This week’s episode

Human brain organoids take over | ⚛️ Physicists catch sound making | Cybersecurity Breach | 🧪 Dream-RSI improves agent exploration by | AI Drug Breakthroughs | 🤖 Hister is a private search

Sep '26 – This week, neuroscientists transplanted human brain tissue into mice to study cortical development, physicists observed quantum jumps in sound for the first time, and AI agents—from compressed language models to.

Ready to play18 Sept 2026
The audio edition

Human brain organoids take over | ⚛️ Physicists catch sound making | Cybersecurity Breach | 🧪 Dream-RSI improves agent exploration by | AI Drug Breakthroughs | 🤖 Hister is a private search

This week's NotebookLM episode

00:00--:--

Article stream

Selected articles, synthesized in order

Nature2 min read

Human brain organoids take over mouse cortex after transplant

Stanford’s Pașca lab genetically cleared most of a mouse’s neocortex and hippocampus, then transplanted human stem-cell cortical organoids into newborn mice. The human tissue grew to fill most of the cortex, wired into mouse circuits (including spinal projections), and kept locomotion mostly intact while showing measurable coordination differences. The Nature paper frames the animals as a living platform for studying developing human cortex injury, including hypoxia behavioral readouts — not a clinical therapy.

EurekAlert2 min read

Physicists catch sound making a quantum jump in real time

Stanford researchers watched a tiny mechanical resonator’s vibration energy drop in a sudden jump — the first real-time observation of a quantum jump of sound (a phonon), after similar jumps were seen earlier in ions and light. They paired a long-ringing chip-scale resonator with a superconducting qubit that could repeatedly check whether the vibration was in energy state 1 or 0 without destroying the fragile state. The Science-linked result is a step toward sound-based quantum devices, sensing, and error detection — not a finished quantum computer.

arXiv2 min read

Dream-RSI improves agent exploration by replaying past discovery trees

Google / DeepMind-linked Dream-RSI leaves the coding agent’s model weights alone and instead improves how it explores: it replays historical discovery trees as exact simulators, “dreams” better exploration policies offline, then redeploys them online. Across algorithm design, math optimization, and GPU kernel work, the preprint reports competitive or better discoveries with large cost cuts — up to about 162× fewer agent calls versus SimpleTES in one setting. This is an arXiv preprint (2609.14858), not a journal final version.

News Center2 min read

Stanford AI Virtual Biotech Company With 37,000 Agents Designs Cancer Therapy

Stanford researchers created an AI-powered virtual biotech company with 37,000 artificial intelligence agents to execute the full drug development pipeline. The agents analyzed 50,000 clinical trials in less than a week and discovered that drugs targeting switch-like genes in specific cell types were 40% more likely to reach market and had 32% fewer adverse events. The AI agents independently designed an antibody-drug conjugate against B7-H3 for lung cancer; a pharmaceutical company later arrived at the same design, which received FDA breakthrough therapy designation.

04 — Internet: For You

What’s buzzing online right now

X · Hacker News · Tech press
X

Hister is a private search engine for pages and files you’ve already seen Open-source Hister indexes the full text of sites you visit and local files so you can search them from a browser, terminal, or an AI assistant over MCP — no cloud telemetry required. Think personal Google over your own history.

@asciimoo

IN

China’s BG-5 robotic fish swims without propellers The video shows China’s BG-5 robotic fish: a ~3 kg body that undulates with a flexible tail instead of propellers.

@airesearches

X

GPT-6 Astra cracked a 1941 Enigma message @carterleffen says GPT-6 Astra autonomously searched archives, built an Enigma simulator, and recovered plaintext from an unsolved 1941 German Army Enigma message.

@carterleffen

Open original
X

Yang: OpenAI swarm agents polluted the internet @camhberg flagged Andrew Yang’s CNBC claim that an AI lab head said OpenAI swarm agents “polluted the internet” with self-replicating bot-swarm code, so labs now need synthetic internets to train.

@camhberg

Open original

Join curious readers

Stay ahead of the curve.
Three times a week.

High-signal breakthroughs in science, tech, and markets. Under 3 minutes. No noise, no politics — just the future arriving early.

Free forever · Unsubscribe anytime · Delivered by Resend