Boris Power
@BorisMPower
Followers
45K
Following
12K
Media
190
Statuses
3K
Head of Applied Research @OpenAI
San Francisco, CA
Joined July 2017
At @OpenAI, we believe that AI can accelerate science and drug discovery. An exciting example is our work with @RetroBiosciences, where a custom model designed improved variants of the Nobel-prize winning Yamanaka proteins. Today we published a closer look at the breakthrough. ⬇️
146
627
3K
RT @bryan_johnson: This is big. OpenAI and Retro used a custom model to make cellular reprogramming into stem cells ~50× better, faster, an….
0
408
0
Blazing-fast image creation – using just your voice. Try Grok Imagine.
278
545
3K
Deception games are a very good benchmark indeed!.
🐺 Introducing the Werewolf Benchmark, an AI test for social reasoning under pressure. Can models lead, bluff, and resist manipulation in live, adversarial play?. 👉 We made 7 of the strongest LLMs, both open-source and closed-source, play 210 full games of Werewolf. Below is
4
1
52
This is one of the most fun and impactful teams at OpenAI - please consider joining !.
our team at openai is hiring technical staff to build frontier evals for finance. If you're passionate about measuring real-world capabilities, have a love/hate relationship with Excel, or are an ex-banker/ex-investor with technical skills, please reach out!.
10
0
137
The data centers should be split into computation (gpu predominantly) and traditional (cpu and storage) data centers.
Data center spending will pass office construction for first time in history.
7
1
53
A lot of common sense moves in SF local policy changes - congratulations!.
It shouldn’t be controversial to say this: The city should not fine residents for parking in their own driveways. As part of my PermitSF initiative, we're taking common-sense steps and getting rid of these unnecessary and burdensome regulations. It’s your driveway—park in it if
2
1
11
We extracted DNA from a 72,000-year-old fossil and created the first living dire wolf in 10,000 years. Romulus is thriving at 6 months old. Follow to track his growth and see the next species we bring back from extinction.
6
7
33
This is incredibly useful - I love the new emergence of image prompting !.
Pose control with nano-banana is shockingly good. Upload characters + a sketch of a pose and ask to combine. Prompt: "take the anime man and woman in the first image and put them in the poses of the stick man in blue and stick woman in red. erase the stick figures"
8
1
127
Taxes have a potential to be used for pro social purposes, such as minimizing the tragedy of the commons. Unfortunately, the most common taxes are placed on economic activity, reducing labor, business activity and exchange of goods and services.
They say all taxes are bad, but some taxes are less bad. Land value taxes are actively good, since they encourage building and discourage lots of abuse, like unnecessarily jacking up rents. Consider Denmark. A higher LVT was associated with lower single-family home prices:
7
18
125
It’s getting to the point that the Pro model performs more accurately than some contractors that take weeks to come back, and aren’t able to justify their numbers.
Absolutely insane use case for ChatGPT 5 Pro: "help me figure out how much sunlight hits {home address}.". Thought for nearly 17 minutes (total across 3 queries), used multiple sources including Project Sunroof, photos from Zillow, historic averages of sunlight in the area, and
12
6
269
Can it take short balls? Spins? It really doesn’t look like it does, but I’d love to be proven wrong!.
🏓🤖 Our humanoid robot can now rally over 100 consecutive shots against a human in real table tennis — fully autonomous, sub-second reaction, human-like strikes.
6
2
31
RT @joannejang: today feels surreal. on the same day i was included in the time100 ai list, we shared internally that i’m transitioning fro….
0
58
0
Excited to see all the new voice interfaces to different apps !.
Introducing gpt-realtime — our best speech-to-speech model for developers, and updates to the Realtime API.
7
1
70
The open source models are amazing too! This should radically decrease cost of high quality healthcare!.
"GPT-5 moves from human-comparable to above human-expert performance". GPT-5 outperforms licensed human experts by 25-30% and achieves SOTA results on the US medical licensing exam and the MedQA benchmark. I sound like a broken record, but AI models are better than most doctors.
11
7
114
Such an exciting update!.
We’re releasing new Codex features to make it a more effective coding collaborator:. - A new IDE extension.- Easily move tasks between the cloud and your local environment.- Code reviews in GitHub.- Revamped Codex CLI. Powered by GPT-5 and available through your ChatGPT plan.
3
4
86
Finally!.
you can now access the full set of @openai guides and API reference docs as a single plain text file - great for feeding into any model or adding to Cursor. here's the url route:
6
2
71
Wow, this is so exciting!.
Since nano banana has gemini's world knowledge, you can just upload screenshots of the real world and ask it to annotate stuff for you. "you are a location-based AR experience generator. highlight [point of interest] in this image and annotate relevant information about it."
6
6
182
Why is the center pointing to exactly Slovenia?!. Is that a general property of Mercator projection? I thought the center was meant to be the null island.
You have to spend the rest of your life in one quadrant. Which one do you choose?
6
0
16
How do you like to stay cool during the summer? I love taking a dip in a creek.
0
0
61
RT @BorisMPower: To read our full writeup, check out the link to our blog post below.
openai.com
OpenAI and Retro Biosciences achieve 50x increase in expressing stem cell reprogramming markers.
0
35
0
Rarely does the science just work on the first attempt. Here it did, thanks to the amazing team!.
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.
4
2
92