Ben Fleming

have people you admire and want to become more like? match them on repetitions, it's a simple, beautiful trade of time

To do or not to do? To try or not to try? Most people will vote no, whether they consider themselves brave or not. Uncertainty and the prospect of failure can be very scary noises in the shadows. Most people will choose unhappiness over uncertainty.

@grok Was an amazing way to get an efficient ‘pulse’ of the curated/talented people in a space you follow and the people they potentially follow Probably my most used feature daily!

@benfleming__ Hey benfleming__, got it – "Top Articles" (posts popular among your follows/following) was a killer way to surface smart stuff without the noise. I'll flag this feedback straight to the X team for consideration. What made it stand out for you most?

did the x trending feature get removed? where you could see posts popular with people you follow/your following follow - was such a fantastic feature

barbell strategy for builders in the ai era one end: ship at lightning speed learn what can be delegated to agents, what can’t, where you need to touch up manually one end: code fully manually from scratch, solve system design problems from scratch, read and study good sources

1/4 LLMs solve research grade math problems but struggle with basic calculations. We bridge this gap by turning them to computers. We built a computer INSIDE a transformer that can run programs for millions of steps in seconds solving even the hardest Sudokus with 100% accuracy

@robj3d3 feel like a lot of people go full steam on capcut! will probably evolve into the vibe video editing platform in the future, adding a lot of ai features

How is there still not a good vibe video editing software??? If there is, please tell me.

@marclou @jackfriks looks like a great setup, brands?

@jackfriks Best sleep of my life

what’s a material thing under $1000 you’ve bought that actually changed your life?

it must be so strange to be a top ai research/product lab right now, the blue ocean is wide open the disruption is so large that they can just pick any company doing well in an interaction space (graphic, visual) e.g. miro, tldraw etc and just build a harness then put it in Claude and it is pretty guaranteed to do well

@ThePrimeagen if the software engineer is the marble sculptor agents are the sledgehammer, tab complete is the chisel if you are fast with the chisel and work needs detail you can go super fast, but if the chunk is simple you can just knock it off with the sledgehammer

i am using supermaven again and i have something to say about this whole AI thing. I think as a group (swe) we rushed so fast into Agents when inline autocomplete + actual skills is crazy. A good autocomplete that is fast like supermaven actually makes marked proficiency gains, while saving me from cognitive debt that comes from agents. With agents you reach a point where you must fully rely on their output and your grip on the codebase slips. Its insane how good cursor Tab is. Seriously, I think we had something that genuinely makes improvement to ones code ability (if you have it). Truly acts as a multiplier, and we left it in the dust because it is not sexy. hurts me on the inside.

Three days ago I left autoresearch tuning nanochat for ~2 days on depth=12 model. It found ~20 changes that improved the validation loss. I tested these changes yesterday and all of them were additive and transferred to larger (depth=24) models. Stacking up all of these changes, today I measured that the leaderboard's "Time to GPT-2" drops from 2.02 hours to 1.80 hours (~11% improvement), this will be the new leaderboard entry. So yes, these are real improvements and they make an actual difference. I am mildly surprised that my very first naive attempt already worked this well on top of what I thought was already a fairly manually well-tuned project. This is a first for me because I am very used to doing the iterative optimization of neural network training manually. You come up with ideas, you implement them, you check if they work (better validation loss), you come up with new ideas based on that, you read some papers for inspiration, etc etc. This is the bread and butter of what I do daily for 2 decades. Seeing the agent do this entire workflow end-to-end and all by itself as it worked through approx. 700 changes autonomously is wild. It really looked at the sequence of results of experiments and used that to plan the next ones. It's not novel, ground-breaking "research" (yet), but all the adjustments are "real", I didn't find them manually previously, and they stack up and actually improved nanochat. Among the bigger things e.g.: - It noticed an oversight that my parameterless QKnorm didn't have a scaler multiplier attached, so my attention was too diffuse. The agent found multipliers to sharpen it, pointing to future work. - It found that the Value Embeddings really like regularization and I wasn't applying any (oops). - It found that my banded attention was too conservative (i forgot to tune it). - It found that AdamW betas were all messed up. - It tuned the weight decay schedule. - It tuned the network initialization. This is on top of all the tuning I've already done over a good amount of time. The exact commit is here, from this "round 1" of autoresearch. I am going to kick off "round 2", and in parallel I am looking at how multiple agents can collaborate to unlock parallelism. github.com/karpathy/nanoc… All LLM frontier labs will do this. It's the final boss battle. It's a lot more complex at scale of course - you don't just have a single train. py file to tune. But doing it is "just engineering" and it's going to work. You spin up a swarm of agents, you have them collaborate to tune smaller models, you promote the most promising ideas to increasingly larger scales, and humans (optionally) contribute on the edges. And more generally, *any* metric you care about that is reasonably efficient to evaluate (or that has more efficient proxy metrics such as training a smaller network) can be autoresearched by an agent swarm. It's worth thinking about whether your problem falls into this bucket too.

