@athletic_coder

mostly ml here;

heyyanshuman.com
Joined February 2020
  • Tweets10.5K
  • Following1.1K
  • Followers20.4K
1.2KMedia
I need VC funding. C=Compute
2
5
793
they cost "a lot" to run
Michael Burry Stock Tracker ♟
@burrytrackerOct 27
Breaking: Michael Burry backs Ken Griffin's AI bubble warning Griffin says AI chips are like airline seats: • Once built, they cost almost nothing to run • Build too many and prices go "into freefall" • He says AI is "almost certainly" in a bubble Burry backed Griffin saying "This is obvious to most with a working brain cell"
2
1,288
never leaving this app
1
4
887
never leaving this app
1
4
887
banger
Mehul Mohan
@mehulmptOct 23
can't wait for people to share guides on how to host 3 trillion kimi k3.1 pro model on their laptop getting 1000s/t speed
716
RSI is the new AGI
3
740
I put together a comprehensive resource list for learning model post-training and building useful applications. now i'll be sharing the resources one by one. this is my post-training resources series. save this as your starting point. links in thread 🧵
1
5
9
1,180
very curious if they used deepseekharness for automationbench maxxing
DeepSeek
@deepseek_aiOct 17
🚀 DeepSeek-V4.1-Flash-1017 is here! Stronger benchmark performance, faster task completion, and shorter outputs than the previous version—a Pareto improvement in performance and cost. Now live on our website, app, and API. 🛠️ Optimized for @DeepSeekHarness, Tencent CodeBuddy, WorkBuddy, and more. Upgrade to DeepSeek Harness v0.2.1-rc.1 or later for the best experience with the new model. 1/4
6
1,737
$0.75 / 1000 searches ???????????? exa is cooked.
DeepSeek
@deepseek_aiOct 17
Replying to @deepseek_ai
🔍 Search & Fetch APIs are now open to everyone! Powered by our in-house search engine, serving hundreds of millions of requests daily worldwide. 💰 Each endpoint: $0.75 per 1,000 calls. 📖 Docs: api-docs.deepseek.com/guides… 📌 Search & Fetch calls via server-side tools in DeepSeek Harness, Claude Code, and Codex are currently free for DeepSeek API users. Billing starts Oct 31, 2026, at 16:00 UTC. 3/4

Search & Fetch | DeepSeek API Docs

The Search API and the Fetch API give agents and RAG applications access to live web content:

api-docs.deepseek.com
51
85
3
3,912
280,981
who's doing this as a service
merve
@mervenoyannOct 17
I've been saying biggest moat for companies is ability to further post-train on your own domain using open models and now open RL envs (both cost a ton!) I don't want to call for a fomo but if you aren't doing this you're missing out
3
3
17
3,437
I put together a comprehensive resource list for learning model post-training and building useful applications. now i'll be sharing the resources one by one. this is my post-training resources series. save this as your starting point. links in thread 🧵
1
5
9
1,180
more replies
explore multimodal RL for vision, speech, and image or video generation. evaluate on held-out tasks, study safety and alignment, and read open model training reports.
1
100
the reading order starts with the fundamentals, then builds toward the systems and research behind modern post-training. full resource list: github.com/athleticcoder21/p…

GitHub - athleticcoder21/post-training-resources: A curated resource list for learning post training

A curated resource list for learning post training - athleticcoder21/post-training-resources

github.com
101
NVIDIA released a transformer that can decode in auto-regressive and diffusion mode. That means INSANELY FAST INFERENCE. I was curious if i post-train a model in auto-regressive mode, do gains carry over to diffusion decoding mode? turns out it does: heyyanshuman.com/posts/rl_ar…
4
8
1
32
2,108
NVIDIA released a transformer that can decode in auto-regressive and diffusion mode. That means INSANELY FAST INFERENCE. I was curious if i post-train a model in auto-regressive mode, do gains carry over to diffusion decoding mode? turns out it does: heyyanshuman.com/posts/rl_ar…
4
8
1
32
2,108
@nvidia 's nemotron labs trained this tri-mode LM arxiv.org/abs/2607.05722

Nemotron-Labs-Diffusion: A Tri-Mode Language Model Unifying...

We introduce Nemotron-Labs-Diffusion, a tri-mode language model (LM) that unifies AR, diffusion, and self-speculation decoding within a single architecture. Trained with a joint AR-diffusion...

arxiv.org
1
3
328
and in this paper @JensenHuang n team show how you can teach an auto regressive model how to diffuse.
3
160
Load more