Federico
@federicovaggi.bsky.social
F_vaggi on Twitter. Senior staff scientist at Google X, previously Amazon.
This is a point I desperately wish more people (especially here) would grasp. There are very few low hanging fruits or pareto improvements when it comes to big policy questions, almost all policy decisions involve pissing at least some people off.
Governing means making trade offs and disappointing people. People tend to see the glass half empty rather than being grateful for what they get. So you get a lot of disappointed people.
January 18, 2026 at 6:35 PM
This is a point I desperately wish more people (especially here) would grasp. There are very few low hanging fruits or pareto improvements when it comes to big policy questions, almost all policy decisions involve pissing at least some people off.
Reposted by Federico
I wrote something up for AI people who want to get into bluesky and either couldn't assemble an exciting feed or gave up doomscrolling when their Following feed switched to talking politics 24/7.
The AI Researcher's Guide to a Non-Boring Bluesky Feed | Naomi Saphra
How to migrate to bsky without a boring feed.
nsaphra.net
April 26, 2025 at 1:31 AM
I wrote something up for AI people who want to get into bluesky and either couldn't assemble an exciting feed or gave up doomscrolling when their Following feed switched to talking politics 24/7.
This is a good write up, but, I think it's missing an important angle. The challenge is that people *want* to believe it's true, and that post was perfectly calibrated to appeal to people's priors, in spite of having a lot of obvious "tells" that had nothing to do with AI.
The author of a viral Reddit thread alleging fraud at a food delivery company tried to back up his claim by sending me AI-generated documents. Today I'm publishing those documents in the hopes that it helps other reporter see what we're up against in the age of AI www.platformer.news/fake-uber-ea...
January 7, 2026 at 3:59 PM
This is a good write up, but, I think it's missing an important angle. The challenge is that people *want* to believe it's true, and that post was perfectly calibrated to appeal to people's priors, in spite of having a lot of obvious "tells" that had nothing to do with AI.
Of course it would probably help if I pasted the right url: federicov.github.io/uncertainty-...
January 6, 2026 at 9:51 PM
Of course it would probably help if I pasted the right url: federicov.github.io/uncertainty-...
Reposted by Federico
My first foray into systems biology -- this work was led by Boya Hou, a postdoctoral researcher at UIUC, now on the academic job market. She presented an early version of these results at the @nitmb.bsky.social MathBio Convergence Conference in August of 2025.
January 6, 2026 at 4:04 AM
My first foray into systems biology -- this work was led by Boya Hou, a postdoctoral researcher at UIUC, now on the academic job market. She presented an early version of these results at the @nitmb.bsky.social MathBio Convergence Conference in August of 2025.
Following @nsaphra.bsky.social - will try to do more science posting here as well. Here is a little project I worked on during the Christmas break. Using semantic entropy as a regularizer to calibrate LLM uncertainty: federicov.github.io/uncertainty-...
federicov.github
January 6, 2026 at 3:57 PM
Following @nsaphra.bsky.social - will try to do more science posting here as well. Here is a little project I worked on during the Christmas break. Using semantic entropy as a regularizer to calibrate LLM uncertainty: federicov.github.io/uncertainty-...
Reposted by Federico
I find it hard to square the preciousness about "why would someone ever stay on Twitter?" when everyone on this site knows that the community here intentionally chased off lots of people.
January 5, 2026 at 5:41 PM
I find it hard to square the preciousness about "why would someone ever stay on Twitter?" when everyone on this site knows that the community here intentionally chased off lots of people.
Reposted by Federico
will do the same!
I'm going to start doing this every month to help new users dipping their toes in. If you post about AI research here, like this post. I'll follow you.
(Like even if I already follow you, to help others find you too.)
(Like even if I already follow you, to help others find you too.)
January 5, 2026 at 9:32 PM
will do the same!
Reposted by Federico
What if you could train agents on a 𝗱𝗲𝗰𝗮𝗱𝗲 of driving experience in 𝘂𝗻𝗱𝗲𝗿 𝗮𝗻 𝗵𝗼𝘂𝗿, on a single GPU?
Excited to share 𝙋𝙪𝙛𝙛𝙚𝙧𝘿𝙧𝙞𝙫𝙚 2.0: A fast, friendly driving simulator with RL training via PufferLib at 𝟯𝟬𝟬𝗞 𝘀𝘁𝗲𝗽𝘀/𝘀𝗲𝗰 🐡 + 🚗
youtu.be/LfQ324R-cbE?...
