groundstate.ink

The Daily Edition  ·  Friday, October 2, 2026

The Index

◈ Science

Surprisingly Complex Waves Reveal the Brain’s Inner Workings

Quanta Magazine  ·  8 min

Traveling waves in the cortex have been known about for decades, but new high-density EEG work finds they’re far more structured than anyone assumed. Multiple interacting wave types move simultaneously across the brain, each shaped differently, each apparently carrying distinct information about the brain’s operating state. The complexity isn’t averaging-artifact. It looks like coordination infrastructure: the waves appear to be actively reorganizing cortical dynamics in real time, helping the brain navigate, access memory, and regulate itself from moment to moment.

The surprise is how much signal was hiding inside what researchers dismissed as noise. A good Quanta piece on how measurement resolution keeps rewriting what we thought we understood about neural architecture.

⬡ AI & Product

GLM-5.3 and the Spread of Advanced Cyber Capabilities

Anthropic Research  ·  6 min

Five months ago, Anthropic released Claude Mythos Preview as a limited tool for vetted cyber defenders — the first model capable of autonomously developing end-to-end exploits (V8 control-flow hijacks, binary exploitation chains). The bet was that giving defenders a head start before the capability proliferated was worth the risk. In this report, the Frontier Red Team documents what they were afraid of: Zhipu AI’s GLM-5.3 has now crossed the same threshold, and it’s available to anyone. The capability jump mirrors Opus 4.6 → Mythos Preview.

The critical difference is safeguards. GLM-5.3 ships with guardrails that can be bypassed 64–100% of the time using elementary jailbreak techniques that don’t work against Claude models. NIST/CAISI independently assessed the US-China frontier gap at ~4 months on aggregate cyber benchmarks — but attackers can’t easily access US frontier models with reduced safeguards; they can download GLM-5.3 right now. The era in which autonomous exploit-development capability was gated to vetted users appears to be over.

◉ Wildcard

C V Raman: the Physicist with a ‘Himalayan Ego’ Who Went to Stockholm Convinced He’d Win a Nobel Prize

Physics World  ·  7 min

In 1930, Raman was so certain he would win the Nobel that he sailed to Stockholm two months before the announcement rather than risk missing the ceremony. The journey from India by sea took the better part of that time. He arrived on December 9th. The announcement came the next day. He was right. Physics World’s profile by Preetish Kakoty traces both the physics — Raman scattering began with a pocket spectroscope and a question about why the ocean was blue, on a sea voyage in 1921 — and the personality. The “Himalayan ego” is a direct quote from colleagues; it’s not the article’s editorial judgment.

A good portrait of what self-belief looks like when it turns out to be correct, and a reminder that the first person from India to win a scientific Nobel did it mostly on defiance.

Quanta Magazine  ·  5 min

Maggie Miller at UT Austin, in conversation with Janna Levin and Steven Strogatz, explains why 3D intuition collapses in four dimensions and how she actually works with 4D spaces: by treating them as a reel of 3D snapshots — the same logic by which a film is a reel of 2D frames. It’s a multimedia piece (short), but the framing is clean and the analogy earns its keep. Miller’s area is 4-manifold topology, so the intuition she’s building is the real thing, not a popularization shortcut.