Andriy Mulyar
@andriy_mulyar
Followers
12K
Following
1K
Media
252
Statuses
4K
letting humans touch latent space @nomic_ai
New York, NY
Joined July 2019
when the nyc apartment rental verification app that requires you to oauth with your bank and your payroll provider has a create-react-app favicon 💯.
2
0
7
I remember when gibbons were dangerous
I remember when GPT-2 was dangerous.
1
0
1
The gap is crazy between retrieval agents in consumer AI apps (ChatGPT, Claude) and what can be cooked up with attention to detail.
1
1
8
Inference time scaling of diffusion models.
We will be presenting our work on steering and scaling diffusion models at #ICML2025!. Bonus:.- FK steering can beat grad guidance?! 🤯.- Boltz 🧬adapted FK steering to BioML!!! . East Ex. Hall A-B #E-1308.(Tue 15 Jul 11 a.m. PDT — 1:30 p.m. PDT)
0
0
5
nothing like a Friday evening annotating data in a vibe coded annotation interface.
1
0
10
today's learning: most document processing VLMs (layout, reading order, etc) can be backboned from distills of large OCR models.
0
0
4
RT @SkyLi0n: Just opened a PR yesterday that will reduce the binary size PyTorch by 40% by adding 1 flag to NVCC . With ~50M monthly of dow….
0
106
0
Those who can't teach do,.Those who teach can't do. - southern man, OBX, 4th July 2025.
1
1
5
Feel free to drop questions you want asked. I'm personally curious how many laptop chargers he's collected over the past few years.
2
0
16
In March 2024, I hired Soham Parekh (@realsohamparekh). I chatted with him earlier today and he has some things he wants to clear up. Join us for a short AMA at 3pm EST today.
17
13
137
RT @Suhail: PSA: there’s a guy named Soham Parekh (in India) who works at 3-4 startups at the same time. He’s been preying on YC companies….
0
3K
0
RT @chunkrai: this is an awesome use case. good parsing + structured outputs is a such powerful combination. if you don't want to send yo….
0
1
0
attending my first ever pride parade today in NYC, land of the free baby.
2
0
8
diffusion llm tech report.
✨The Mercury tech report from Inception Labs is now available on the Arxiv. It took us a bit of time to get this one out, but it’s a nice complement to the blog post with many more experiments. Stay tuned for more updates soon!
1
0
5
RT @nomic_ai: so exciting to get a chance to collaborate with @Wikipedia & @Wikimedia on the first full multilingual wikipedia map! even mo….
0
15
0
all hail tokens.
A 24-trillion-token web dataset with document-level metadata just dropped on @huggingface . License: apache-2.0. ESSENTIAL-WEB v1.0 collects 24 trillion tokens from Common Crawl. Each document is labeled with a 12-field taxonomy covering topic, page type, complexity, and quality
1
0
4