People dunking on this have a kind of Stockholm Syndrome from the difficulty of using computers. Being able to just say what you want or need and have it happen was always the dream of the Computer Revolution. The original pioneers would be elated the difficulty has melted away.
My 12 year old son just made this @Tesla Roadster game from scratch by himself using AI.
AI is nuts.
Ty retweeted
The funny thing is everyone was saying AI slop code will cause a cyber crisis but it’s actually all the old human slop causing it and the AI is just cleaning it up
Ty retweeted
This dude MADE UP a story about me after ripping my video pretending I use some second brain grok bot bullshit 😭
I am not affiliated with this guy or his grift I use Hermes on a dual spark cluster and of course it scrapes my email for leads but not through some obsidian BS
FFS
25-year-old developer copied 6 years of his emails, client calls and notes, pasted them into Grok Bot and told it to build a brain that thinks exactly like him.
20 minutes later - a complete second brain.
Then he boarded a 3-day train across America with no Wi-Fi. In his backpack: a laptop and a $4,000 box the size of a sandwich running a 200B-parameter model. One Ethernet cable between them. The whole brain lived inside that box.
Somewhere in Nebraska, zero signal, he typed “what am I forgetting?” It found 3 client follow-ups he’d dropped months ago, the exact price he’d quoted each one, and wrote every reply in his own voice.
Sent them at the next station. Two paid before Denver. $6,400 from a train seat with no internet.
Then he built a machine.
The brain now reads every email and call as it happens, finds him new clients, writes his proposals, makes his decisions and plans his week. Zero cloud, zero subscriptions, nobody ever sees his data.
Month one - $6,400 from a train. Month six - it runs his business for him
Most people still use AI to fix typos in their emails.
A second brain makes you money, remembers everything and solves problems before you even notice them, and almost nobody has built one yet.
Ty retweeted
The guy who gave us QLoRA, says DeepSeek's inference system is on par with OpenAI's stack and better than what Anthropic runs, at least at their scale.
Now check the license. Apache 2.0.
Until today a production stack at that level existed inside a handful of companies. It was the thing that made their API fast and their margins fat, and none of them ever showed it to anyone.
Today any startup, any university, any cloud in any country can download it, read it and run it. It runs on NVIDIA Blackwell and on Huawei Ascend, so it isnt tied to 1 chip vendor either.
This is what I mean when I say open weights are about more than downloading a chatbot. DeepSeek is giving away the knowledge that only a few labs had, and that changes who can build an AI company at all.
Respect @deepseek_ai
Insanity
Introducing statistically-lossless Qwen3.8-27B!
Three models, 23.7–30.5 GB: up to 57% smaller weights than BF16.
98–99% measured EAR, versus 97.71% for native FP8.
EAR measures overlap with BF16’s next-token probabilities. A 99% EAR means 99% average probability overlap, here, measured over the reference’s top-10 tokens after renormalization.
The search optimizes EAR directly, assigning more precision to sensitive projections while keeping activations in BF16. The goal is to find the smallest checkpoint while keeping fidelity as high as possible.
vLLM + Humming • Apache 2.0
GGUF versions coming next week!
Get the models:
huggingface.co/collections/k…
Based on SLQ (arxiv.org/abs/2605.02404) from ISTA’s DASLab (github.com/IST-DASLab), the group behind GPTQ, GSQ and RCO, with customizations to improve the balance between size and fidelity.
More details soon.





















