Raveesh Bhalla

For those working on AI/ML/LLM datasets, is there any particular dataset that could use some help with labeling tasks?

Built something that I’d like to test with on an open dataset that would benefit.

Convinced everyone who abandoned TikTok for RedNote should have their internet access permanently taken away for being too dumb.

Ever wanted to host your own Olympics?

olympify.xyz

Project CETI •-- Home CETI is a nonprofit organization applying advanced machine learning and state-of-the-art robotics to listen to and translate the communication of sperm whales.

Some actual projects on this! www.projectceti.org

The more you use Jupyter Notebooks, the more you appreciate how valuable they can be for prototyping.

And, equally, the more you hate the general python ecosystem.

Hahahahahahahahaha

Interesting angle - where I see humility as important is being aware that, regardless of what you accomplish, there are people you can learn from and who can do things better than you can.

Zuck had a good way to frame it saying at any point he knows there are more “great people” not at meta than at

Morning sunrise over the San Francisco Bay

If there’s one thing I’m grateful for this year it’s becoming a morning person and getting to enjoy these views on a daily basis.

Can empathize. My one worry with abandoning Twitter, though, would be the over-moderation here given what happened with @alpindale.bsky.social. Definitely feel like it’s easier to have a quiet community here until some flashpoint.

“Solved” the propaganda issue there by moving back to Following tab.

Yep used that this morning and worked great!

Wow that’s really neat!

The author of this dataset repo (@AlpinDale) has been banned by the @moderation.bsky.app for nonsense reason. For a tech team that aims to rival X but doesn't bother to learn about the basics of training and be reflective about their own data policy, this is totally unacceptable. @howard.fm

3. Shared block lists is a great idea that Twitter should have adopted long ago (pre-Elon)
4. Excited to try the custom feeds – been interested in algorithmic choice for some time. Worried that there isn't enough content though, & Top Articles as a tab on Twitter might be good enough

Few things I noticed as I tried to port over my network:
1. 41% of my (very small) following list has an account there
2. Definitely more active than before! Following tab actually has content (& better than Threads)

Thank you!

Questions from a Bluesky noob:
1. What are some useful lists of people to follow? Topics I like to read about include AI/ML, tech, economics, SF. Also open to other niche ones!
2. How do I use the “block lists” feature? Want to make sure I keep my feed here as healthy as possible

Arrested development meme about how people know No Dead Ends doesn’t work, except for their idea. Hint: it doesn’t

Let’s see if this does better than X - that too with a link!

“No dead ends” is a commonly stated product/design “principle” for Search and Recommendation products.

Unfortunately, it’s also wrong more often than not.

raveesh.substack.com/p/no-dead-en...

Guess I’m giving this a shot? Kinda wish there was a decent “port your follow list” tool built in or highly recommended in the onboarding flow