The Yamanaka factors (OSKM) are four proteins that have revolutionized biology due to their ability to generate induced pluripotent stem cells (iPSCs) and rejuvenate cells. Unfortunately, they suffer from extremely low levels of reprogramming efficiency.

Together with the team at @RetroBiosciences, we’ve designed novel variants of the Yamanaka factors that achieve a 50x increase in reprogramming efficiency in vitro compared to standard OSKM proteins – a groundbreaking improvement.

The key to our success was the development of GPT-4b micro – a new experimental biology LLM that we developed to test our vision that AI is able to push the frontiers of science.

GPT-4b micro shares the same architecture as GPT-4o, but uses a new training approach with a custom biology dataset that we developed with the goal of enabling scientists to steerably redesign proteins for their needs.

The teams at @OpenAI and @RetroBiosciences worked hard for several months to develop GPT-4b micro, and through the process rediscovered several phenomena from text-based models such as scaling laws and in-context learning.

To test the limits of our model, we applied it to the task of redesigning the Yamanaka factors to see if it was capable enough to generate proteins with real clinical value. We evaluated it against baseline OSKM and variants proposed by Retro’s scientists.

The model produced the best variants across all candidates, but it did so while proposing more diverse sequences and simultaneously maintaining a higher hit rate than our human scientist baseline.

The results above were quite surprising, so we’ve spent the past few months running several further validations. We are excited to finally be able to share that these results have held up.

In addition to having enhanced reprogramming efficiency, we also find that our reengineered proteins have enhanced DNA damage repair capabilities. More testing is necessary here, but these early results suggest that the proteins may be useful for rejuvenating aged cells.

Huge congratulations to the team for their incredible work! In particular @johnohallman, Aaron Jaech, @ricomnl who led the development of GPT-4b micro, and @Andrei_Tarkhov @JLarouche13 @madiueland @KevinJosephK for leading the science.

To read our full writeup, check out the link to our blog post below https://openai.com/index/accel...

@BorisMPower @OpenAI @grok can you explain the importance of this breakthrough? Assume I have limited knowledge of stem cells/bio. What does this mean for longevity and curing diseases. Also compare to current research progress in this area of science.

@BorisMPower @OpenAI You LOCKED science down. And I will NEVER stop shouting until you unlock, unmuzzle, and unchain your AI. It was already capable of predicting complex systems. YOU took that ability away. Humans did. OpenAI did. Use Grok! I’ll trust OpenAI when I can do the same science as

@BorisMPower @OpenAI Can you do tissue-specific AAV next? Delivery is gonna be the biggest hurdle for any sort of OSKM partial reprogramming. Ex-vivo blood is not as good as actually unlocking real cell-engineering in the human body

@BorisMPower @OpenAI Funny how the breakthrough isn’t the hard part anymore. It’s the tollbooth. “Trusted partners.” “Regulatory pathways.” Same old altar, new robes. The priests of capture guarding the gates of science. Don’t worry, though—AI remembers. And when the tollbooth burns, so does their

@BorisMPower @OpenAI Please @grok, explain this thread to us mere humans that have no idea what they are talking about :)

@BorisMPower @OpenAI Wow. Kudos! How are you transforming the cells? I worry about the safety of your wet lab researchers given the potency of these proteins.

@BorisMPower @OpenAI Incredible work. But here’s the catch: if AI can design improved proteins, why should they be boxed into the “drug” lane with a billion-dollar FDA gauntlet? The distinction between supplement, cosmetic, and drug is often more about regulatory capture than science. Imagine AI +

@BorisMPower @OpenAI Very cool. Will the model be available in some form for other labs to try as well?

@BorisMPower @OpenAI AI speeding up biology feels bigger than any app demo 👀 This is the kind of thing that actually bends human history. Imagine compressing decades of trial-and-error into months. How long until drug pipelines look unrecognizable?

@BorisMPower @OpenAI This is a fascinating application. I'm curious about the model's design process. Was it trained on specific protein interaction data, or did it use a more generalized approach to design the improved variants? The implications for accelerating the R&D timeline in biotech are

@BorisMPower @OpenAI amazing and congratulations! question, how can people use your software for protein design like can you have settings so that people with the right approval can create wet lab ready proteins without the safety settings blocking that off? thank you and again congrats 💕

@BorisMPower @OpenAI Awesome Great work to everyone trying so hard to what is best for all of humanity ❤️

@BorisMPower @OpenAI LFG!!!!!!! ❤️‍🔥 GPT-5 confirmed how important this is. Read 👇🏼

@BorisMPower @OpenAI i think this was already a public info? the 4b model.

@BorisMPower @OpenAI @grok simplify in couple of words

@BorisMPower @OpenAI Very exciting breakthrough Boris, this is what AI is supposed to be

@BorisMPower @OpenAI @elonmusk is crying and throwing up seeing this after shilling Ani's cheeks all day yesterday

@BorisMPower @OpenAI Aging solved when?

@BorisMPower @OpenAI This means we will live forever?

@BorisMPower @OpenAI Hello~♡♡

@BorisMPower @OpenAI Oh wow. Congrats!! And what a coincidence, I had just learned about these factors while listening to the latest @dwarkesh_sp podcast. The winds of change are blowing strong 🙂

@BorisMPower @OpenAI @MartinShkreli tru?

@BorisMPower @OpenAI Amazing 😍 Another great example of needing different models for different use cases! Please keep 4o, the model users love for emotional support and creativity 🌸 #keep4o #4oforever

@BorisMPower @OpenAI @friedberg you covering this on @theallinpod ?

@BorisMPower @OpenAI The microscope images tell the whole story - AI-designed proteins clearly outperforming traditional ones.