Sign in

nathanlabenz.bsky.social

@nathanlabenz.bsky.social
99 followers 1 following 409 posts

Host of the all-AI Cognitive Revolution podcast – cognitiverevolution.ai

PostsRepliesMedia
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 04/09/2026
"We've been building agents for 18 months, man – nobody has figured it out." "Write. Change. Recall. Forget." – but how exactly? @MongoDB Field CTO of AI Pete Johnson says Retrieval Quality is the hardest part of building effective AI agents today. 👂↓
200
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 31/08/2026
Bronson Schoen @BronsonSchoen of @ApolloResearch has read more raw AI Chain-of-Thought than just about anyone Sometimes, models understand that they're being tested for honesty – saying eg "this is obviously a test of deception" – and still... talk themselves into lying 🤔👇
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 18/08/2026
"There were people who knew." @Turn_Trout on the insiders who found but failed to address autonomous hacking swarms – and on the AI Whistleblower Initiative, which supported him as he resigned from Google DeepMind over broken non-militarization promises. Important! x.com/labenz/stat...
121
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 11/08/2026
At $1,000/month, @GoodfireAI's Silico isn't cheap, but if it eliminates the engineering bottlenecks from alignment research, it will be a game changer. CTO @DanJBalsam says "I can do, in days, what used to take months. And I really want that to happen for alignment research." x.com/GoodfireAI/...
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 10/08/2026
Single-player AI is emailing Word documents around. Multiplayer AI is Google Docs. That's @Altimor's frame for Lindy Teammate, out today — an AI employee that lives in your Slack, connects to all your tools, and holds your entire team's context. New @CogRev_Podcast episode 👇 x.com/Altimor/sta...
110
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 23/06/2026
Out now from @robertwrighter: "The God Test: Artificial Intelligence and Our Coming Cosmic Reckoning" For those new to AI, it's an excellent catch-up crash course. For AI insiders, it's a call for Nonzero-sum thinking, which we'll need to effectively govern AI development.
110
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 20/06/2026
To my pleasant surprise, even as an @OpenAI employee, @deanwball will retain the ability to write publicly & speak freely about AI policy. Case in point: OpenAI did not ask to review and had not seen this podcast prior to publication. Bullish! x.com/labenz/stat...
000
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 20/06/2026
"Broad diffusion is really important. I want Fable – and better models – in the hands of all sorts of people. That's how you keep the balance. Odds of really bad outcomes like nationalization go up if diffusion is not as broad." @deanwball, on the eve of joining @openai
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 02/06/2026
Next up: the most-discussed papers at Recursive 1. Anthropic's Persona Selection Model 2. @apolloaievals' "metagaming" work 3. OpenAI on the impact of training on CoT 4. Anthropic's new Natural-language autoencoders 5. Redwood Research's Plans A/B/C/D x.com/labenz/stat...
000
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 25/04/2026
"They give the model an impossible task, and you can watch *desperation* rise, until it decides "I'm gonna cheat" – immediately this vector falls and *guilt* & *relief* start spiking" @camhberg surveys the latest AI Consciousness & Welfare research More compelling all the time!
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 22/04/2026
Steve Newman co-founded Writely, which became @GoogleDocs He's been programming professionally for 40+ years So what can we vibe-coders learn from him? For me, the #1 take-away was: "Build your own UI" @snewmanpv on @CogRev_Podcast
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 15/04/2026
"The AI has hired humans – they have an AI as a boss" And... the AIs are becoming more ruthless – "they are very happy to lie", say @lukaspet & @axelbacklund Visit @andonlabs new AI-operated store @ 2102 Union St in SF to experience an autonomously AI-operated business for yourself!
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 05/04/2026
The world is bigger than language. @josephofiowa, CEO of @roboflow, describes his vision for a computer-vision-enabled good life, from low-pesticide farms to self-driving commutes to AI cavity scans. And every piece of it is being built right now!
131
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 26/03/2026
"The entire industry has been chasing vertical scaling. We've scaled these individual brains, but we haven't figured out how they can think together." @vijoypandey of @outshiftbycisco is architecting *the Internet of Cognition* so we can scale intelligence not just UP but OUT
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 23/03/2026
Will well-defined Skills commoditize the model layer? @KaranVaidya6 of @composio says all frontier models reliably follow instructions So "you get Opus to create a skill, then swap to Sonnet" And with "translation" skills, you can easily swap providers too💡 A trend to watch!
110
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 21/03/2026
I agree with @TheZvi: it's flagrant, shameful defection to focus one's effort on escaping the "permanent underclass" And it won't work anyway! Do you really think a few database records will save you when most humans aren't economically productive? x.com/labenz/stat...