Excited to share 𝙋𝙪𝙛𝙛𝙚𝙧𝘿𝙧𝙞𝙫𝙚 2.0: A fast, friendly driving simulator with RL training via PufferLib at 𝟯𝟬𝟬𝗞 𝘀𝘁𝗲𝗽𝘀/𝘀𝗲𝗰 🐡 + 🚗
youtu.be/LfQ324R-cbE?...
PufferDrive 2.0 release
YouTube video by Daphne Cornelisse
youtu.be
December 30, 2025 at 4:12 PM
What if you could train agents on a 𝗱𝗲𝗰𝗮𝗱𝗲 of driving experience in 𝘂𝗻𝗱𝗲𝗿 𝗮𝗻 𝗵𝗼𝘂𝗿, on a single GPU?
Excited to share 𝙋𝙪𝙛𝙛𝙚𝙧𝘿𝙧𝙞𝙫𝙚 2.0: A fast, friendly driving simulator with RL training via PufferLib at 𝟯𝟬𝟬𝗞 𝘀𝘁𝗲𝗽𝘀/𝘀𝗲𝗰 🐡 + 🚗
youtu.be/LfQ324R-cbE?...
Excited to share 𝙋𝙪𝙛𝙛𝙚𝙧𝘿𝙧𝙞𝙫𝙚 2.0: A fast, friendly driving simulator with RL training via PufferLib at 𝟯𝟬𝟬𝗞 𝘀𝘁𝗲𝗽𝘀/𝘀𝗲𝗰 🐡 + 🚗
youtu.be/LfQ324R-cbE?...
Reposted by Federico
I appreciate the forthright acknowledgement of the mistake from the author, and look forward to decisive action by @princetonupress.bsky.social to correct this error, and any others, that made it past their robust editorial process.
December 29, 2025 at 5:36 PM
I appreciate the forthright acknowledgement of the mistake from the author, and look forward to decisive action by @princetonupress.bsky.social to correct this error, and any others, that made it past their robust editorial process.
I actually think people are earnest when they believe that <claim> is false. The process of justification is something like:
- <claim> is upheld by people I don't like, which is very strong evidence it's false.
- <claim> undermines something I think is important or implies a trade off
- <claim> is upheld by people I don't like, which is very strong evidence it's false.
- <claim> undermines something I think is important or implies a trade off
Person A: <claim>
Person B: Did you honestly say <claim>? You idiot. You moron.
Person A: What is incorrect about <claim>?
Person B: I can't tell you, for Secret Reasons.
***
I don't understand why Person B thinks this is effective rhetoric.
Person B: Did you honestly say <claim>? You idiot. You moron.
Person A: What is incorrect about <claim>?
Person B: I can't tell you, for Secret Reasons.
***
I don't understand why Person B thinks this is effective rhetoric.
December 24, 2025 at 8:33 PM
I actually think people are earnest when they believe that <claim> is false. The process of justification is something like:
- <claim> is upheld by people I don't like, which is very strong evidence it's false.
- <claim> undermines something I think is important or implies a trade off
- <claim> is upheld by people I don't like, which is very strong evidence it's false.
- <claim> undermines something I think is important or implies a trade off
Behold, the king of all academics: you might enjoy this @lastpositivist.bsky.social
December 22, 2025 at 8:19 PM
Behold, the king of all academics: you might enjoy this @lastpositivist.bsky.social
Reposted by Federico
Deleted a post that was going viral and bringing an undesirable amount of attention from anonymous accounts. But, I'm going to be honest, the toxic folks here have won. So much of the usage of the site is about avoiding their attention. I'll just use this site to get book recs going forward
December 21, 2025 at 1:33 AM
Deleted a post that was going viral and bringing an undesirable amount of attention from anonymous accounts. But, I'm going to be honest, the toxic folks here have won. So much of the usage of the site is about avoiding their attention. I'll just use this site to get book recs going forward
Great interview between Ezra Klein and the Ellie Hassenfield - the leader of Givewell:
www.nytimes.com/2025/12/16/o...
Gifting the article here - highly encouraging everyone to donate to them, they are an amazing organization staffed by amazing people.
www.nytimes.com/2025/12/16/o...
Gifting the article here - highly encouraging everyone to donate to them, they are an amazing organization staffed by amazing people.
Opinion | What Does It Mean to Give Well?
www.nytimes.com
December 16, 2025 at 4:11 PM
Great interview between Ezra Klein and the Ellie Hassenfield - the leader of Givewell:
www.nytimes.com/2025/12/16/o...
Gifting the article here - highly encouraging everyone to donate to them, they are an amazing organization staffed by amazing people.
www.nytimes.com/2025/12/16/o...
