Daniel Yurkin (@danyurkin) / X

Daniel Yurkin

419 posts

Daniel Yurkin

@danyurkin

building @atomic_chat_hq - local AI app and inference engine find me: discord.gg/8wGSsvmg4V

San Francisco, CA

linkedin.com/in/danyurkin

Joined July 2025

Daniel Yurkin
@danyurkin
Jul 30
good company to be in! ai leadership comes down to how many people can actually run these models and build on top of them thanks @Microsoft for pulling this together and to every team building in the open
atomic.chat
@atomic_chat_hq
Jul 30
Atomic Chat signed the Open Weights letter! We believe everyone should be able to run AI on their own device. When a model is open, thousands of teams fine-tune it, quantize it and build new tools on top of it. Together we move AI faster than any one company ever could
Daniel Yurkin
@danyurkin
Jul 22
brooooos lmao 🤣🤣🤣 anthropic scraped literally the entire internet to train fable and now they’re out here talking about IP theft
Director Michael Kratsios
@mkratsios47
Jul 22
We have information that Moonshot AI distilled Anthropic’s Fable for the development of its K3 model. To do this they developed a sophisticated internal platform to conduct large scale distillation against U.S. models, allowing them to quickly switch between multiple methods of
Daniel Yurkin
@danyurkin
Jun 19
yo @kilocode team! thanks for shipping this live 🤝 huge step forward - love seeing integrations like this happen
atomic.chat
@atomic_chat_hq
Jun 19
Run KiloCode on open models powered by Atomic Chat @kilocode is trusted by 3M+ developers. Code, debug, and build entirely on your device — no cloud, fully private, free and open-source with @atomic_chat_hq local AI provider
00:00
Daniel Yurkin
@danyurkin
May 29
running lfm2.5-8b-a1b locally in our atomic chat on m3 pro 18gb (Q6_K) - 70 tok/s. only ~1.5B of 8.3B params fire per token, trained on 38T tokens, built for local workflows and tool calling a dense 8B at the same memory footprint would be 3-5x slower
00:00
Liquid AI
@liquidai
May 28
Today, we're releasing LFM2.5-8B-A1B, a device-optimized model designed to power real-life applications on phones, laptops, PCs, robots, and fast & lightweight server-side use-cases. > 8B MoE, 1.5B active > Expanded 128K context > LFM2.5 flagship hybrid MoE architecture >
Daniel Yurkin
@danyurkin
May 26
running this locally on an iPhone is kinda crazy ngl! big deal @PrismML
PrismML
@PrismML
May 26
Today we’re releasing 1-bit and Ternary Bonsai Image 4B. A new family of image-generation models designed to run high-quality diffusion inference on local hardware: from laptops to phones.