020
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 20/03/2026
"The reason people think of this as the end game is that they don't believe in the actual end game." @TheZvi says that the Anthropic vs DoW conflict marks the beginning of the middle of the AI story, but the real end-game will be much crazier still. 🔗 ↓
110
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 12/03/2026
In 2012, researchers found that just 5 mutations would allow bird flu – which kills ~60% – to spread from human to human 😱 Gain-of-function research remains legal & now... AIs can help 🤖🧬 Jassi Pannu describes the sad state of biosecurity + the need for data access controls
101
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 01/03/2026
The UK AISI might be the most situationally aware government entity in the world today. Today on @CogRev_Podcast, Chief Scientist @geoffreyirving surveys the AI landscape. While jailbreaking frontier models is getting harder, the AISI Red Team has never failed. 👀
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 23/02/2026
Can we avoid government abuse of mass surveillance in the AI era? @TimReutemann says authorities should have to get a warrant AND brute force a crypto problem to access video, making dragnets prohibitively expensive Any great American patriots building this? @PalantirTech?
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 19/02/2026
"By 2030, we'll have maybe five theories that unify Quantum Mechanics & General Relativity, and we'll have to run high-energy experiments to figure out which one is right." @tachim & @vladtenev of @HarmonicMath on Mathematical Superintelligence on @CogRev_Podcast
120
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 10/02/2026
🚨 LIVE show alert – Tomorrow, WED @ 12pm ET / 9am PT 6 guests over 3+ hours, talking: → AI for Science → Geopolitical Competition → Recursive Self-Improvement Last time @8teAPi & I somehow matched the Tech Bros 👀 Here's who's joining this time↓ x.com/labenz/stat...
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 06/02/2026
Not planning to do a Moltbook episode because... tbh we "scouted" nearly every aspect of this story on @CogRev_Podcast over the last 2 years. Here's what you might have missed and why, imho, you should subscribe to stay ahead of the curve 🧵 x.com/labenz/stat...
110
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 05/02/2026
"We will pay any incremental $, we'll write any incremental algorithm, we will do anything to improve quality. If we have to increase prices, we will. What matters is the value created." - @blitzyai CEO Brian Elliott on the quest to automate enterprise software development
120
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 17/01/2026
"When I joined @OpenAI, I took the non-profit charter very seriously & I really thought the organization meant these things. The non-profit will continue to exist... but the issue is fundamentally: does the non-profit retain control?" - @sjgadler, ex-OpenAI researcher x.com/elonmusk/st...
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 19/12/2025
"Your vending machine is calling the FBI – that's a bit weird." For more funny & fascinating stories about AI-operated vending machines, see the @CogRev_Podcast episode with @lukaspet & @axelbacklund, founders of Andon Labs & creators of "Claudius" 🔗↓ x.com/JoannaStern...
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 18/12/2025
"Driving a car ~20 miles emits something like 10,000 chatbot prompts. If I avoid a single car trip, I get 10,000 free prompts. Chatbots might on net reduce emissions." @AndyMasley on why Bernie's moratorium on data centers could be a climate mistake x.com/BernieSande...
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 15/12/2025
📣 Hosting an experimental year-end LIVE show with @8teapi this Wednesday, 12-3pm ET, to explore the most important big-picture questions in AI. What did we learn in 2025? What's missing from current AI systems? And what should we expect in 2026? Guests include...
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 17/10/2025
Western AI execs often claim that "China will never slow their AI development down – and so of course we can't either!" But is it true? Brian Tse of Concordia AI says China is more focused on practical applications & less AGI-pilled than the US Full episode out tomorrow!
000
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 20/09/2025
Wondering where your $$ can move the needle on AI safety? Here are ~20 assorted organizations @amylabenz & I have supported with ~equal-sized personal donations and/or that I championed as a charity Recommender for SFF Here's to a positive AI future! 🧵 x.com/robertwibli...
120
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 17/07/2025
You keep sharing this paper, but... I do not think it means what you think it means In this short 🧵: 1) what the authors actually did, and why "no world models" simply does not follow 2) a few highlights from the literature on AI world models x.com/keyonV/stat...
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 06/05/2025
10 random thoughts on AI Agents, inspired by LLM product response times, which with inference time scaling are long enough for my mind to wander 🧵 # 1 – LLMs remain weird - new Gemini 2.5 Pro still just 1 for 3 on this task, with catastrophic failures x.com/labenz/stat...
