Erik Kristian Sverre Uri
@eksuri_
Azure OpenAI, CS @ University of Washington
You might like
![cesarepm's profile picture. Curious fan of the universe, eternal optimist, geek & family man doing his best(ish) to be a decent person.
[Nintendo Treehouse. Tweets are my own.]](https://pbs.twimg.com/profile_images/1269519726138417154/CaPXTDEX_x96.jpg)
We built Things, Notion, Todoist... And this person said "nah, txt file is fine" Unironically brilliant. jeffhuang.com/productivity_t…
It's not officially summer here until after the Fourth of July. 😎🐾
Larry Page and Sergey Brin tried to sell Google for under $1m so they could return to their PhDs. Paul Allen had to convince Bill Gates to drop out of Harvard. Bill Hewlett and Dave Packard took day jobs until pushed by a professor to start HP. Steve Wozniak wanted HP to build…
The idea of convincing people to become entrepreneurs is ridiculous. If you need to be convinced, you shouldn’t do it. Starting a company is totally irrational—you need to be obsessed and committed to a painful, decade-long grind. Most people are better off as employees.
Many talented people ontop of myself are putting everything we have into making this happen within ~18months 🙏 We have the team and are going to manifest the compute
Happy 100 years to the Montlake Bridge! Here's to another hundred of serving as the "gateway to the University of Washington". 🥳 🎉
The University of Washington is ranked #⃣8⃣ in the world and #⃣2⃣ among U.S. public universities by U.S. News & World Report’s 2025 Best Global Universities list just released this week. #GoHuskies | lnkd.in/giiku4_v
Few months ago researchers from Meta empirically found out that LayerNorm acts similar to a TanH, squeezing in the weights that are too high into more tractable values They tried to replace the LayerNorm with a TanH, and achieved similar results at higher speed What??? [1/8]
![gabriberton's tweet image. Few months ago researchers from Meta empirically found out that LayerNorm acts similar to a TanH, squeezing in the weights that are too high into more tractable values
They tried to replace the LayerNorm with a TanH, and achieved similar results at higher speed
What??? [1/8]](https://pbs.twimg.com/media/GuFBIv4WwAA2640.png)
![gabriberton's tweet image. Few months ago researchers from Meta empirically found out that LayerNorm acts similar to a TanH, squeezing in the weights that are too high into more tractable values
They tried to replace the LayerNorm with a TanH, and achieved similar results at higher speed
What??? [1/8]](https://pbs.twimg.com/media/GuFBOH0WcAAa5yL.jpg)
Introducing Codestral Embed, the new state-of-the-art embedding model for code.
Unfamiliar with CSS or HTML? @its_just_chen shares a feature you have to try in @code, straight from #MSBuild. @msdev
I think the Palace of Fine Arts in San Francisco, built in a Classical style in 1915, is proof that authenticity doesn’t really matter at all in buildings. If it looks good you can just build a Greek temple in the middle of a West Coast American city and people will love it.
Beauty endures. When wielded in the service of scale, the effect is a humbling grandiosity making us feel small, not vulnerable, but in awe. This is an immensely difficult feat to pull off, but when achieved, is close to a religious experience
Working at Ai2 is probably the closest thing to this btw Tweeted this over a year ago while I was still at Apple, cool to see I kinda figured it out
Is there any way at all I can get paid $200k/yr to go back to school. Pls don’t laugh
Waymo crashed in SF. Police questioning the self-driving car. Surreal. (Wasn't waymo's fault. Actually tried to protect the passengers.)
32 year old Yann Lecun shows off the world's first CNN in 1993 for text recognition.
OK this is good
This looks really good and pretty surprised to see a college class this ~up to date with frontier stuff
Want to learn the engineering details of building state-of-the-art Large Language Models (LLMs)? Not finding much info in @OpenAI’s non-technical reports? @percyliang and @tatsu_hashimoto are here to help with CS336: Language Modeling from Scratch, now rolling out to YouTube.
Sam Altman on how to arbitrage your time
The Daft Punk score for TRON: LEGACY might be my all-time favorite x.com/FreddsCorner/s…
Apple releasing research on all sorts of scaling laws is NOT what I expected in 2025. So insightful Lets fucking go 🍎
Scaling Laws for Native Multimodal Models - Early-fusion exhibits stronger perf at lower param counts, is more efficient to train, and is easier to deploy, compared w/ late fusion. - Incorporating MoEs allows for models that learn modality-specific weights, significantly…
United States Trends
- 1. Alabama 71,7 B posts
- 2. Indiana 83,7 B posts
- 3. Indiana 83,7 B posts
- 4. Mendoza 24,3 B posts
- 5. DeBoer 7.008 posts
- 6. Ty Simpson 5.839 posts
- 7. Curt Cignetti 6.378 posts
- 8. Saban 13,6 B posts
- 9. #iufb 6.746 posts
- 10. #RollTide 6.507 posts
- 11. Oregon 58,5 B posts
- 12. Notre Dame 24,7 B posts
- 13. Austin Mack 1.304 posts
- 14. #CFPplayoff 1.535 posts
- 15. #NeverDaunted 4.720 posts
- 16. Grubb 3.235 posts
- 17. The SEC 54,8 B posts
- 18. Tuscaloosa N/A
- 19. Georgia 63,2 B posts
- 20. Mamdani 312 B posts
You might like
-
Poroburu
@poroburu -
Nikola Barjaktarevic
@Coach_Nikola -
Marooncat23
@marooncat23 -
Dan
@musicman662 -
Adi Seredinschi
@AdiSeredinschi -
Paul Cagnina
@CagninaPaul -
freshies.ch
@freshiesCrew -
HealX PT
@HealXPT -
k kless ward
@ayasgirls3101 -
Fabo
@fabo_web -
CJ in CO
@cjonesCO -
![cesarepm's profile picture. Curious fan of the universe, eternal optimist, geek & family man doing his best(ish) to be a decent person.
[Nintendo Treehouse. Tweets are my own.]](https://pbs.twimg.com/profile_images/1269519726138417154/CaPXTDEX.jpg)
César | 🇲🇽+🇺🇲
@cesarepm -
sal
@salFreeApps -
bps
@brittanyschmelz -
Dave Farley
@DaveFarley1986
Something went wrong.
Something went wrong.