←
Ashaya Sharma1,112 posts
Ashaya Sharma avatar

Ashaya Sharma

@ashaya

Me according to North: 🇨🇦🐝🇳🇵🏒🤖📈⚽💵💻♟️ | Co-Founded @honeycomb_app | MTS @cohere rn | @yaletownpodcast | @VoltSafe

◷ Joined March 2018▧ 157 Photos and videos
Ashaya Sharma avatar
Ashaya Sharma@ashaya·3h

🔥🔥🔥

We put 16 AI models in charge of buying for a produce wholesaler, every morning for 90 days. Only 3 made more money than a simple reorder formula. The models noticed their mistakes and wrote them down, but then they mostly kept making them.

0:50
Ashaya Sharma avatar
Ashaya Sharma@ashaya·Nov 17

Stationary bike riding for rehab has been a big part of my treatment for ACL rehab and now I can just talk to my agents while doing it. - sent while on bike

Ashaya Sharma avatar
Ashaya Sharma@ashaya·Nov 16

We might be sleeping on Apple becoming publicly more infra pilled and k8s pilled. their devices can already do serious ai inference and act as terrific cpus. this needs to be studied. imagine apple as a sandbox provider for remote execution or inference. if they apply this properly to their own products the potential is huge. if they start offering it more broadly to other companies 👀

Post attachment
SpaceX avatar
⟳ Ashaya Sharma retweeted
SpaceX@SpaceX·Nov 14

Mechazilla catches Starship! Congratulations to the entire SpaceX team on this incredible achievement!

1:33
Ashaya Sharma avatar
Ashaya Sharma@ashaya·Nov 14

Vancouver one

0:22
toki@tokifyiNov 13

fable 5.2 is insane at visual stuff one-shotted this w a single prompt: "draw san francisco at golden hour in pure javascript. canvas only, no images, no svg, no libraries. include the bridge, skyline, fog, reflections, and a cable car."

Ashaya Sharma avatar
Ashaya Sharma@ashaya·Nov 13

Some internal NVDA slack channel is about to move the markets hundreds of billions in either direction

Ashaya Sharma avatar
Ashaya Sharma@ashaya·Nov 7

This is beautiful

i got gpt-2 running purely in css! all 124m parameters inferencing at 15 sec/tok, with zero lines of js. i also used a custom font as a BPE tokenizer, and variable fonts for matmuls :)

0:19
donald avatar
⟳ Ashaya Sharma retweeted
donald@donaldjewkes·Nov 3

Introducing CandyBench: a new frontier eval in applied robotics we got 7 robots to give out 3057 halloween candies and graded them on performance here's what happened:

1:56
Rikiya Takehi avatar
⟳ Ashaya Sharma retweeted
Rikiya Takehi@rikiyatakehi·Nov 2

We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: rikiyat.github.io/blog/weigh…

Post attachment
Ashaya Sharma avatar
Ashaya Sharma@ashaya·Oct 31

Video showing some uploads.

0:07
Ashaya Sharma avatar
Ashaya Sharma@ashaya·Oct 31

I was talking to a cracked ex-apple engineer this week which got me investigating this. The clanker did the rest. This probably exists already but 🤷🏽‍♂️