COMO ASSIM BICHO
O modelo de IA da PREFEITURA DO RIO bateu os modelos da Qwen e da Deepseek nos benchmarks e virou um dos melhores modelos de IA de código aberto ????????
This is actually a pretty cool idea having your own cultural curator for your Insta or Tiktok feed.
You can one up it: One could have different personas scrolling through different feeds. I would argue that's more for marketers though.
I send codex on doom scroll quests to instagram on my behalf everyday. I love discovering hidden gems from influencers, but I hate this new trend where every reel is:
"comment LINK and I'll dm you my list." So I built a Codex skill to watch bookmarked reels for me, and turn them into field notes for:
👗 clothes I should buy
📚 books I should read
🍜 places I should visit
It identifies the brands, products, restaurants, and locations mentioned in the reel, finds the original product pages, pulls the actual product photos from the source website, and organizes everything into structured notes with links and any useful creator comments.
Instead of doomscrolling for 45 mins and forgetting everything, I have delegated brain rot to codex!
I really liked Eric's take on why alpha go is profound: A 10-layer network can only do 10 sequential steps of thinking, by construction. And yet those 10 steps can "amortize and approximate to very high fidelity a nearly intractable search problem."
SF Native here- the whole post is about SF tech. Go to the redwoods. Go to Santa Cruz. Make friends with dog walkers, doctors, bartenders. Get involved in the food culture (and the after parties). Go see some live music. Stare out at the bay while eating seafood in mission bay. Have a picnic with friends in Golden Gate Park. Ride a boat in stow lake. There’s so much more to the city than the tech climb and grind. You can be in tech without having it swallow you whole.
Devs need to get outside of the tech bubble and go solve real problems for non-tech people. We are an overserved market. Go ask your brother-in-law how his software is at his non-tech job and watch the wheels start turning.
@rcwhiteley@Object_Zero_ Yep, traditional image recognition will do the job for most of those use cases.
And at that point is dealing with edge cases that will take most of the dev time.
New fastest shortest-path algorithm in 41 years!
Tsinghua researchers broke Dijkstra’s 1984 “sorting barrier,” achieving O(m log^(2/3) n) time. This means faster route planning, less traffic, cheaper deliveries, and more efficient networks - and a CS curriculum revamp =)
Announcing fasttransform: a Python lib that makes data transformations reversible/extensible. No more writing inverse functions to see what your model sees. Debug pipelines by actually looking at your data.
Built on multi-dispatch. Work w/ @R_Dimm
https://t.co/OGDrBFhnfP
I got Cursor to build the same thing using my annual review and journal notes from Evernote.
Thanks for the idea! @danshipper
I adjusted the prompt for more introspection.
my weekend project:
deep background—a public, static site full of all of my notes, quotes, sources, and references. I feed the URL to deep research when I ask it a question and it uses ONLY my sources to write answers.
it's really fucking cool
Thanks to AI: Ozempic without side effects
A newly discovered peptide from Stanford Medicine combats obesity in a similar way to Ozempic, but without its side effects.
The naturally occurring 12-amino acid peptide acts specifically in the hypothalamus and reduced food intake by up to 50% in animal experiments. Obese mice lost 3 grams of mainly adipose tissue in 14 days.
AI algorithms were crucial to the discovery. Researcher Katrin Svensson founded a company for clinical studies on humans.
How can web browsers render something like this at 60 FPS but if I add a few rows to an HTML table the page freezes for half a second while it repaints?