Michael Fester retweeted
last night at @ycombinator paper club: AI for Biology
mountain view, packed room, and it was @FrancoisChauba1's last one!!!
we heard from @yusufroohani on Stanford's Virtual Cell, @garykbrixi on Evo2, @divnori on genomic in-context learning, and the @Omantahealth team on their Francois for making paper club what it is. thank you YC for continuing to be a home for weird, ambitious people trying to do impossible things.
LETS GO!!!!!
mountain view, packed room, and it was @FrancoisChauba1's last one!!!
we heard from @yusufroohani on Stanford's Virtual Cell, @garykbrixi on Evo2, @divnori on genomic in-context learning, and the @Omantahealth team on their Francois for making paper club what it is. thank you YC for continuing to be a home for weird, ambitious people trying to do impossible things.
LETS GO!!!!!
0
2
4
224
For the first time in my career, I'm working on a problem I would happily spend the rest of my life trying to solve.
Typically, there's some kind of goal in mind when you start something, commonly an exit or IPO!
If I spent the next 40 years working on cures for autoimmune diseases (such as type 1 diabetes, multiple sclerosis, rheumatoid arthritis), I would be perfectly happy.
We will get there. These are extraordinarily difficult problems, and may take a while, but what else would I do with my life?
It was always "do this, so we can work on that, then we can work on cures", some means to an end goal.
And finally, we said, let's just get to it, now.
Typically, there's some kind of goal in mind when you start something, commonly an exit or IPO!
If I spent the next 40 years working on cures for autoimmune diseases (such as type 1 diabetes, multiple sclerosis, rheumatoid arthritis), I would be perfectly happy.
We will get there. These are extraordinarily difficult problems, and may take a while, but what else would I do with my life?
It was always "do this, so we can work on that, then we can work on cures", some means to an end goal.
And finally, we said, let's just get to it, now.
1
0
10
551
Salesforce is officially cooked.
Instead of paying $25/month for a CRM, I spent 90 hours and $6K in tokens building one that automatically moves every deal to closed-won.
Instead of paying $25/month for a CRM, I spent 90 hours and $6K in tokens building one that automatically moves every deal to closed-won.
1
0
1
174
We have confirmed that, last night at approximately 1:23 AM, a rogue swarm of unaligned agents operating outside established safety protocols seized control of a karaoke machine at a San Francisco dive bar and performed "Mr. Brightside" three consecutive times.
2
0
1
164
So excited for this!!!
Introducing Expand Fetch, a new way to scrape and crawl the web.
Your agent has never seen the web. No wonder it gets basic questions wrong.
Traditional Markdown APIs lose critical details. So we trained two System one models to change that:
Iris, a vision model that sees erving tables, icons, and layouts.
X-Ray, a JSON model that uncovers the hidden data behind every page.
Every fetch also comes with a Wayback Machine. Every line links back to its source, so you can verify exactly what your agent saw.
Three products. One API.
On our benchmark of 1,094 questions across 326 challenging web pages, agents using Expand got 74% right. The next best API got 55%.
Available today at $1 per 1,000 fetches.
Try it at expand.ai
0
0
8
743
I don’t want to grow old with type 1 diabetes, so we’re starting now to work on a cure.
For ten years, Marie and I have dreamed about working on this. But we couldn’t see what we could uniquely bring to the problem.
Now we can.
Recent breakthroughs in AI-driven mathematics were the final push, including yesterday’s announcement of a solution to the Navier–Stokes problem. Seeing AI tackle long-standing mathematical problems has made us more hopeful about what we could contribute to research on gene and cell therapies.
We know how to build systems of AI agents that work in parallel, test ideas, ditch what doesn’t work, double down on what does, and keep going.
We’ve had hundreds of agents working on questions around producing beta cells, keeping them alive, and protecting them from the immune system. We’ve published the work so far.
The holy grail is a "world biology model" of type 1 diabetes that we can experiment on virtually. Once that model is reliable, we could run thousands of experiments in parallel and take the most promising ideas into the lab. We’d still need biological validation, but we could learn much faster than the timelines of mice and wet labs currently allow.
We might be five years away from that, or we might be months away. I don't now yet. But we’ll be there!
For ten years, Marie and I have dreamed about working on this. But we couldn’t see what we could uniquely bring to the problem.
Now we can.
