πΌπππππ πππππππ ππππ ππ’ πππππ π πππππ πππππ-ππππ https://nitter.cf/t.co/BjeBCRVHcI https://nitter.cf/t.co/SuEfJVnn2P
Earth
Joined July 2009
- Tweets7.1K
- Following2.4K
- Followers5.6K
- Likes33.9K
Bio compute being shipped to the masses. Brain cells hooked up to electrodes processing data, creating visuals.
How do you know you are not just a consciousness hooked up for observation and computation?
today, we introduce a little mischief to the world.
for four years, weβve been brewing real brain cells on electrodes to understand how they compute. now, weβve turned those discoveries into a video model anyone can use.
from 0.5Β’ a second, mischief 1 is at least 5x faster than its base and the cheapest image-to-video model on @fal and @OpenRouter. soon to be on @awscloud marketplace.
itβs not a delusion anymore. biocompute is real, commercial and this is just the beginning.
happy halloween.
Readers added context they thought people might want to know
This is a post-train of Wan, an open weights model, and is not running on actual neurons.
The authors claim they used βinsights from the neuronsβ to make the model cheaper, but many post-trained video models exist with no relation to bio-compute. unite.ai/tbc-mischief-nβ¦
Itβs not exclusively about efficiency, itβs also about self governance and a sense of sovereignty.
Individuals have a right to build their own things and use them how they see fit.
Is Sam suggesting we lay down and let him consolidate power?
sam: local inference is a terrible allocation of scarce resources
βa lot of people want local inference, and somehow they assume that the problem is solved now because we have open-source models, we're going to be able to run them on our computers, and so we're not going to need all the stuff in the data centre
but from a global allocation perspective, actually it's quite terrible to have everybody, like, have a bunch of the world's limited supply of DRAM sitting in their own computer and being used five percent of the time
you want maximum utilisation in the cloud if you're trying to deal with this very supply-constrained environment
somehow many people's intuition is the opposite of thatβ
Data centers are being built with human brain tissue because itβs so much more power efficient per computer cycle.
Not sure how I feel about that, what does that remind you of?
youtu.be/mwISkfdRN4o
A small part of a larger project, fully powered by a Single DGX Spark β‘οΈ
Running Qwen3.8 27B AEON Ultimate Uncensored
Do not underestimate the power of a good Qwen3.8 27B finetune
A new day; a new model!
Intriguing for several mobile, edge and finetuning use cases.
It wonβt be the smartest but it might be smart enough and that makes all the difference.
Anyone have experience with this one yet?
Introducing π»π΅πΌπΉ: our new generation of Liquid Foundation Models. With a new tri-hybrid architecture and new multimodal encoders, the LFM3 family is the ideal model suite for specialized intelligence.
> 6 model instances, across text, audio, vision, and omni modalities, generative and decision models.
> π»π΅πΌπΉ-π·.π½π±: text, 256k context window.
> π»π΅πΌπΉ-π
π»-πΈπ±: vision-language model with video-input support for spatial understanding
> π»π΅πΌπΉ-π°ππππ-πΈπ±: tool calls, voice chat, reasoning in 18 languages.
> π»π΅πΌπΉ-πΎπππ-πΈπ±: an omni model combines vision, audio, and language.
> ππ·-πΈπ± and ππ·-πΎπππ-πΈπ±: decision models with typed outputs and calibrated probabilities in a single pass. Omni adds audio input.
> fine-tune on 1 GPU.
> a new tri-hybrid architecture
> architecture optimized for 14 processor classes across GPUs, CPUs, and NPUs.
> π»π΅πΌπΉ-π
π»-πΈπ± is a beast. It performs competitively with Gemma 4 12B (6x larger), and Gemini 3.5 Flash Lite, and GPT-5.6 Luna, with 1.9B parameters.
> π»π΅πΌπΉ-π·.π½π± is #1 on the Pareto frontier of device intelligence πππππππ.ππππππ.ππ
Every modality for specialized intelligence. Pick the one that fits your use case, customize it, and run it on-device.
1/ π§΅
You do have to be careful running an Abliterated model and understand the risks of wielding raw unconstrained power. I recommend limiting exposure externally to your models and enabling your own agentic guardrails to safeguard from unintended execution of malicious instructions.
RANDOMNESS is NOT ENTROPY.
It just happens to be exponentially more likely that true randomness will be high in entropy.
Entropy is the measure of chaos
Randomness can be a string of 0s and a single digit (low entropy)
You want high ENTROPY +true randomness + high bit length
I believe EVERYONE has a right to transact. NO ONE needs permission to buy sell or transact value of any form.
The people vote with their money and their will. Let the people decide where the money goes.
Organizations like Patreon, Payment Providers, and Centralized platforms are so prideful and arrogant to think they have the right to decide who can and can not participate in society.
Let the people decide!
Well Iβm getting complete radio silence from @Patreon
They have run off with nearly $500 in community contributions (you all better get refunded) I hadnβt added my bank account yet to withdraw.
Building their replacement now, drop next level feature requests below π
Last remnants of the previous timeline, new orthogonal cutover starts now.
The sun looks so strange today in the sky.
It seems AI is bringing to light what has been happening in the shadows for ages.
The math and science institutions have been stonewalling anyone not on the in. Brilliant ideas dismissed, solutions discarded. Especially when discoveries challenge the classical model.
I have a calling to liberate the people, this is a spark to the flame π₯
Time to start building more freedom power and creator empowering tech!
Itβs not over, we are just getting started!
I want to make beautiful things that make the world a more exiting place to be!
Iβve submitted an appeal hopefully they actually review it and reverse the decision. This was going to be a key to getting more time allocated to actually working on AI and delivering bigger and better work for everyoneβ¦