danfiru

Cofounder & Product @ Quadric. More with More.

[danfiru @danfiru]

an engineer ported DOOM over the weekend. 36fps on a 50MHz fpga emulation. your NPU runs neural networks. ours runs programs. full article in the comments

[danfiru @danfiru]

Power to the local

[David Hendrickson @TeksEdge]

🌞This is big Local AI news! A new open-source Computer-Use LLM has just launched.

Holo 3.1 is H Company’s (🇫🇷) new local computer-use agent model that beats Qwen3.5-397B, Kimi-K2.5, and Sonnet 4.6!

Since it is built for local deployment →
⬩ Runs fully on your machine (MacBook, Windows PC, DGX Spark, RTX Spark)
⬩ Based on Qwen architecture, specialized for GUI understanding & computer control
⬩ Optimized checkpoints: NVFP4, FP8 & Q4 GGUF (0.8B to 35B sizes)
⬩ Strong gains: 79.3% on AndroidWorld benchmark (35B model)

💻 Comparison to Qwen3.5:
Holo 3.1 is fine-tuned specifically for computer-use agents (screen understanding, planning, clicking, navigation). Better at real GUI tasks than general-purpose Qwen3.5, especially when running locally.⚡

[danfiru @danfiru]

A tungsten cube would fix me.

[danfiru @danfiru]

@HuggingModels I love MoEs. I'm infatuated with the idea of running intelligence at home. MoEs are a great unlock for that.

[danfiru @danfiru]

@MTSlive @fleetingbits honestly the first movers have to be one. the company that can build robots that don't fucking suck will establish a good lead for themselves

[danfiru @danfiru]

@TiezhuCrypto the value of intelligence is still far greater than the value of the memory. the short-term kicker is dram availability.

[danfiru @danfiru]

@teortaxesTex 3d integration is the dream is you can get thermals to work

[danfiru @danfiru]

@The_AI_Investor i've had chips sitting waiting for all kinds of reagents in the past. i can't imagine the level of control required at that scale.

[danfiru @danfiru]

@FirstSquawk 800 bln won is a bold move.

[danfiru @danfiru]

Agent pilled engineers are modern day wizards.

[danfiru @danfiru]

@akshay_pachaar This is a good video.

[danfiru @danfiru]

@niccruzpatane luce is a vanity project for the 1%. model 3 is a practical supercar for the rest of us. (i love mine)

[danfiru @danfiru]

@NousResearch this is cool. impact on model performance?

[danfiru @danfiru]

@SMB_Attorney living to over 100 seems possible

[danfiru @danfiru]

@ihtesham2005 nothing beats walking, tbh

[danfiru @danfiru]

@yacineMTB local models will win in the long

[danfiru @danfiru]

@elonmusk going to try

[Josef Chen @josefchen]

Launching our new paper on arXiv: we trained the largest multilingual food model ever built.

4.1M recipes. 7 languages. 1,790 ingredients. 300 dimensions.

All of human cooking compressed into 2 megabytes.

[danfiru @danfiru]

@_TommyMason i'll take the leaf and a rack of npus for tokens at home.

[danfiru @danfiru]

@brfootball Dean Huijsen is the only snub, tbh. Madrid have always been a more international club than Barcelona.

[danfiru @danfiru]

This would fix me.

[Tom's Hardware @tomshardware]

Enthusiast runs 1-trillion parameter LLM from 768GB of Intel Optane DIMM memory sticks https://t.co/RbPzqpjAk6