Prasang Upadhyaya

(1/2) After over 7 years, I resigned from Twitter last Friday. I got to work on many things here, but building the Home Timeline was the most fun I have had in my career. A large part of it was the amazing people I had the good fortune to collaborate with and who made me better.

Incredible case study that's readable yet filled with practical approaches and tactics: ✔ Use GEPA to get near-frontier performance at 1/100th the cost. ✔ Quickly evaluate new models, without having to rewrite prompts. ✔ Evolve & maintain your programs w/o full rewrites

💡 Two arXiv papers published in recent days (one from us, one from TUD) reach the same conclusion: LLMs can now generate C++ code for SQL processing that outperforms classical database systems. ⚙️ Our code generator is based on Claude Code and exploits multiple agents working in parallel. Each agent performs tasks typically associated with different components in a #DBMS, such as workload analysis, query optimization, or physical design tuning. 📊 We compare to various classical #DBMS such as DuckDB, ClickHouse, Umbra, MonetDB, and PostgreSQL, finding that the agent-generated code is often significantly faster. Code generation costs are moderate (<$20), making the approach practical for frequently executed queries. 🤖 Analyzing generated code, we find that agents exploit various optimization techniques, including query-specific data structures, as well as low-level optimizations that are specific to the hardware cache hierarchy of our server. 📃 Paper: arxiv.org/pdf/2603.02081 💾 Code: github.com/SolidLao/GenDB 🌐 Site: solidlao.github.io/GenDB @lojil192574 #LLM #Databases #AI #DB

👇Some cool stuff from my team! We bring search to more modalities and applications with this release!

We just rolled out our Spring ’25 release with some big updates to Dropbox Dash, our AI search and knowledge management product. Here’s what’s new: 📷 Search across video, audio, and images — no more digging for cryptic file names. You can even use Dash to search for people you work with. (1/2)

@RohitSSJain @Twitter 🫡

After an incredible 7-year journey, I have decided to leave @Twitter. Today marks my last day. It has been a privilege to work and learn alongside such amazing colleagues. It was so much fun working on the home timeline, and extremely proud of building the conversations ranking team. Thanks to @arvind__t @mjahr @paraga @xuexb @aandryey @lientm @gungorpl @sagemintblue @satanjeev @prup @xiaohuyan @Zhengxiangyi @ienjoythebeach @rachelxiaoyu @rishabh_misra_ @christine_su @sal_gandhi @tyleraromero @asthanaut and many many others for taking a bet on me during my time here!

@sushnt Congratulations Sushant!

Thrilled and honored to be a part of this year's cohort of Sloan Fellows. Deep gratitude to my collaborators and students without whom the work I'm part of would be impossible, and to my mentors and my family for their unwavering support. 🙏🏽 #SloanFellow

@justinli_bj @thesteggie 😂

@thesteggie the only real 'personalized' boosting

I have never before criticized a competitor by name on the @1Password blog. This is an exception. blog.1password.com/not-in-a-milli…

@xiaohuyan Thank you for all you did to make Home a great team to be a part of. I was lucky to get to work with you!🫡

@whoisxun Is this the one where they accidentally transfer 2x of your price and then ask for x back?

One scammer insisted Zelle but can't speak over the phone.

Trying to sell some furniture online. Facebook marketplace is full of scammers. I am not kidding. More than half reached out are scams. Meta, do your job! Moved to Offer Up. Hopefully a better experience.

(2/2) I leave Twitter with many fond memories and wish the best to the team who will continue taking Twitter to new heights in the challenging times ahead. #lovewhereyouworked 💙

(1/2) After over 7 years, I resigned from Twitter last Friday. I got to work on many things here, but building the Home Timeline was the most fun I have had in my career. A large part of it was the amazing people I had the good fortune to collaborate with and who made me better.

@Mishael40 @nabokov7 They left the base unspecified ;)

@nabokov7 you look well for a 30 year old 😁 businessinsider.com/pro-elon-musk-…

@venusatuluri Amazing run Venu!

My time at Twitter has ended. 🫡💙 10 years 3 months. Many thanks to so many people who've taught me so many things. But it's time for me to try and learn something different. And a big thanks to those of you who'll keep this ship running in the stormy weeks and months ahead.

@Cyan_Dai Very grateful that we got the chance to work together.

🫡

@ghostM Sorry to hear that Mattie.

@i_pascu Really enjoyed working with you! Excited to see what you do next and I wish you the best for it :)

@satanjeev @__lucab Maybe writing SQL inside pandas. Or the other way round ;)

@__lucab Which of these is hardcore?

do ds managers have to write sql or is pandas good enough?

@adityagp @ponderdata @pandas_dev @modin_project @lux_api Congratulations Aditya and team!

So incredibly excited for the official launch of @ponderdata today, taking our research on making @pandas_dev usable at scale with @modin_project and @lux_api to the next level ...

We describe a method for maximising long term engagements by using model-based reinforcement learning. The policy is then used to make decisions about whether to send push notifications or not. More details are in the paper if you are interested. arxiv.org/abs/2202.08812