founder, coder, hacker; crypto systems + ml
San Francisco, CA
Joined April 2012
- Tweets525
- Following803
- Followers354
- Likes847
Aaron Li retweeted
wtf
this guy just had Claude Sonnet 5 perform brain surgery on a living mouse
the goal was to treat the mouse's epilepsy so it would stop having so many seizures.
here's what Claude did:
> it looked at 19 possible treatments and picked one that boosts the brain's natural calming chemical
> it chose where to inject it based on where these seizures are thought to start
> it worked out the exact injection point by averaging three research papers
> it lowered the needle 4.3 millimeters into the mouse's brain
> it injected a 20,000th of a teaspoon of the treatment over five minutes
> it checked the live camera 32 times to watch for things like bleeding
and the mouse woke up and recovered well!
interestingly, Opus 5.5 and Opus 5 both refused the task because it involved real biology
so Sonnet 5 got the job.
and as far as the researcher knows it's the first time an AI has picked a treatment and then injected it into the brain of a living mammal.
Last night an AI agent performed brain surgery on a mouse.
We gave it 19 gene therapy vectors in a spreadsheet, a mouse model of epilepsy, access to a motorize surgery rig with cameras and one goal: reduce seizures.
It picked the drug. It picked the brain region. It derived the coordinates from three papers. Then it drove the stereotaxic arms into the brain and injected 250 nanolitres while watching the entry site through the rig camera for bleeding.
A licensed vet gated every irreversible step and held a physical kill switch to stop the AI agent.
Thread on what it decided, what it saw, and the moment it refused to let me override it.
[Writing this on the spot out of excitement, but I've been 18hrs in front of my laptop coding and coordinating this, so keeping it brief and descriptive, more cohesive thoughts on this soon...]
Aaron Li retweeted
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.
block.xyz/dessa
Looked into mmWave, THz scanner, Faraday bags for dealing with hacked ledger as a fun hackathon project with @Aileengf1 at Frontiers 2026. Still WIP - the methods are promising - cheap, non-intrusive enough for every day use. Maybe @Ledger should have a built-in Faraday housing
- Kitchen scale: [HOTO](hototools.com/products/smart…)
- Milligram scale ×2: [MAXUS](amazon.com/dp/B07ZQZ962K)
- Rigid ruler: [Benchmark](taytools.com/benchmark-rigid…)
- Window Faraday bag: [Mission Darkness](amazon.com/dp/B01A7JLN7C)
- RF display/controller: [Flipper Zero](flipper.net/products/flipper…)
- RF detector module (DIY software required; currently unavailable): [RF Meter Click](mikroe.com/rf-meter-click-cl…)
- RF antenna: [Nooelec UWB Surveyor](nooelec.com/store/sdr/sdr-ad…)
- Antenna cable: [SMA male–male](nooelec.com/store/sdr/sdr-ad…)
- GPIO jumper wires: [Adafruit](adafruit.com/product/1954)
- mmWave radar kit: [Infineon BGT60UTR11AIP](digikey.com/en/products/deta…)
- THz system (quote required): [Menlo TERA ASOPS](menlosystems.com/products/th…)
- THz imaging attachment: [Menlo TERA Image](menlosystems.com/products/th…)
better mmWave radar kit [Infineon BGT60TR13C]: mouser.com/en/ProductDetail/…
time for 2 more sparks - wish there are cost efficient high bandwidth alternatives at $20-30k level
GLM‑5.3 full, the big one, running at home on 4 DGX Sparks.
Same galaxy prompt as my Flash test. One self‑contained Three.js file.
• Quant: @vic305's EXL3 3.38bpw SAGE (319 GB)
• Engine: TensorFold TP4 + DFlash2, 4× DGX Spark
• 85.7K tokens (68.9K thinking)
• 42.4 min · ~34 tok/s
• 939 lines · 44 KB HTML
• Thinking max, no output cap
• 1 fix: exposure 1.15 → 0.85 (one line); core dive still bright
Thanks @vic305 for the quant + 4‑Spark recipe, and @ashxhart for tensorfold
Aaron Li retweeted
Pair the R1 with your 3D printer and you can make transparent acrylic art at home.
Thundercloud Lamp by Daniel4d
Layered Clock by Bambu Lab
share.bambulab.com/4dSBo83
Product idea: a hyperlocal temporary marketplace for small tasks - like a Taskrabbit for conferences - where people can offer or get paid doing tasks on demand such as:
- waiting in line to get coffee / food
- attend a talk and take photos / videos / notes, observe vibe
- collect merch / contacts / info from booths
- find and reserve charging places, seats, meeting tables
- "tour guides"
- ...
Aaron Li retweeted
Meet EmbeddingGemma 2, our first natively multimodal open model for on-device embeddings.
It expands beyond text to unify code, images, audio, and video in a shared space. 🧵
Aaron Li retweeted
Today we’re announcing Personal Agent Protocol — an open standard @Meta and @SierraPlatform are developing along with industry partners at @Genesys, @instinct, @RocketOTD, @Shopify, @stripe, and @Walmart. It will help define how personal agents interact with businesses and is open for anyone to implement.
Read more: sierra.ai/blog/introducing-p…
Much simpler onchain & without hopping layers and bridges
Migration complete. Derive is running on its new V3 architecture.
Market makers and traders are reconnecting as liquidity builds back up. Withdrawals for some assets may take up to a week to resume.
We'll share more about V3 in the coming days, including a closer look at the new architecture and what’s next for users and builders.
Aaron Li retweeted
We've been busy.
→ Deribit, the world's largest crypto options exchange, is now part of Coinbase
→ Soon, a first: US + global derivatives in one regulated liquidity pool
→ Next: options, spot margin + unified portfolios
And Coinbase Pro returns later this year.
Aaron Li retweeted
Meet Bambu Lab R1. Big job. Light work.
Starting at $2,499
Pre-orders now open share.bambulab.com/4dSBo83
Aaron Li retweeted
What I like about Jev: for years, large generalist models sucked almost all the oxygen out of AI.
But there are huge opportunities in much more specialized and customized models, built for specific tasks and languages and as a result, orders of magnitude cheaper, faster and more optimized. There are 3 million of them publicly available on @huggingface.
Let's build a much more diverse AI ecosystem!
Aaron Li retweeted
Steam Frame is here! Our wireless VR headset + controllers, available now in the following models:
Steam Frame 256GB
Steam Frame 1TB
Sign up now and learn more on @Steam store.steampowered.com/hardw…
Tried polymarket.us at Blizzcon (Overwatch Championship) for the first time - fast, smooth, crypto invisible throughout the whole user experience
