Stephan Ledain

Old Twitter was funnier. bsky.app/profile/sledai…

@elonmusk lol no. no it won't.

Tesla will be one of the companies to make AGI and probably the first to make it in humanoid/atom-shaping form

@naval @JTLonsdale This is an obviously lazy take, but more so reads as if you've never tried to model a large amount of behavioural data. There are consistent, replicable patterns even where our capacity to theorize lags.

@JTLonsdale “Mental illness” is to illness as “social science” is to science.

This reply below is simply wrong, and an idiot aspect of our current justice system. Yes, I believe in second chances. But I don’t care what’s wrong with them: if they’ve committed a second violent crime, don’t let them out to hurt our friends and family again. Not for decades.

@karpathy This is brilliant, @karpathy. Grateful for your generosity in sharing insights. I wonder if a useful guardrail for an LLM council is to let go of deep analysis and just “trust its gut” (limit the time in analytical reasoning) to avoid overthinking.

As a fun Saturday vibe code project and following up on this tweet earlier, I hacked up an **llm-council** web app. It looks exactly like ChatGPT except each user query is 1) dispatched to multiple models on your council using OpenRouter, e.g. currently: "openai/gpt-5.1", "google/gemini-3-pro-preview", "anthropic/claude-sonnet-4.5", "x-ai/grok-4", Then 2) all models get to see each other's (anonymized) responses and they review and rank them, and then 3) a "Chairman LLM" gets all of that as context and produces the final response. It's interesting to see the results from multiple models side by side on the same query, and even more amusingly, to read through their evaluation and ranking of each other's responses. Quite often, the models are surprisingly willing to select another LLM's response as superior to their own, making this an interesting model evaluation strategy more generally. For example, reading book chapters together with my LLM Council today, the models consistently praise GPT 5.1 as the best and most insightful model, and consistently select Claude as the worst model, with the other models floating in between. But I'm not 100% convinced this aligns with my own qualitative assessment. For example, qualitatively I find GPT 5.1 a little too wordy and sprawled and Gemini 3 a bit more condensed and processed. Claude is too terse in this domain. That said, there's probably a whole design space of the data flow of your LLM council. The construction of LLM ensembles seems under-explored. I pushed the vibe coded app to github.com/karpathy/llm-c… if others would like to play. ty nano banana pro for fun header image for the repo

LLMs are not the bridge to enterprise AI adoption. They're helpful at times, not transformational. We need genuinely emergent applications that are rooted in new ways of working to realize the AI valuation narrative.

@Heyits_Gaurav @bindureddy Or further exacerbates the anxiety... because they'll be left with the weight of a self-determined life without the support of a work-related institution. There's evidence to suggest that for some, this could cause more harm.

@sledain @bindureddy Exactly. Most people underestimate how much of their “purpose” is borrowed from their profession. Take that away, and we’ll realize UBI doesn’t solve existential anxiety it just removes the excuse for not facing it.

AGI + UBI Is A Powerful Combination - AGI will take care of all the heavy lifting - UBI will provide basic resources to all humanity Humans can choose to pursue whatever they desire, including working on their passions. It will take a decade to get there, but we need a President who is determined to implemet UBI ASAP

@Heyits_Gaurav @bindureddy The data supports this take more than any others

@bindureddy Everyone says AGI + UBI will “let people follow their passion.” But 90% don’t know what that passion is because work gave them identity. When robots take the grind, humans will face the mirror. That’s scarier than unemployment.

Yes and that is by design. No one wants to be anything like LA. This is the gift you've given to us.

This is one of the most beautiful and perfect game 7s ever.

@foundmyfitness @PeterAttiaMD It's really cool how much you're drawing attention to the large body of creatine literature. I'd love to see more human trials explore adverse effects. Anecdotally, I've felt pain in the kidney area from taking more than 5g.

From my recent appearance on The Drive with @PeterAttiaMD! Full episode here. youtube.com/watch?v=VhbpSc…

Supplementing with protein during strength training can increase gains in muscle size by approximately 27%, even when baseline protein intake is already considered adequate. The current RDA for protein, 0.8 grams per kilogram of body weight per day, is meant to prevent deficiency, not to optimize muscle growth or performance. Increasing intake to 1.5x the RDA leads to greater improvements in both muscle size and strength when combined with resistance exercise. Doubling the RDA is even better. Meta-analyses suggest that around 1.6 g/kg/day of protein is sufficient to maximize gains in muscle strength and fat-free mass, with potential (though smaller) benefits observed at higher intakes up to 2.2 g/kg/day for those aiming to fully optimize muscle growth and recovery (like athletes). This highlights how conservative the RDA truly is and why higher protein intakes are important, particularly for individuals engaging in regular strength training or other forms of physical activity.

Here's what I think the future looks like written by Harry Shaan AI x Creativity: Stephan Ledain and the Probability of Wonder linkedin.com/pulse/ai-x-cre…

@emollick This also highlights the safety risks of capabilities outpacing alignment and maybe counterintuitively, a pause to focus on alignment may be more lucrative for the big research houses.

When and if AI development plateaus (and no indication that is happening yet), it may actually accelerate AI integration into our lives, because then it becomes easier to figure out what products & services are needed to complement AI. Right now capabilities are changing too fast

on @OpenAI GPT-5: irrespective of whether you consider GPT-5 better or worse than prior models (and beyond some of the technical failings which i’m sure will get fixed at some point) a lot of the pushback i’m seeing is along the lines of the fact that it is *different* than what people are accustomed to in other words, people have spent the past 1+ years deeply integrating LLMs into their lives to such a degree that they learned how to work with them- including an understanding of their strengths and weaknesses, and how you need to handle them to get the most out of them when the models change significantly in how they engage with you in a new release, it disrupts that experience. it’s like getting a new coworker. it doesn’t *feel* right anymore the future of these models has to be some kind of “personas” which you can control, so that engagement is highly tailored to your preferences, and the logic gets upgraded on the back end with subsequent models while the engagement style with you remains the same i don’t know when we get there, but that is my takeaway from the GPT-5 situation so far

People are literally more anxious and chaotic in this era

@RealDoctorMike @EricTopol @PeterAttiaMD Can we talk about the trust and resistance obstacles standing in the way of effective AI adoption in medicine? Especially in the era of scientific doubt.

Strong (and valid) criticism from @EricTopol on @PeterAttiaMD on longevity supplements, full-body MRIs, and more. Full convo on The Checkup Podcast live now 🎥🙏🏼

In what universe does a 60 million dollar opening for an original studio movie warrant this headline?

@aliciakatz We stand on guard for thee.

can i say something without everyone getting mad?

@NaomiSharon What are the Tariffs on these, the best of Toronto's exports?

good times at Rick

@cathsimard_ Sorry you're going through this. It probably should be a multi-pronged approach. There's intense evidence for CBT, SSRIs (for a limited duration), breathwork and meditation, and most importantly, as you're doing here, community.

Any tips to treat chronic anxiety? Only way way for me to get better and find some peace is complete social isolation but this isn’t sustainable nor healthy. I live a healthy lifestyle (exercice, therapy and eat well) but it doesn’t help.

Lmao Ramit responding to rebuttal in this thread is generational

A model for when things happen to us. What? So What? Now What?