Gifting the article here - highly encouraging everyone to donate to them, they are an amazing organization staffed by amazing people.
Reposted by Federico
There's a sucker born every minute as the saying goes. People who bet on the president's meme coins being their ticket to riches have been sorely disappointed. $TRUMP is down 92% since its peak and $MELANIA was down 99%.
At this point, you can only blame yourself if you fell for this grift.
At this point, you can only blame yourself if you fell for this grift.
Donald and Melania Trump’s Terrible, Tacky, Seemingly Legal Memecoin Adventure
No one wants to claim credit for helping the first couple launch cryptocurrencies that plummeted more than 90% from their peak.
www.bloomberg.com
December 16, 2025 at 12:49 AM
There's a sucker born every minute as the saying goes. People who bet on the president's meme coins being their ticket to riches have been sorely disappointed. $TRUMP is down 92% since its peak and $MELANIA was down 99%.
At this point, you can only blame yourself if you fell for this grift.
At this point, you can only blame yourself if you fell for this grift.
Reposted by Federico
Extraordinary courage from Ahmed al Ahmed, a Muslim, 43-year-old father of two, who bravely risked his life to save his neighbors celebrating Hanukkah.
Praying for his full & speedy recovery.
And so deeply inspired by his example.
Praying for his full & speedy recovery.
And so deeply inspired by his example.
December 14, 2025 at 2:37 PM
Extraordinary courage from Ahmed al Ahmed, a Muslim, 43-year-old father of two, who bravely risked his life to save his neighbors celebrating Hanukkah.
Praying for his full & speedy recovery.
And so deeply inspired by his example.
Praying for his full & speedy recovery.
And so deeply inspired by his example.
Reposted by Federico
Incredible piece on Oliver Sacks. If you were ever awed at his supposedly true stories (I remember being stunned by the account of the autistic twins who rattled off large prime numbers), read this. He told wonderful stories, but they were in large part fiction.
www.newyorker.com/magazine/202...
www.newyorker.com/magazine/202...
Oliver Sacks Put Himself Into His Case Studies. What Was the Cost?
The scientist was famous for linking healing with storytelling. Sometimes that meant reshaping patients’ reality.
www.newyorker.com
December 12, 2025 at 10:33 PM
Incredible piece on Oliver Sacks. If you were ever awed at his supposedly true stories (I remember being stunned by the account of the autistic twins who rattled off large prime numbers), read this. He told wonderful stories, but they were in large part fiction.
www.newyorker.com/magazine/202...
www.newyorker.com/magazine/202...
Reposted by Federico
I have purchased the PACER documents so everyone can read the impact statements submitted by victims of Do Kwon and the 2022 Terra/Luna meltdown. Collapses like these ruin lives.
www.courtlistener.com/docket/67087...
www.courtlistener.com/docket/67087...
![described UST as a stablecoin that would always hold its $1 peg, and those protocols as safe, reliable way to earn 19–40% yield. It didn’t feel like gambling; it felt like the answer we had been praying for. We sold everything at discounted prices: the house, the business, the cars. We scraped together just over $190,000 — every dollar of seventeen years of sacrifice — and in late April 2022 I converted it all to UST and locked it into Anchor and Mirror, trusting the promises that had been repeated for months. Just two weeks later, on May 9, 2022, I watched UST fall off its peg. In panic, I almost sold everything to limit the damage, but then I saw Mr. Kwon’s tweet: “Deploying more capital— steady lads.”. The price started recovering that day. I trusted him again and decided to keep the UST locked. What followed were two weeks of pure terror. The Terra blockchain halted. Mirror Protocol liquidity vanished. I couldn’t withdraw a single cent. When the system finally let me out on May 23, UST was trading below $0.07. Our $190,000 had become less than $13,000. Seventeen years of our lives — gone in two weeks. People say you should never risk money you can’t afford to lose. But this never felt like risk. The founder of the project had called it “too big to fail,” boasted that the Luna Foundation Guard would protect every dollar. To a desperate father trying to feed two families in a foreign country, it sounded like the safest investment ever. The money was only the beginning of the catastrophe. Less than two weeks after the collapse, my wife packed her bags. She could no longer trust that I could keep her and her mother financially safe. A few months later the divorce papers arrived. Our teenage sons chose to stay with me and my parents, to hold what was left of the family together. College was never mentioned again. Both boys quit school and now work as car mechanics to put food on the table. Today I am [redacted]. I live in [redacted] with my elderly parents, taking wh…](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:exrxvyu6bpoym6mbnctke5tn/bafkreidoc2b4cylajtyz2bnt74nc6q5egitftutriae3wi6lsgopvtxa7q@jpeg)
December 11, 2025 at 10:16 PM
I have purchased the PACER documents so everyone can read the impact statements submitted by victims of Do Kwon and the 2022 Terra/Luna meltdown. Collapses like these ruin lives.
www.courtlistener.com/docket/67087...
www.courtlistener.com/docket/67087...
