Field Notes · 5 Jul 2026
Best explained
Why larger networks tolerate unstructured architectures: as network scale grows, the model has enough capacity to compensate for suboptimal structure, but at small scale tight architectural choices are critical for performance.
By the numbers
Liquid Neural Networks can perform autonomous car driving with 19 neurons, parallel parking with 12 neurons, and drone navigation with 30 neurons.
New today — Evidence added
Liquid AI published the closed-form solution to liquid neural network neuron-interaction equations in Nature Machine Intelligence in November 2022.
Ramin HasaniLiquid Neural Networks can be as small as 1-25 megabytes in file size and run on a CPU or Raspberry Pi for specialized predictive applications.
Ramin HasaniNew today — Sharp calls
Ryan Sean Adams on dual token equity structures: shareholders get Delaware law, token holders get a pinky promise.
Ryan Sean AdamsSierra's engineering interview now issues candidates a $150 token budget to build an app using any coding agent and tools of their choice, signaling a new industry benchmark for AI-native hiring.
Clay BavorNew today — Useful context
Marc Andreessen on products outpacing companies in global expansion.
Marc AndreessenNew today — Watchlist
Ondo targets $1.5 to $2 billion in tokenized stocks by end of 2026.
Ian De BodeMicroStrategy released a Digital Credit Capital Framework, signaling a strategic shift toward operating as a broader crypto hedge fund rather than a pure Bitcoin treasury.
Ryan Sean AdamsTool watch
Liquid AI's AFMD automated architecture search produced the double gated convolution as the optimal efficient architecture, eliminating hand-tuned gating from models like Mamba and gated delta nets.
Worth quoting
“You cannot sell Bitcoin. If you are selling Bitcoin, the whole system breaks.”Michael Saylor
“The reality is that products and technology travel faster than companies can.”Marc Andreessen
“It's not the stress that burns us down, it's the response to stress.”Martin Picard
“We believe it would be good for the world to have the option to slow or temporarily pause frontier AI development.”Jason Calacanis
“You're spending a trillion dollars a year and the revenue on AI talk about maybe 70 80 billion.”Rajiv Jain
“40 to 41% of employees saying they ship AI work that they couldn't explain if asked. This is bot f***.”Rebecca Hinds