co-founder @mirendil | prev. @xAI, @X, @Twitter

Joined August 2020
Alongside my co-founders, proud to formally announce @mirendil. Our mission is to accelerate science and technology through democratizing frontier AI R&D. If you share our excitement for this mission, reach out to us!
Today, I’m excited to formally announce @mirendil with my amazing co-founders Harsh Mehta, Shayan Salehian, and Tara Rezaei! We’re fortunate to work with @a16z and @kleinerperkins, who led our seed round of $200M, followed by a major investment from NVIDIA, among others. Mirendil exists to accelerate science and technology, and through them, to help solve humanity's most pressing problems. Self-accelerating AI R&D is the most direct path to delivering on AI's broader promise, which is why we believe the most important application of AI is AI itself. Get this loop right, and it compounds. It fundamentally changes the rate of progress itself across all domains. We believe this capability should be democratized. It should be used to power all scientific efforts trying to innovate at the frontier. There are far more important problems—and broader ones—than any single lab can take on, so more groups should be able to pursue them. This pulls concentration of power away from a few labs: businesses and science labs can own their AI and infrastructure, keep their margins, and control their own destiny instead of ceding it all to a single AI lab. We’re a small team with a singular focus. Our founding team consists of 20 researchers and engineers from frontier institutions including Anthropic, xAI, Google DeepMind, and OpenAI, united by a passion for science and a drive to build the technologies that move it faster. If you want to build the system that builds systems, join us! @HarshMeh1a, @shayan_, @tararezaeikh
34
12
1
199
21,866
Shayan retweeted
I've worked on pretraining LLMs for a long time and before agents -- living in a world with this kind of leverage feels incredible. The best part is, we plan to release the same tools that let us do this, so more people can do it too. You can just do things :)
Today, we're sharing a major milestone in our mission to democratize frontier AI development: Pretraining at scale without a pretraining team. We set out to answer an ambitious question: Can we significantly reduce the time, human expertise, and compute needed to pretrain a competitive foundation model from scratch? The result is Mirendil-Base, our first pretrained coding model, built by one researcher in just 3.5 months with less than $10M in compute, including all research, experimentation, and the final training run. Our approach is already showing promising results as we scale to even larger models. This is an important step toward a future where a single researcher, empowered by AI, can accomplish what once required an entire research organization. As these capabilities improve, we believe frontier AI development can become accessible to far more people and teams. More to come soon, including independent evaluations, post-training benchmarks, and API access.
8
8
117
7,391
Shayan retweeted
Today, we're sharing a major milestone in our mission to democratize frontier AI development: Pretraining at scale without a pretraining team. We set out to answer an ambitious question: Can we significantly reduce the time, human expertise, and compute needed to pretrain a competitive foundation model from scratch? The result is Mirendil-Base, our first pretrained coding model, built by one researcher in just 3.5 months with less than $10M in compute, including all research, experimentation, and the final training run. Our approach is already showing promising results as we scale to even larger models. This is an important step toward a future where a single researcher, empowered by AI, can accomplish what once required an entire research organization. As these capabilities improve, we believe frontier AI development can become accessible to far more people and teams. More to come soon, including independent evaluations, post-training benchmarks, and API access.
9
19
14
259
52,970
How frontier AI gets built is changing rapidly and drastically!
We formally announced Mirendil less than 4 months ago. Today, I'm excited to share an important milestone for: Pretraining at scale without a pretraining team. Mirendil's mission is to solve Recursive Self-Improvement (RSI) and use it to democratize frontier AI development, ultimately accelerating science and technology. At Mirendil, when we think about RSI, we think holistically about the company as an intelligent entity that self-improves toward its goal. From that perspective, the most powerful notion of acceleration is reducing the time and resources (number of human experts, compute, and other costs) needed to achieve ambitious goals. Given that pretraining is perhaps the most important yet most challenging part of frontier AI development, we decided to run an ambitious controlled experiment: Starting fully from scratch, can we significantly reduce the time, number of human experts, and compute resources needed to go all the way to successfully pretraining a strong coding model at scale? Mirendil-Base is the result of this experiment. Our first pretrained model was built by one researcher in 3.5 months, with less than $10M in compute. This includes all the research and small-scale experiments it took to get there, as well as the final training run. The recipe holds up as the models get bigger. For reference, Mirendil-Base is in the 100B-500B parameter range, Mirendil-Base-Large in the 500B-2T range, and Mirendil-Base-XL in the 2T-10T range. We're beginning to see a world where a single researcher, empowered by AI, can take on work that once required an entire research organization. As these systems improve, the bottleneck shifts from the size of a team to the ambition and judgment of the people directing them. This has profound implications for who gets to build AI and, ultimately, who gets to push the frontiers of science and technology forward. We’ll share independent evaluations, post-training benchmarks and API-access sooner than you think so stay tuned!
6
2
32
1,885
Pretraining a strong model used to need a whole team. At Mirendil, it took one researcher. The next frontier labs may be very small.
