Is Genie 4 publicly available today?

Not broadly. DeepMind describes a research preview for trusted testers first, with later access planned for Google AI Ultra subscribers via Project Genie and early-access Gemini API / Google AI Studio partners. A waitlist is mentioned.

Did OpenAI prove the Exponential Time Hypothesis is false?

OpenAI released AI-produced claims and Lean formalizations framed as raw research data for community examination — including a claim that ETH is false. An external group including ETH originators is examining the work. Treat it as under review, not settled consensus.

What do we know about Anthropic’s S-1 financials so far?

Anthropic filed an S-1 intending Nasdaq ticker ANT; share count and price remain TBD. Day-after reporting (e.g. SF Examiner) attributes June-quarter sales and operating-profit figures to the prospectus. IPO timing chatter remains reported rumor, not an Anthropic confirmation.

AI Daily · Oct 28, 2026 — Genie 4, ETH claim, Anthropic S-1

World models — Google DeepMind announces Genie 4

DeepMind published Genie 4, framed as a new generation of world model aimed at more realistic real-time interactive simulation. The headline shift versus earlier Genie work is multimodality baked into the world loop: ambient sound, music-like audio, and interactive dialogue generated alongside the visual scene — not bolted on afterward.

Architecture notes in the blog lean on the Gemini family: Gemini Omni for real-world appearance and behavior, other Gemini models for coherence and continuity, and Gemini 3.8 Flash described as a guiding agent-in-the-loop for multi-step events. While a session runs, testers can steer with natural language — change weather, move the camera, insert people or objects — without restarting the simulation.

DeepMind also claims improved physics and human-action realism, and positions Genie 4 as the first real-time Genie with native audio plus interactive dialogue. Safety tooling called out in the post includes guardrails and SynthID watermarking on generated video and audio.

Access path: research preview for trusted testers first; later intended for Google AI Ultra subscribers through Project Genie, plus early-access Gemini API / Google AI Studio partners. A waitlist is mentioned — treat “coming” as gated rollout, not a general public download today.

Known soft spots (DeepMind’s own list): object permanence and visual drift over longer sessions, complex physics, legible on-screen text, and spatial accuracy remain active research areas. That matters if you hoped to use world models as a drop-in replacement for game engines or robotics digital twins without human review.

Why it matters here: testers already explored pre-visualization, historical-photo → simulation, game prototyping, and robotics / enterprise scenario sims. If you evaluate interactive AI products on our chatbot scorecard, Genie 4 is a different product class — score it on controllability, session length before drift, watermarking, and access gates, not on chat benchmarks.

Research claims — OpenAI’s AI-produced science under academic shepherding

On October 26, OpenAI published a post on building scientific understanding with academic communities. It remains the major science story for this week. The framing is important: three AI-produced results in math, computer science, and physics are presented as raw research data that still need community understanding — not as settled theorems stamped by the field.

Math: models produced a proof of the local smoothing conjecture in every dimension, which would imply Kakeya and Stein restriction in all dimensions. Natural-language exposition and Lean formalization were released.

Computer science: a claim that the Exponential Time Hypothesis (ETH) is false, with Lean formalization released. An external examination group led by Impagliazzo, Paturi, and Williams — ETH’s originators plus Williams — is reviewing the work. Natural language and Lean artifacts were made public on GitHub.

Physics: existence proofs claimed for certain interacting quantum field theories including QCD and a Standard Model completion; Yang–Mills existence formalized. A shepherd group led by David Gross and others is engaged; physics artifacts are being held back pending that group’s process.

Buyer / builder filter: these are claims under independent examination. Do not rewrite product roadmaps, crypto hardness assumptions, or “ETH is dead so we can ship X” narratives until shepherds publish verdicts. Lean artifacts raise the bar for verification — they do not skip peer scrutiny.

Capital markets — Anthropic S-1 day-after: reading the tables, not the memes

Yesterday’s October 27 briefing covered Anthropic’s announcement that it filed a Form S-1 for a proposed IPO of Class A common stock, intending Nasdaq Global Select listing under ANT, with share count and price still blank, and Goldman / Morgan Stanley prospectus desks named for distribution. Today’s angle is the financial follow-through in reputable reporting — not a second announcement from Anthropic.

SF Examiner (Troy Wolverton, Oct 27) attributes prospectus-linked figures including roughly $11.6B in June-quarter sales (up about 144% from the March quarter), operating profit around $974M in Q2 2026 after prior operating losses, large reported net losses driven largely by convertible-debt fair-value accounting, and roughly $3.4B of cash generated by operations in H1 2026 after capex. The same piece cites The Information on IPO timing chatter around the week of November 16 — label that as reported speculation, not an Anthropic confirmation.

Fortune’s Oct 27 coverage leaned into prospectus language themes (including “age of miracles” / existential-risk framing). Useful for tone; not a substitute for the financial tables.

What we deliberately skip: circulating “$100B raise / multi-trillion valuation” digest headlines without a primary S-1 table citation. Prefer Examiner-attributed operating figures and keep valuation rumors soft or out.

Why it matters on this site: Anthropic already ranks near the top of our AI chatbot scorecard for careful writing and privacy posture. Better disclosure helps reviewers; ticker narrative does not change model quality overnight. Keep product evaluation on rubrics — see Reviews — while the roadshow noise spikes.

Editor takeaways (actionable)

Genie 4: put it on the interactive-sim shortlist for pre-viz and game prototypes — after you confirm access tier, session-length drift, and SynthID requirements for your pipeline.
OpenAI science posts: bookmark the Lean repos and wait for shepherd write-ups. Do not treat “ETH false” as a product or cryptography roadmap input yet.
Anthropic S-1: use attributed operating metrics for diligence; ignore valuation gossip without a table cite. Yesterday’s “filed, not listed” rule still holds.
Site housekeeping: keep scoring tools and chatbots on published rubrics (Tools, Reviews) instead of chasing every launch into a new #1.

Primary sources

DeepMind blog — Genie 4: a new generation of world simulation
OpenAI — Building scientific understanding with academic communities (Oct 26, 2026)
Anthropic announcement — Files S-1 for proposed IPO
SF Examiner reporting (Troy Wolverton, Oct 27) — day-after S-1 financial attribution; IPO-timing chatter cited from The Information (reported, not Anthropic-confirmed)
Fortune (Oct 27) — prospectus theme coverage (optional context on language / risk framing)

Topic selection may overlap other AI digests on the same calendar day. Wording, sectioning, and recommendations here are written for wordok.top. Prior issue: 2026-10-27.