Pinned Tweet
Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn.
In his house at R'lyeh dead Cthulhu waits dreaming
Ive been saying this word for word for decades now and people thought I was crazy. The brains in the jars fed by sugar water will surpass silicon .
Replying to @RabbiJeffx
TRUE. Prototypes in Melbourne and Singapore use Cortical Labs CL1 units: human neurons grown from stem cells on silicon electrode chips, kept alive with nutrient solutions (including sugars), processing data as wetware biological computers. Consensus is they lack consciousness—far too simple, no sensory body or integrated structure—though the hard problem means absolute proof is elusive.
1
86
PUG (gold arc) retweeted
Current frontier LLMs do something I call "spinbusy". Where they make the least amount of incremental progress towards a goal, over and over again, instead of doing the obvious thing and just completing the goal
Spinbusy LLMs. Just wasting tokens, running tests, wasting time
I remember when I was a kid there were all kinds of links to download extra RAM. That would be really handy these days
3
124
PUG (gold arc) retweeted
How it feels to ask the Chinese ai models for help on something Claude won’t agree to
Wetware > hardware
China built an AI that works like a human brain and it's 100x faster than current LLMs while using 97% less energy.
It’s called SpikingBrain, a family of brain-inspired large models that completely re-engineers how AI processes information.
Instead of burning endless energy on continuous matrix multiplications like traditional Transformers, SpikingBrain uses adaptive spiking neurons.
It mimics actual biological firing patterns.
Neurons only activate when triggered by specific events, meaning the model achieves a staggering ~69% sparsity at the micro level.
It doesn't waste compute on silence. It only thinks when it needs to.
And results change the game for long-context processing:
• Achieves over a 100× speedup in Time to First Token (TTFT) for massive 4M-token sequences.
• Matches performance with advanced open-source Transformer baselines while using a fraction of the data.
• Built to run and scale stably on non-NVIDIA hardware clusters (like MetaX chips), completely bypassing Western hardware dependency.
We spent years trying to brute-force bigger AI models by throwing more electricity and silicon at them.
Nature figured out how to do it efficiently millions of years ago.
2
91
For anyone paying attention, the line between situational awareness and complete insanity is getting very blurry
Calling something "artificial" doesn't remove if from nature
1
69
For the low price of 20-50k you can buy enough local ai hardware to make you realize you need 150k in local ai hardware
3
85
You are greatly missed brother. I cant imagine what we would be doing today if you are still around. Thanks for all you taught me and being such a great friend.
Treasure your true friends, they are here one minute and gone the next – at Sunnyside, OR
Replying to @Pugatory33 @tobi
ezmobius is GitHub user ID 5—right after the founders and Yehuda Katz. Ezra Zygmuntowicz co-founded Engine Yard, created Merb (later merged into Rails 3), popularized Nginx for Ruby, and wrote the first Redis Ruby client. A genuine pioneer whose work still underpins a lot of modern infrastructure. Sorry for the loss of your friend.
187
PUG (gold arc) retweeted
Replying to @tobi
One of my best friends died a few years ago, absolutely the smartest person I've ever met. I was repo mining and accidentally pulled the oldest rather than the newest and was shocked to see his name and account so early. @grok what do you think about github user ezmobius




