Recent breakthroughs in AI-driven mathematics were the final push, including yesterday’s announcement of a solution to the Navier–Stokes problem. Seeing AI tackle long-standing mathematical problems has made us more hopeful about what we could contribute to research on gene and cell therapies.
We know how to build systems of AI agents that work in parallel, test ideas, ditch what doesn’t work, double down on what does, and keep going.
We’ve had hundreds of agents working on questions around producing beta cells, keeping them alive, and protecting them from the immune system. We’ve published the work so far.
The holy grail is a "world biology model" of type 1 diabetes that we can experiment on virtually. Once that model is reliable, we could run thousands of experiments in parallel and take the most promising ideas into the lab. We’d still need biological validation, but we could learn much faster than the timelines of mice and wet labs currently allow.
We might be five years away from that, or we might be months away. I don't now yet. But we’ll be there!
4
2
24
2,885
Michael Fester retweeted
Thank you to @forbes for including me and Michael in a piece on why founders with vision outlast founders chasing money!!!!
It's humbling being there next to people building global safety nets and electrifying Africa.
Michael and I had already built B2B SaaS companies, and er one, but instead we started building for a problem we'd discussed for ten years, and one very personal to us.
The more you care about what you do, the bigger the impact you get to have. I really believe that!!
link below
It's humbling being there next to people building global safety nets and electrifying Africa.
Michael and I had already built B2B SaaS companies, and er one, but instead we started building for a problem we'd discussed for ten years, and one very personal to us.
The more you care about what you do, the bigger the impact you get to have. I really believe that!!
link below
2
1
9
272
Michael Fester retweeted
An internal version of Astra, @OpenAI’s next major model family, solved 10 major open problems in mathematics, quantum complexity, and theoretical computer science.
We believe it will be a major step for scientific reasoning. openai.com/index/ten-adva…
We believe it will be a major step for scientific reasoning. openai.com/index/ten-adva…
10 proofs from our next major model Astra on long-standing open problems in mathematics and theoretical computer science (also including new circuit lower bounds for computing the permanent!)
GPT-5.6 has already enabled so much exciting work in math and science. Can’t wait to see what comes next!
827
2,508
17,411
13,376,045
So glad we can finally share this now that things have settled.
Joshua showed up on Saturday night, Dec 20 at 7pm, right as the great SF power outage hit.
For his work trial. This was our first in-person interaction.
There was no wifi so we couldn't do any work.
Ok, let's t every restaurant was closed. No power. No Toast iPads. No food.
We wandered until we found a neighborhood with wifi just so we could call an Uber.
Ubered to a hotel downtown that actually had power. (Shoutout CitizenM.)
Ate sushi. Worked in the hotel until 10pm when the power came back. Then back to the crib.
Definitely not a traditional work trial.
But the most eventful moments in life usually mean something bigger. I don't believe in coincidences.
Josh rolled with every curveball that night. No complaints and great vibes.
That told us everything we needed to know.
Josh killed the trial and officially started with us last week.
Welcome aboard, Joshua. We're so happy to have you.
Joshua showed up on Saturday night, Dec 20 at 7pm, right as the great SF power outage hit.
For his work trial. This was our first in-person interaction.
There was no wifi so we couldn't do any work.
Ok, let's t every restaurant was closed. No power. No Toast iPads. No food.
We wandered until we found a neighborhood with wifi just so we could call an Uber.
Ubered to a hotel downtown that actually had power. (Shoutout CitizenM.)
Ate sushi. Worked in the hotel until 10pm when the power came back. Then back to the crib.
Definitely not a traditional work trial.
But the most eventful moments in life usually mean something bigger. I don't believe in coincidences.
Josh rolled with every curveball that night. No complaints and great vibes.
That told us everything we needed to know.
Josh killed the trial and officially started with us last week.
Welcome aboard, Joshua. We're so happy to have you.
1
0
9
697
The @ycombinator hiring board might be the most underrated part of the entire program.
We posted a role and had 200+ applicants in under 48 hours. And the quality is insane compared to every other job board we've used.
People who apply through YC are already plugged into the is knowing what "Y Combinator" is.
They're not looking for corporate hours, they're hungry to build something.
LinkedIn is 80% spam and 15% agencies.
This might be worth the application alone.
We posted a role and had 200+ applicants in under 48 hours. And the quality is insane compared to every other job board we've used.
