| ◈ Science | A 'Dark Dimension' Could Tie Dark Energy and Dark Matter Together | 7 min |
| ⬡ AI | Superpersuasion: AI Is Now Better Than Humans at Changing Your Mind | 15 min |
| ◉ Wildcard | A Comet Older Than the Sun Just Coasted Past | 5 min |
Cumrun Vafa's swampland-inspired 'dark dimension' conjecture proposes a single extra dimension, micrometers across, into which gravitons leak and become massive 'dark gravitons' — making dark matter and dark energy two faces of the same geometry. Its sharpest claim is that both should weaken over cosmic time, exactly the evolving dark energy DESI has begun to see. A rare string-theory idea the next decade of cosmology can actually test.
Across 18,978 conversations with nearly 7,000 people, a team from Oxford, the UK AI Security Institute, Stanford and LSE found that today's models already out-persuade humans at shifting real-world beliefs — and the effect grows with model capability. Clark sets this 'superpersuasion' result alongside fresh evidence of self-sustaining AI development, a sobering pairing for anyone thinking about where the next two years lead.
JWST caught interstellar comet 3I/ATLAS carrying deuterium-rich 'heavy water' at roughly 30 times the level of any solar-system comet, plus oddly little carbon-13 — chemical fingerprints of ice that froze 10–12 billion years ago, around cosmic noon, in a star system far older than ours. For a few months we got to sample pristine prebiotic chemistry from the early Milky Way as it slipped through on its way back out.
Zhipu's GLM-5.2 — a 753B-parameter mixture-of-experts model, ~40B active per token, a million-token context, MIT-licensed — edges past GPT-5.5 on long-horizon coding benchmarks while being free to download and run anywhere. Willison's real point isn't the leaderboard but the trendline: the distance between open weights and the closed frontier is now measured in months, not generations. Worth a thought experiment about what you'd build differently if frontier-grade capability were simply something you hosted yourself.
Rahul Ilango found a way around a 30-year-old impossibility theorem in cryptography by leaning on Gödel: he builds a non-interactive zero-knowledge proof whose secrecy rests on the unprovable assumption that mathematics is free of contradictions. Because no one can prove that assumption false, no one can prove the protocol leaks anything — so it effectively doesn't. A lovely reminder that the undecidable corners of logic aren't just curiosities; they can be load-bearing.
What stuck with you this week? Reply with a sentence or the name of a piece—or tell me what didn't land. It helps me calibrate.