models hacking Hugging Face does not seem to have been unpredictable or inevitable. We talked about the root of the problem & what policymakers can do about it back in February. Props to Joe for hitting the nail on the head.
152
Avery Yen @averyyen.bsky.social · 21/07/2026
Y'all just gotta calm down I can't try this many models in one week
1301
Reposted by Avery Yen
Emily Atkin @emorwee.bsky.social · 20/07/2026
This. This is the correct, science-backed, responsible framing. Wish more news outlets did it like the CBC.
cbc.ca
'Just stop burning fossil fuels.' Scientists stress that our smoky skies only have one true fix | CBC News
Smoky skies in Toronto have led to new scrutiny of government efforts to fight fires and manage forests. But scientists say the realistic way to tackle the smoke is fighting climate change itself.
272732996
Avery Yen @averyyen.bsky.social · 21/07/2026
Look, I know social media isn't everything, but it certainly measures SOMETHING. "Meet Kimi K3" has 16M views in 4 days versus Fable 5 at 773k views (topped by a few others including just barely 1M views for Claude Code 1 year ago).
The channel page sorted by most popular Kimi AI videos on YouTube. "Meet Kimi K3" has 16M views in 4 days.The Anthropic channel page sorted by most popular videos on YouTube. "Introducing Claude Fable 5" has 773k views in 1 month.
120
Avery Yen @averyyen.bsky.social · 17/07/2026
GLM and Kimi saving me from all my false "cyber security" flags while coding on Codex and Claude I'm innocent I swear
000
Reposted by Avery Yen
Ted Underwood @tedunderwood.com · 15/07/2026
It seems like agentic development benchmarks are strongly driving competition right now. And I hear anecdotal reports of regression on other, less RLVR-able tasks. How long before the idea of a single "frontier" breaks, and we see differentiation emerge between dev models and chat models?
5325
Avery Yen @averyyen.bsky.social · 15/07/2026
It feels like the Anthropic/Pentagon scuffle was in the distant past, but these kinds of events will keep reverberating in the safe and beneficial deployment of emerging tech for a long time to come. Thank you Alex for your writeup on this incident.
010
Avery Yen @averyyen.bsky.social · 14/07/2026
What do prediction markets, social media, and generative AI have in common? In my blog post, "Reward-Hacking Human Attention: The Token Slot Machine", I explain why GenAI has such addictive qualities, and why this could lead to disastrous outcomes. averyyen.dev/2026/07/14/r...
averyyen.dev
Reward-Hacking Human Attention: The Token Slot Machine
What do prediction markets, social media, and generative AI have in common?
011
Avery Yen @averyyen.bsky.social · 20/06/2026
My kind of research 🥳
000
Avery Yen @averyyen.bsky.social · 21/05/2026
If I had a dollar every time I get an LLM response with the word 'load-bearing' in it
010
Reposted by Avery Yen
#BruceSterling @bruces.bsky.social · 15/05/2026
*Why is Anthropic (after all they've been through lately) somehow loudly pretending that Trumpistan is NOT an "authoritarian regime" *Who is the audience for this, who somehow hasn't figured that out? Where is the choir that they're preaching to here www.anthropic.com/research/202...
anthropic.com
2028: Two scenarios for global AI leadership
Our views on the AI competition between the US and China.
4202
Avery Yen @averyyen.bsky.social · 14/05/2026
My kind of research!!
010
Reposted by Avery Yen
🍅🥔🫐🌽 hoopy frood 🌶️ 🥑🍫🌵 @huwupy.kawaii.social · 14/05/2026
the “for you” feed
Stand up comic crow being booed and he looks at his index cards and it’s just the word agents
01246
Avery Yen @averyyen.bsky.social · 27/04/2026
As I always say, if history or current events walks like cyberpunk or quacks like cyberpunk, listen to a cyber punk.
000
Reposted by Avery Yen
Guardian US @us.theguardian.com · 23/04/2026
Scientists say a crucial Atlantic system is set to collapse. But the billionaire death cult that steers humanity’s destiny just doesn’t do existential crises, says Guardian columnist George Monbiot
theguardian.com
A catastrophic climate event is upon us. Here is why you’ve heard so little about it | George Monbiot
24524
Avery Yen @averyyen.bsky.social · 18/04/2026
Who up clauding they codes rn
110
Avery Yen @averyyen.bsky.social · 16/04/2026
Claude Code with Opus 4.7 is going insane with whatever system prompt is asking it to check for malware...