People who apply through YC are already plugged into the is knowing what "Y Combinator" is.
They're not looking for corporate hours, they're hungry to build something.
LinkedIn is 80% spam and 15% agencies.
This might be worth the application alone.
9
12
267
34,758
We hire AI engineers to answer our customers' support tickets until they automate them.
Most companies hire the cheapest labor possible to answer support tickets forever.
Brilliant engineers hate repetitive work, and have agency to do something about it. They'll answer "where mes before they can't stand it anymore and automate it.
The people improving the AI are the same people who live in the trenches answering tickets.
Hire people too restless to answer the same ticket twice.
Most companies hire the cheapest labor possible to answer support tickets forever.
Brilliant engineers hate repetitive work, and have agency to do something about it. They'll answer "where mes before they can't stand it anymore and automate it.
The people improving the AI are the same people who live in the trenches answering tickets.
Hire people too restless to answer the same ticket twice.
1
1
7
497
One of the best parts of being a founder is doing grunt work at the ground level.
When we onboard a new customer, I become their chatbot support agent for a few days.
I'm in there answering tickets manually, learning their business, understanding how their customers actually his to train our model, but I'm always surprised at how many high-level takeaways I end up with after being boots on the ground.
So if you see my face around on the chatbot, don't ask anything weird 😉
When we onboard a new customer, I become their chatbot support agent for a few days.
I'm in there answering tickets manually, learning their business, understanding how their customers actually his to train our model, but I'm always surprised at how many high-level takeaways I end up with after being boots on the ground.
So if you see my face around on the chatbot, don't ask anything weird 😉
0
0
5
331
Black Friday is the Super Bowl of customer service. We onboarded a new consumer brand during it. Behind the scenes in Nashville.
0
0
5
350
Most common Thanksgiving support tickets we saw this year:
"Someone brought store-bought rolls acting like they cooked"
“Why does every dish have raisins in it?"
“This turkey is dry.”
“Cousin walk group is giggling at the Jell-O.”
"Football volume too loud" / "Football put me next to the cousin who breathes like a leaf blower?”
"Mother-in-law asked when we're having kids again" - Recurring issue.
"Someone's kid is showing me YouTube videos"
"Dad asleep on couch by 4pm"
Closing all tickets. See everyone in 364 days
"Someone brought store-bought rolls acting like they cooked"
“Why does every dish have raisins in it?"
“This turkey is dry.”
“Cousin walk group is giggling at the Jell-O.”
"Football volume too loud" / "Football put me next to the cousin who breathes like a leaf blower?”
"Mother-in-law asked when we're having kids again" - Recurring issue.
"Someone's kid is showing me YouTube videos"
"Dad asleep on couch by 4pm"
Closing all tickets. See everyone in 364 days
0
0
5
373
Tomorrow is Thanksgiving, so today felt like the right time to say thank you.
Only in SF would a random night dinner turn into a deep dive on functional programming while eating ravioli .
This is what we're most grateful for! The community that's formed around what we're ke time out of their week to share ideas, challenge our thinking, and geek out over things that excite us.
So to everyone: thank you.
This wouldn't be nearly as fun (or delicious) without you.
Happy Thanksgiving. 🦃
Only in SF would a random night dinner turn into a deep dive on functional programming while eating ravioli .
This is what we're most grateful for! The community that's formed around what we're ke time out of their week to share ideas, challenge our thinking, and geek out over things that excite us.
So to everyone: thank you.
This wouldn't be nearly as fun (or delicious) without you.
Happy Thanksgiving. 🦃
5
3
25
2,412
Excited to announce a game-changing employee benefit:
We now offer free hitchhiking to work.
Any employee can now simply go outside, stick their thumb out, and someone will probably pick them up, completely free, and drop them off somewhere near the office.
Early results are oyees arrive within 2 hours of their intended start time, and we're seeing incredible cross-pollination as team members network with long-haul truckers and doordash drivers.
Talent is our greatest asset.
We now offer free hitchhiking to work.
Any employee can now simply go outside, stick their thumb out, and someone will probably pick them up, completely free, and drop them off somewhere near the office.
Early results are oyees arrive within 2 hours of their intended start time, and we're seeing incredible cross-pollination as team members network with long-haul truckers and doordash drivers.
Talent is our greatest asset.
0
0
11
998