Reposted by Federico
Can vision transformers learn without images?🤔👀
Our latest work shows that pretraining ViTs on procedural symbolic data (eg sequences of balanced parentheses) makes subsequent standard training (eg on ImageNet) more data efficient! How is this possible?! ⬇️🧵
Our latest work shows that pretraining ViTs on procedural symbolic data (eg sequences of balanced parentheses) makes subsequent standard training (eg on ImageNet) more data efficient! How is this possible?! ⬇️🧵
December 10, 2025 at 11:06 AM
Can vision transformers learn without images?🤔👀
Our latest work shows that pretraining ViTs on procedural symbolic data (eg sequences of balanced parentheses) makes subsequent standard training (eg on ImageNet) more data efficient! How is this possible?! ⬇️🧵
Our latest work shows that pretraining ViTs on procedural symbolic data (eg sequences of balanced parentheses) makes subsequent standard training (eg on ImageNet) more data efficient! How is this possible?! ⬇️🧵
Reposted by Federico
Listening to Ballmer reflect on running Microsoft. He highlighted key mistakes:
1. Hubris in trying to compete everywhere.
2. Spreading too thin without the talent.
3. Failing to decouple new efforts from core.
4. Ignoring Wall Street investors.
Mistakes a lot of tech companies still make today
1. Hubris in trying to compete everywhere.
2. Spreading too thin without the talent.
3. Failing to decouple new efforts from core.
4. Ignoring Wall Street investors.
Mistakes a lot of tech companies still make today
The Steve Ballmer Interview | Acquired Podcast
The complete Steve Ballmer interview (and transcript!) on Microsoft’s history and strategy
www.acquired.fm
December 11, 2025 at 4:24 AM
Listening to Ballmer reflect on running Microsoft. He highlighted key mistakes:
1. Hubris in trying to compete everywhere.
2. Spreading too thin without the talent.
3. Failing to decouple new efforts from core.
4. Ignoring Wall Street investors.
Mistakes a lot of tech companies still make today
1. Hubris in trying to compete everywhere.
2. Spreading too thin without the talent.
3. Failing to decouple new efforts from core.
4. Ignoring Wall Street investors.
Mistakes a lot of tech companies still make today
Reposted by Federico
I find this sort of messaging so much better than insisting that the USA is somehow poor in aggregate.
In the richest city in the richest country on the planet, the thing we can’t afford is to forget those who are left hungry.
November 28, 2025 at 9:04 PM
I find this sort of messaging so much better than insisting that the USA is somehow poor in aggregate.
Reposted by Federico
Mentioning the amount of grant money you have won in a way that makes the department chair happy without coming across too blatantly as bragging that you piss off the other professors in audience.
November 21, 2025 at 12:25 AM
Mentioning the amount of grant money you have won in a way that makes the department chair happy without coming across too blatantly as bragging that you piss off the other professors in audience.
Reposted by Federico
Basically saying "if you do crime for me, I won't only pardon you--I'll give you millions of dollars of public money." Truly unprecedented level of corruption
www.bloomberg.com/news/article...
www.bloomberg.com/news/article...
Michael Flynn, DOJ in Settlement Talks Over $50 Million Claim
The Justice Department has been discussing settlements with two former officials from Donald Trump’s first term who — like the president — claim they’re owed major payouts from the US government as vi...
www.bloomberg.com
November 14, 2025 at 7:04 PM
Basically saying "if you do crime for me, I won't only pardon you--I'll give you millions of dollars of public money." Truly unprecedented level of corruption
www.bloomberg.com/news/article...
www.bloomberg.com/news/article...
I fully understand the causal point, but, if in the real world, candidate quality ends up being very tightly correlated with ideology, then, filtering by ideology is completely viable. This is quite obviously what's going on - there is a very simple DAG too.
Yes, these are legitimate **candidate quality issues** for Dems. But they are not about ideology! If your evidence is “they’re bad candidates,” that’s not an ideology critique — and actually the very thing I have been saying people should pay attention to more than ideology!
November 12, 2025 at 4:27 PM
I fully understand the causal point, but, if in the real world, candidate quality ends up being very tightly correlated with ideology, then, filtering by ideology is completely viable. This is quite obviously what's going on - there is a very simple DAG too.