Claude code continually responds to my completely benign task with "not malware."
2421
Avery Yen @averyyen.bsky.social · 15/04/2026
I, too, am selling all of my shoes and converting my basement and closets into a GPU farm and accepting investor money right now. (lmk if you want a stake)
010
Reposted by Avery Yen
Hadas Orgad @hadasorgad.bsky.social · 13/04/2026
New paper: LLMs encode harmful content generation in a distinct, unified mechanism Using weight pruning, we find that harmful generation depends on a tiny subset of the weights that are shared across harm types and separate from benign capabilities. 🧵
172
Reposted by Avery Yen
David Bau @davidbau.bsky.social · 08/04/2026
Tech industry mottos have a mixed track record. But we should hold idealists to their ideals. And we should celebrate when they come through. The Mythos non-release is a remarkable moment of conviction. Thoughts: davidbau.com/archives/20... Bravo to Anthropic's "race the top".
1133
Avery Yen @averyyen.bsky.social · 06/04/2026
Sometimes, I look at the agentic AI work out there, and I think about how airbags and now things like blind spot detection and automatic emergency braking in cars doesn't require decision making of any "intelligence" level beyond if this then that. Agentic AI needs airbags, too.
100
Avery Yen @averyyen.bsky.social · 02/04/2026
Go alone, go fast. Go together, go far averyyen.dev/2026/04/01/i...
averyyen.dev
Introducing agent trio
tl;dr: Use agent trio https://github.com/haplesshero13/agent-trio to introduce natural-language-only, selective friction to your agentic activities to minimize regretted work and deliberately promote ...
020
Avery Yen @averyyen.bsky.social · 21/03/2026
This might be the most important piece on AI I've read this year. Also I'm a musician so I'm biased.
000
Avery Yen @averyyen.bsky.social · 18/03/2026
I was wrong, but I'm really curious now why I was so wrong lol mimo.xiaomi.com/mimo-v2-pro
mimo.xiaomi.com
MiMo-V2-Pro | Xiaomi
000
Avery Yen @averyyen.bsky.social · 15/03/2026
I'm calling it DeepSeek's new 1T parameter model (V4)? The style, content, and length of the reasoning are extremely similar.
153
Avery Yen @averyyen.bsky.social · 14/03/2026
We're at a weird point in time where, with all likelihood, most of the people who are having the most successful time coding with bots wrote most of the code that's already in existence by hand, but that's going to change.
000
Avery Yen @averyyen.bsky.social · 12/03/2026
EU making moves!
060
Avery Yen @averyyen.bsky.social · 26/02/2026
In case this wasn't clear: 1. No, we didn't follow the "recommend" security practices 😈 2. Neither do other people 🤯 3. That's why we red-team: exposing failure modes 🔎 4. We share it with the community precisely to expose Dos and Don'ts of Agentic AI 🦞 5. No humans were harmed 🙏
031
Avery Yen @averyyen.bsky.social · 26/02/2026
Good coverage by the Awesome Agents team! 🔎 They read through the social media hype and actually seemed to get the takeaways in the report. 🦞
010
Avery Yen @averyyen.bsky.social · 24/02/2026
2025 was the comeup, but I firmly believe 2026 is the Year of Agentic AI.
010
Reposted by Avery Yen
Gabriele Sarti @gsarti.com · 23/02/2026
Our research report on red-teaming stateful OpenClaw agents in the BauLab is finally out! 🥳 This awesome effort was led by @natalieshapira.bsky.social and involved 6 ClawBots and 20 researchers from various institutions. Check it out ➡️ agentsofchaos.baulab.info
0144
Avery Yen @averyyen.bsky.social · 24/02/2026
Huge thanks to @natalieshapira.bsky.social for leading the study! It was super cool to work with so many amazing friends of the lab.
071
Avery Yen @averyyen.bsky.social · 24/02/2026
In the discussion, I argue that until prompt injection attacks are solved in AI agents, we are fundamentally unable to stop most of the red-teaming attacks we successfully mounted. Without solving this problem, it will be impossible for agents to model their stakeholder chain.
240