It makes zero sense that we already make it so difficult for foreign students, who we spend tax dollars educating, to stay in the US after graduation. And this would make things significantly worse. Just one stupid brain drain acceleration after another.

Douglas Natelson@nanoscaleviews.bsky.social · 6d ago

🧪⚛️ The administration is considering charging someone (graduates? employers? unclear) $100K for OPT, the part of the international student visa that allows graduates to work for some time after completing their degrees. www.wsj.com/politics/pol...

Bluesky handles 3 minute videos, so this is a drumming account now, sorry Midnight at McGuffy's from Dance Gavin Dance

Exactly the type of old school data science project that I love, with the help of new school tools www.iandmacomber.com/blog/gopro-g...

Teaching LLMs Taste: How I Built an Automated Cycling Ride Recap with GoPro, Garmin, and Gemini — Ian Macomber

Or as my wife says: How I managed to turn 3 hours of boring GoPro footage into 30 seconds of boring GoPro footage

iandmacomber.com

Shouldn’t this be free just like car parking?

Streetsblog NYC@nyc.streetsblog.org · 3w ago

The end of lugging your bike up the stairs to your fifth-floor walkup is near 😤 @nyc-dot.bsky.social is bringing 500 secure bike parking lockers to New York City's streets by the end of 2027! And there's still time to request one for your block 👇 youtube.com/shorts/v8PBb...

I guess I should invest in an air purifier

A piece of advice I've given over and over to my fellow data people is: The better you are at the engineering part, the more effective you'll be over all. The person who can figure out how the system generating the data is working is going to beat the pants off the person who can only do SQL.

aly@aly.codes · 3w ago

I always forget how bad data people are at code sometimes (sorry to any data people reading this)

It’s oddly vindicating(?) that our Agentic Age rewards those who invested in software engineering best practices. Tests, CI/CD, platform APIs, infrastructure as code. They’re all very well-aligned with agent utility, which is maybe not surprising; these are all done in the name of automation

Thinking back to that time that Thrice wrote a 9/11 truther song youtu.be/FWHRTKnk7LY

Broken Lungs

YouTube video by Thrice - Topic

youtu.be

Is @citibikenyc.bsky.social on summer vacation?

🧪⚛️ New reporting about unprecedented mid-year cuts at NSF confirms bad news and raises many questions. nanoscale.blogspot.com/2026/07/bad-...

Bad to worse at NSF? (July 2026 edition)

When I wrote this post last month , I really did not want it to be first in a series. The inciting incident to write that post was a news ar...

nanoscale.blogspot.com

Transformers are bottlenecking transformers

Irene De la Torre Arenas@inari-ta.bsky.social · 3w ago

NEW: The century-old device choking the AI push Transformers are fundamental to the flow of electricity and are essential to powering our modern lives. But the largest and most powerful transformers are facing bottlenecks, driving up prices and putting the relentless race for AI expansion at risk.

Tomorrow is the deadline for submitting a comment to stop the administration from dismantling U.S. science. There are nearly 280,000 public comments filed so far! But! Do not assume they are all, or even most, against the proposal. This is a key pillar of Project 2029 - wind down federal R&D.

Costa Samaras@costasamaras.com · 4w ago

Here’s a great resource to help write a comment this weekend before the deadline, by @colesci.bsky.social and @cdelawalla.bsky.social: www.standupforscience.foundation/policy-and-a...

Yes, there are about 250 million cars and SUVs in the U.S. that nearly all run on gasoline and they make about 1,000 million metric tons of carbon dioxide each year and they stay on the road for 12-20+ years.

Hypervisible @hypervisible.blacksky.app · 4w ago

“Cornell researchers found that at the current rate of AI growth, the burgeoning industry could represent 24 to 44 million metric tons of carbon dioxide emissions by 2030, the equivalent of adding five to ten million cars to US roadways.”

It’s been 10 years since this post and nothing’s changed 😭 www.ethanrosenthal.com/2016/07/20/l...

Eugene Vinitsky 🍒@eugenevinitsky.bsky.social · 4w ago

Tools from operations research solve so many problems and the fact they’re not super widely deployed really tells you that many bottlenecks are not technical

Tools from operations research solve so many problems and the fact they’re not super widely deployed really tells you that many bottlenecks are not technical

Slides for our ICML tutorial on Memorization and Generalization of Diffusion and Flow Matching Models are now available ! 🌀 memorization-generalization.github.io @quentinbertrand.bsky.social

Very happy to announce our new release of MVEB our new benchmark in MTEB for evaluating video embeddings! Whether you want to evaluate embeddings for clustering video, matching video with text, or using them within a retrieval system, this benchmark should cover you.

Slides for an introductory talk on IA for maths. speakerdeck.com/gpeyre/ia-fo...

IA for theory

Slide for an introduction talk on IA for maths.

speakerdeck.com

1/2 minute precision 🤌

$21 for a hot dog at JFK is downright offensive

About to try my hands DIY fixing our Toniebox with a soldering iron... we'll see how this goes

Look at that beautiful status line

Alex Gude @alexgude.com · last mo.

It's great!

TIL about /statusline in claude code code.claude.com/docs/en/stat...

Customize your status line - Claude Code Docs

Configure a custom status bar to monitor context window usage, costs, and git status in Claude Code

code.claude.com

Looking for a software engineer to help maintain your OS repo? I currently have availability for a part-time open-source engagement. Preferably, focused on data, evaluations, NLP/ML/LLM. I work as a consultant through my company OxyKodit. DM me if you'd like to explore working together!

🧪⚛️ What is going on within NSF? What is an X-Lab? More on today's news of cuts. nanoscale.blogspot.com/2026/06/bad-...

Bad to worse at NSF? (June 2026 edition)

This is a US research ecosystem post.  Feel free to skip if this isn't your cup of tea. I'm showing my internet age in thinking that the rig...

nanoscale.blogspot.com

When all you have are trillions of parameters, apparently every problem is a “wedge”

👉👈

I think it’s under appreciated how many interesting, diverse, and deep technical problems there are when training and serving frontier models. People think it’s all just pytorch and indexing errors, but there is soooo much other stuff going on across the entire stack.

Vicki@vickiboykis.com · 2mo ago

the actual most revolutionary thing the ai labs have actually done is make it look like it's insanely easy to duct tape together python microservices with all their concurrency issues and trained models with all their transformers issues and have them run as one api call

the actual most revolutionary thing the ai labs have actually done is make it look like it's insanely easy to duct tape together python microservices with all their concurrency issues and trained models with all their transformers issues and have them run as one api call

Kimi’s got the salient bit

austin@aparker.io · 2mo ago

fun? I am in there. intheweights.com