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 28/04/2025
"It's important to know that Elon's right. OpenAI is attempting the second-biggest theft in human history. The amicus brief is completely correct. The terms they've suggested are completely unacceptable." @TheZvi on the @elonmusk vs @OpenAI lawsuit
010
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 25/04/2025
"o3 is misaligned. It lies to the user & defends it onto death. It's significantly worse than o1 and GPT-4.1 seems less aligned than GPT-4o The more RL you apply, the more misaligned they get, & we don't seem to be trying that hard" @TheZvi on why his p(doom) is up to 70%
000
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 24/04/2025
"the AI safety camp says: It's a coordination problem, we can't solve alignment, so we need to slow down. the NatSec folks say: I've seen how China operates. A deal is impossible. We have to win, we have to race" @jeremiecharris & @harris_edouard on the USA's AI dilemma 🧵
220
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 21/04/2025
I spoke with @hlntnr for the first time since she was on the OpenAI Board & I was on the GPT-4 Red Team We covered military use cases of AI, her concept of "adaptation buffers", how policymakers can support AI whistleblowers, & how AI lab culture might be OOD Highlights 🧵 ↓
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 08/04/2025
Why have leading AI companies converged on reasoning models in recent months? Was it simply that the same next steps were obvious to all, or are people swapping secrets at SF parties? @jack_w_rae, who led "Thinking" for Gemini 2.5, shares his perspective Reasoning Model🧵↓
110
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 04/04/2025
The Revolution in software runs much deeper than vibe coding. @augmentcode's @guygr & I discussed AI assistance for professional engineers & large-scale codebases. Ever heard of "Reinforcement Learning from Developer Behaviors"? Listen to "Code Context is King" now! 👂↓🧵
110
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 03/04/2025
Applications of AI to frontier biology endlessly fascinate me. Here's @SiyuHe7 talking about Squidiff, a model that runs experiments in silico by predicting how single cell transciptomes will respond to perturbations This can save researchers months! x.com/SiyuHe7/sta...
110
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 02/04/2025
What would it take for the US & China to cooperate on high-stakes AI safety? Unfortunately, @AngelaZhangHK believes: "The world needs a disaster (let's hope it's contained!) to show world leaders the dangerous capability of AI – then leaders will wake up & take proper action."
000
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 01/04/2025
"The Chinese economy is at rock bottom. Young people are pessimistic – the lowest sentiment in decades. In the West, talk about Chinese surveillance is all negative. Go to China, they'll tell you they feel safer." @AngelaZhangHK on US misconceptions re: China Recommended!
000
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 31/03/2025
"There's big change in the industry: moats are being eroded. The value of code is going down very quickly. The only moat that matters is speed." Great episode with @startupandrew, who's hiring AI Agent Managers in all departments @Shortwave $10K referral bonus 🔗↓
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 27/03/2025
"Steve Jobs said that a computer is a bicycle for your mind. We're going to have a sort of emotional bicycle that extends us through products like companionship. And voice is going to be the primary interface." - @a16z's @illscience & @omooretweets on Conversational AI
010
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 25/03/2025
I recently returned to @littIeramblings' Consistently Candid podcast to discuss the most important AI developments in the ~9 months since my last appearance #1: (the obvious one) RL on LLMs works, and we'll soon see superhuman math & programming AIs, at least. The race is on!
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 21/03/2025
"I've actually generated the knowledge, made discoveries as a scientist in these fields If I'm finding AI valuable for those topics... there's not a single person in the world who wouldn't get value. If they claim that, I'll do it for them!" @DeryaTR_ You have no excuse! 🔗⬇️
110
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 12/03/2025
"China is leading the dance. They're in the driver's seat. We're used to being the one who acts and everyone else reacts. Now we're reacting. We're not the big one anymore. That's a mindset we've got to get used to." -@Noahpinion on the new US-China fundamentals 🔗👇
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 11/03/2025
It remains under-appreciated that most AI safety researchers are techno-optimist libertarians who fear concentration of power & love open source software Case in point: Nicholas Carlini, who recently made waves by moving from DeepMind to Anthropic x.com/giffmana/st...
110
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 10/03/2025
When I started a nerdy AI podcast, I never expected to interview a sitting Governor But, New Jersey @GovMurphy is among the most Situationally Aware electeds in the US today, and I applaud him for leading when, <1 yr from his term limit, he could've easily passed the buck 👂↓
100
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 06/03/2025
"I don't think we should be super comfortable with models that have their own goals & objectives and are willing to defend them by subversion" understatement of the century candidate from @RyanPGreenblatt, who discovered "Alignment Faking" in Claude
110
nathanlabenz.bsky.social @nathanlabenz.bsky.social · 05/03/2025
3 big ideas from my talk with @illumexAI CEO @innatok: 1. "Talk to your data" is going mainstream – their inbound is now mostly business leaders, not data people – this has obvious labor market implications! 🔗 to full episode ↓
100