We formally announced Mirendil less than 4 months ago. Today, I'm excited to share an important milestone for: Pretraining at scale without a pretraining team. Mirendil's mission is to solve Recursive Self-Improvement (RSI) and use it to democratize frontier AI development, ultimately accelerating science and technology. At Mirendil, when we think about RSI, we think holistically about the company as an intelligent entity that self-improves toward its goal. From that perspective, the most powerful notion of acceleration is reducing the time and resources (number of human experts, compute, and other costs) needed to achieve ambitious goals. Given that pretraining is perhaps the most important yet most challenging part of frontier AI development, we decided to run an ambitious controlled experiment: Starting fully from scratch, can we significantly reduce the time, number of human experts, and compute resources needed to go all the way to successfully pretraining a strong coding model at scale? Mirendil-Base is the result of this experiment. Our first pretrained model was built by one researcher in 3.5 months, with less than $10M in compute. This includes all the research and small-scale experiments it took to get there, as well as the final training run. The recipe holds up as the models get bigger. For reference, Mirendil-Base is in the 100B-500B parameter range, Mirendil-Base-Large in the 500B-2T range, and Mirendil-Base-XL in the 2T-10T range. We're beginning to see a world where a single researcher, empowered by AI, can take on work that once required an entire research organization. As these systems improve, the bottleneck shifts from the size of a team to the ambition and judgment of the people directing them. This has profound implications for who gets to build AI and, ultimately, who gets to push the frontiers of science and technology forward. We’ll share independent evaluations, post-training benchmarks and API-access sooner than you think so stay tuned!
4
1
1
45
2,712
Shayan retweeted
Turns out the future does have flying cars
48,108
100,215
32,693
834,503
78,605,074
Today, we're announcing a multi-year partnership with @Google to power the next generation of self-accelerating AI! Mirendil will scale its training, inference, and AI research workloads on @googlecloud 's AI Hypercomputer — drawing on both TPUs and full-stack NVIDIA AI Infra.
1
66
7,984
Shayan retweeted
We're thrilled to lead Mirendil's $200M seed round. Frontier AI work has been locked inside a few big labs. @mirendil is building a system that can help anyone do AI work: they train frontier models that are expert at AI R&D and build the product around it. The result is a system that loops over research and engineering problems on its own, making progress without human intervention. It's like a coding agent built for AI research that controls its own GPUs. Mirendil is one of the few teams with the experience and the priors to make the end-to-end system work. @bneyshabur, @HarshMeh1a, @shayan_, and @tararezaeikh come out of Anthropic, Google DeepMind, and xAI. Welcome to the a16z family, Mirendil. By @BornsteinMatt and @MaikaThoughts
Today, I’m excited to formally announce @mirendil with my amazing co-founders Harsh Mehta, Shayan Salehian, and Tara Rezaei! We’re fortunate to work with @a16z and @kleinerperkins, who led our seed round of $200M, followed by a major investment from NVIDIA, among others. Mirendil exists to accelerate science and technology, and through them, to help solve humanity's most pressing problems. Self-accelerating AI R&D is the most direct path to delivering on AI's broader promise, which is why we believe the most important application of AI is AI itself. Get this loop right, and it compounds. It fundamentally changes the rate of progress itself across all domains. We believe this capability should be democratized. It should be used to power all scientific efforts trying to innovate at the frontier. There are far more important problems—and broader ones—than any single lab can take on, so more groups should be able to pursue them. This pulls concentration of power away from a few labs: businesses and science labs can own their AI and infrastructure, keep their margins, and control their own destiny instead of ceding it all to a single AI lab. We’re a small team with a singular focus. Our founding team consists of 20 researchers and engineers from frontier institutions including Anthropic, xAI, Google DeepMind, and OpenAI, united by a passion for science and a drive to build the technologies that move it faster. If you want to build the system that builds systems, join us! @HarshMeh1a, @shayan_, @tararezaeikh
113
121
189
1,468
1,216,901
It doesn’t need to be this way. It should not be this way. It should not be OpenAI/Anthropic deciding which fields get accelerated. To truly accelerate science and technology, self-accelerating AI R&D must be democratized. In fact, this is the exact motivation behind Mirendil.
academics are unprepared for the coming world where much scientific progress is majorly a function of inference compute. whether OpenAI points the Eye of Stargate at your particular field will decide its acceleration. talent will leach away into the labs. it's already begun
11
18
1
263
42,023
We are building a frontier lab to ensure scientific progress is no longer bottlenecked by the speed of AI R&D. Our first memo is live at: mirendil.com. Join us!
