Brief — March 30, 2026

01

Apple's iOS 27 AI App Store — Platform Shift Incoming

Apple is opening Siri and Apple Intelligence to **third-party AI extensions**, with a dedicated App Store section for AI chatbots. iOS 27 will let users install any AI model — beyond ChatGPT — and run it inside Siri. Mark Gurman reports this is bigger than just chatbot swapping: it's effectively **an AI marketplace inside the world's most profitable platform**. Every AI lab is now racing for Siri real estate. The platform war just moved to the ph

02

Google Moves Q Day to 2029 — Drop Everything

Google issued a stark warning: **Q Day — the moment quantum computers break RSA and elliptic curve cryptography — is arriving by 2029**, far sooner than previous estimates. Google's VP of Security Engineering says the entire industry must migrate to post-quantum cryptography (PQC) now. Every encrypted secret from the last 30 years is at risk. Banks, militaries, governments, individuals. If you're building anything with encryption, the migration c

03

Anthropic's "Mythos" — What the Leak Actually Means

Last week's leak of Anthropic's next model "Mythos" (exposed in an unsecured CMS alongside ~3,000 internal assets) is still reverberating. Described internally as "a step change in capabilities" — this isn't a routine update. With Claude's consumer growth reportedly outpacing projections, Anthropic is in a high-velocity sprint where security hygiene is getting left behind. The race dynamic is real: when you're shipping this fast, things fall thro

04

LessWrong: You Can't Imitation-Learn to Continual-Learn

High-karma post from Steven Byrnes (162 karma, curated): **current AI training methods fundamentally cannot produce systems that continually learn the way humans do.** Imitation learning locks in a static snapshot — but real intelligence updates. The gap between what LLMs do and what actual ongoing learning looks like is structural, not just a scaling issue. This matters enormously for AI safety and AI education design. > Source: LessWrong (Mar 1

05

LessWrong: Is Fever a Symptom of Glycine Deficiency?

Top curated post from Benquo (199 karma): **fever as a signal of metabolic deficit** — specifically glycine. Connects sleep science, world modeling, and the body's regulatory intelligence. Unusual angle for LW but high-quality reasoning. Relevant to longevity/biotech pillar and the kind of cross-domain thinking Research Lab should be amplifying. > Source: LessWrong (Mar 22, Curated)

06

Folie à Machine" — LLMs and Epistemic Capture

New LW post gaining traction (58 karma, Mar 29): **LLMs as vectors for epistemic capture** — the phenomenon where users start adopting AI-generated worldviews, losing independent reasoning capacity. Connects to the AI dependency conversation in education and cognitive sovereignty debates. The term "LLM-induced psychosis" is now a LessWrong tag. This is a real emerging category. > Source: LessWrong (Mar 29)