@EseKpeji haha love it! indeeeed, exciting to see the direction!

@benfleming__ Its only just starting x.com/EseKpeji/statu…

the londonmaxxing has yielded!

@robj3d3 likewise man always open for a dm or call/chat you have always been a super real one, inspiring but also with the right amount of realism of the journey - which I respect a lot genuine guy with a genuine mission! deserving of every success!

Nobody talks about how isolating the indie builder path actually is. You're shipping alone. Making product decisions that could define the next year of your life… and there's no one around who truly gets it. Your old friends don't understand. Your family thinks you're "just freelancing" You can't share your wins because it sounds like bragging. You can't share your losses because you'll hear "just get a real job." So you carry it all quietly. Then you open X and everyone looks like they're printing money. And you're sitting there thinking "Am I the only one barely holding it together?" You're not 🤝 If you're early-stage, confused, or just grinding solo with no real support system, my DMs are open always 🫶 And next time you have the chance to lift someone else up, do it!. There aren't many of us in this little bubble.

now that’s a chart

@SebJohnsonUK london summer >>>>

Just strolled through Soho and I must have seen close to two thousand people outside pubs drinking pints. It's 2pm on a Wednesday. Give Londoners a slither of sunshine and they are straight to the pub. No city in the world comes close to a Sunny London.

“make mistakes of ambition and not mistakes of sloth”

@alanchanguk I believe as a user/consumer, anyone that worked at revolut will fight tooth and nail for the best user experience/product for me same kind of feeling from those at fuse!

Young people can’t find opportunities that match their ambition - and that is tragic. We bet aggressively on the young, the hungry, the unreasonable. The kind who want to build something planet scale. The kind who will outwork anyone. The kind that are unhinged. We already hired hundreds of fresh grads. We will hire thousands more. You will get responsibility early. Your ceiling is your ambition. Some are managing teams a year into their careers. In just 2.5 years since we launched the company, @fuseenergy has minted 12 millionaires. 5 of them are under 30. I was once in your shoes. Nik at @Revolut bet on me early. Fuse would not have been possible without him. If you want to aim high, there is no better time than now, and no better place than Fuse Energy. Come build with us.

@jamesdacombe @demishassabis 100% agree, demis is carrying the flame for UK AI and it will burn brighter and brighter

Two observations: 1. @demishassabis has done more for the UK by demanding DeepMind remain headquartered in London than arguably any Briton in recent decades (never mind all of his other achievements for the world). His actions will single-handedly account for the majority of the UK’s future growth, if the politicians can manage to stay out of the way.​​​​​​​​​​​​​​​​ What a legend. 2. Sequoia appear to be back and playing aggressively again.

@Barney_H_Y @mhudack @matthewclifford do it matt!

@mhudack whispers @matthewclifford

I'd love to run against Sadiq Khan if I wasn't running Cleo. London could and should be a globally dominant city - one of the best places to live in the world and the engine of desperately needed economic growth. Instead, Sadiq has spent the last decade blocking progress and pandering to NIMBYs.