
Building @quanticmachines Working @utillylabs
0000
Joined September 2024
- Tweets180
- Following57
- Followers11
- Likes51
Dax’s post yesterday on compute and OpenRouter’s DeepSeek serving gave an IRL perspective on how much @quanticmachines should be preparing for.
This is an official notice that we @quanticmachines will be serving our models with our own infrastructure by 2027
Welcome to this side of the world. @quanticmachines already did this.
Embeddings were a mistake.
We matched OpenAI’s best embedding model on FiQA without using a single embedding. No vector database. No chunking. No re-indexing.
Today, we're launching @wondersearchai searcheverything your company ever wrote and find the exact line.
@ycombinator @fdotinc @daltonmeon
Loau retweeted
Qwen feels like the middle ground, 16.5t/s, 4.1s startup. Nemotron is 20t/s but taking my ram up to 6.2gb, but just 4.2gb for qwen, more ram for browser tools and all.
All my notes are now context and I can make research on anything in my notepad.
I’m going to build my personal assistant on one of my Instabox. I like Muse, but it is not there yet.
If I do anything spectacular during the process, I will post it on this thread.
Watching a lot of founders come up with ways to provide and make data useful in this AI age, only to realize that it can’t work in Nigeria or Africa as a whole, is painful. The biggest challenge we have is providing useful data for AI.
We are exploring a way to make your phone the computer your agent use while you also use the phone for your day to day @quanticmachines




