jake retweeted
every hardware founder to be absolutely drooling thinking about automating ports rn. like bro you do not want to attract the attention of the waterloo interns they will cook you in six months
jake retweeted
I would love an ELO-based restaurant ranking system
Yelp reviews aren’t very useful. Everything is 3.5 - 4.5 stars. Tells me nothing. Same for Google Maps. The 0-5 scale just doesn’t get used properly because of social norms toward high ratings.
What would work much better is…
jake retweeted
Here is the NVIDIA 4090(sm89) instruction Set.
Please share
kuterdinel.com/nv_isa_sm89/
jake retweeted
I hate how research papers are written.
so much yapping in them.
just get to the point bro
jake retweeted
Perhaps the next bill we should pass is one where we can sue Senators and hold them accountable for the damage they do to their own economic futures.
jake retweeted
Replying to @futuristflower
i oscillate between “this is amazing” to “this is adult coco mellon”
jake retweeted
Replying to @Geiger_Capital
"I don't think there's anything confusing there" says someone totally confused🤦♂️
jake retweeted
Devin just got debunked, no LLMs won't be replacing software engineers news.ycombinator.com/item?id…
jake retweeted

Crémieux@cremieuxrecueil6 Mar 2024
United States
United StatesConnected via United States Android AppAccount-level information, not a live location or per-post device.
SF just voted to
- cut down on bribes to public officials
- give police more power while limiting the bureaucratic nonsense they have to deal with
- to screen welfare recipients for drug use, and
- to bring back algebra.
If sanity can win in SF, it can win anywhere.
jake retweeted
I was continuously distressed at Meta over how few people in the VR org would actually talk to customers using the products, or even read the app reviews and user voice comments they had taken the time to write. Bold New Visions were way more popular than Fixing Broken Shit.
jake retweeted
Have you ever done a dense grid search over neural network hyperparameters? Like a *really dense* grid search? It looks like this (!!). Blueish colors correspond to hyperparameters for which training converges, redish colors to hyperparameters for which training diverges.
jake retweeted
Some of the greatest minds in the world are working at OpenAI, Anthropic, Perplexity, etc.
Why is the best interface they’ve come up with for chatting with AI an unsearchable, flat list of conversations?
Why aren’t they mapping the ideas from your conversations?
Why aren’t they providing a search interface trained on your conversations so you can find previous responses?
Why aren’t they intelligently grouping your archive beyond date?
Especially with the increasing commoditization of the underlying technology
jake retweeted
AI isn't catching up to "us", because AI makes us smarter. Any externalized cognitive device we invent (calculators, chess engines, LLMs...) becomes a tool for us to use -- a part of our "greater brain".


