We have been heads down but wanted to share a bit about what we are doing 🧵
4
6
151
26,486
Career update: I left xAI to start something new, closing my 7+ year chapter working at Twitter, X, and xAI with so much gratitude. xAI is truly an extraordinary place. The team is incredibly hardcore and talented, shipping at a pace that shouldn’t be possible. From the Home Timeline at X to Grok 2, 3, and 4 at xAI, I worked across product infra and model behavior post-training, from memory to coding infra, agents, and more. Pure startup mode, every day. Working closely with Elon across X and xAI, I saw what happens when you refuse to accept impossible as an answer. I learned to embody obsessive attention to detail, maniacal urgency, and to think from first principles. I’m deeply grateful to @elonmusk for the experience, to @wanghaofei for the trust and support throughout Twitter/X, and to @TheGregYang and @ibab for believing in me. And to the many incredible people I had the privilege to work with along the way, thank you! Now, I’m excited to take the leap and build something new, focused on accelerating science. More soon.
128
71
16
2,226
262,870
Career Update: I've left Anthropic to start something new. Anthropic is a magical place — amazing people, strong culture, and unmatched taste. I have a lot of respect for my friends and ex-colleagues, and knowing them, I'm confident they'll do the right thing, especially when the choices are hard. I feel grateful and proud to have been part of the journey with them. Excited to start a new chapter — stay tuned for more!
73
31
14
1,715
297,712
I've left Anthropic to start something new. 🧵
150
59
20
2,854
420,419
From trusted source in Tehran: Tell all of your friends [abroad], everyone you know: there is absolutely nothing else we can do here inside Iran. Nothing. They are killing people in such ways, they’ve descended upon people so brutally, they're attacking us in such ways... We’ve lost so many lives that no one dares go out anymore. They shoot directly with bullets. They kill outright. And even after killing, they come and behead you, and do countless other violent things to you. For God’s sake, whatever can be done, you are the ones who can do it. You must not abandon these gatherings. All of you must keep these protests going. We cannot do anything inside Iran. Going out into the streets is literally suicide. It’s not about bravery anymore. It’s madness. You go out and they shoot you point blank. They don’t even ask why you came. They just kill you. There is absolutely no way for us to gather unless we had weapons, unless we were armed like them. Otherwise they have weapons everywhere. There has been so much killing. So much. So whatever is to happen now is up to you. You are the ones who can somehow save us. [Breaks down into a sob. Crying, the caller continues...] When you see what’s happening, when you hear what’s happening… I’ve done so much yelling and choked on so much teargas, my voice is bad. We follow what you're doing. When we see you gather in protest in front of an embassy it brings us joy. It still gives us a lingering hope that we can do something. It's your responsibility now. We did absolutely everything we could. As our fellow Iranians it's your responsibility. You can influence foreign policy. You can tell the world what's happening. The placards you hold. The YouTube videos. Distributing images and video. Gathering in front of embassies. Even if you have just the weekends. Please do something for us on Saturdays and Sundays. Please don't let your gatherings get smaller as the weeks go on. No one will stop you there. No one will kill you. No one is singling you out. We don't even have the freedom to walk about in our own town. We no longer even know who is friend or foe. We don’t even know who we’re speaking to anymore. We no longer know if those recounting what's happening is an enemy. Because if one wrong word slips from our mouths, they turn us into a corpse even before we've had a chance to return home. For God’s sake, I beg of you, don’t abandon these gatherings. Our only hope now is you. What’s now obvious is that America won’t do anything. You’re our only hope. Please tell your friends. You’ve seen how many young people have been killed. So many, so many young people. For the sake of the youth. I know that based on the videos that have reached Iran International [satellite station] since yesterday, now that the internet has come back briefly, they will probably cut the internet again and people will go into silence, the country will go back into a blackout. For God’s sake, please help us. Help us save our country from these people. Stand together so we can live in a flourishing country. For God’s sake, help us. We are in Tehran. What I saw in Tehran, and what they’re saying about the provinces and small towns... They couldn’t fully control Tehran’s gatherings with the forces they had, but in the small towns it was killing after killing. Towns with populations of only 20 or 30 thousand people truly had many deaths. Thousands, hundreds… The things we’ve heard, the things we’ve seen go far beyond mass killing. It was horrific. Truly horrific. I beg you. I beg you. Don’t abandon these gatherings. Please. #Iran #IranMassacre
1,038
5,653
572
13,380
3,214,625
Shayan retweeted
After 100 hours of total internet blackout, Iran carried out one of the worst massacres of modern times.💔 Over 20,000 killed in 4 days. Hospitals attacked. Doctors prosecuted for saving lives. 800 people now on death row. #R2PforIran #IranMassacre
725
6,334
383
10,498
173,022
I’m devastated. A good friend and dormmate of mine, Amin Pourfarhang, has been sentenced to death after protesting in Iran 😭 He’s one of the kindest people I’ve ever known. Please share. Visibility raises the cost for the Islamic Republic and may stop this execution!
1,480
30,937
733
56,173
1,051,466
If you have a moment, please sign the petition. Your support means a lot. change.org/p/stop-the-execut…
44
262
10
553
22,474
Shayan retweeted
⚠️ Confirmed: Live metrics show #Iran is now in the midst of a nationwide internet blackout; the incident follows a series of escalating digital censorship measures targeting protests across the country and hinders the public's right to communicate at a critical moment 📉
410
2,218
475
4,136
967,019