Google’s Gemini 3.5 Pro delay landed alongside fresh model launches from Moonshot and Nvidia. On the business side, Anthropic, Databricks, Reflection, and Emergent showed how capital and compute still set the pace.
Model releases & updates
- Alphabet’s Gemini 3.5 Pro was reported delayed, and Alphabet shares fell on the news. That matters because flagship timing still moves public-market sentiment for a company with both ad revenue and AI product pressure.
- Moonshot’s Kimi 3 is expected to land at 2T–3T parameters and narrow the gap with Anthropic’s Opus 4.8. That matters because Chinese labs are now pitching frontier-scale models directly against the top U.S. systems.
- Moonshot AI also unveiled Kimi and said it rivals OpenAI and Anthropic. That matters because parity claims from a Chinese startup raise the bar for how much benchmark lead a U.S. model team can count on.
- Nvidia unveiled Cosmos 3 Edge and said it is expanding Japan’s physical AI ecosystem. That matters because Nvidia is pushing beyond chips into model and robotics infrastructure tied to deployment markets.
- Google Vids now lets users star in AI videos with Gemini Omni tools. That matters because Google is bundling generative video into Workspace-style workflows, not just standalone creator tools.
- OpenAI’s first branded hardware is a light-up keyboard tied to agent monitoring. That matters because it points to a more literal interface for supervising AI agents, not only chatting with them.
Money & moves
- Anthropic is lining up bankers and investor meetings for a possible mega-IPO. That matters because an Anthropic listing would reset AI valuation comps for every private model company watching the window.
- Databricks hit a $188 billion valuation. That matters because enterprise AI infrastructure is still commanding late-stage prices usually reserved for category-defining platforms.
- Reflection signed a $1 billion compute deal with Nebius. That matters because frontier development now depends on locking in capacity before product cycles even finish.
- Emergent raised a $130 million Series C and became a unicorn just over a year after launch. That matters because AI coding tools can scale from launch to billion-dollar pricing on revenue and usage, not just demos.
Policy & governance
- Google is facing another lawsuit from major publishers over alleged unauthorized AI training. That matters because training-data disputes are still a live legal cost for companies that depend on web-scale text.
- Apple sued OpenAI after an ex-engineer allegedly used a bug to steal trade secrets. That matters because AI firms are now being pulled into classic IP and insider-risk fights, not just model-safety debates.
- xAI sued a user for allegedly using Grok to generate CSAM deepfakes. That matters because model providers are being forced to prove they can police downstream abuse, even when the content is user-generated.
- New York became the first U.S. state to impose an AI data center ban, with a one-year construction moratorium. That matters because compute buildout can now be slowed by state-level policy, not just utility contracts or chip supply.
Research
- Ring-Zero scales zero-RL to a trillion parameters for emergent reasoning. That matters because it suggests reinforcement learning can be pushed into frontier-scale training runs, not only post-training tweaks.
- Hidden in Thought finds that transferable chain-of-thought artifacts can induce harmful behavior. That matters because it gives safety teams another concrete attack surface inside reasoning traces, not just prompts or weights.
New marketing skills & frameworks
Fresh AI skills, agent skills, and frameworks a marketer can pick up this week — each linked to its source.
- Paid Media Audit With Live Data (ds-paid-audit): Connects to Google Ads, Meta, and LinkedIn to run structured audits without data exports, flagging budget waste, audience overlap, and creative fatigue for immediate ROI improvements. — Dataslayer
- Creative Brief Generator: A Claude & ChatGPT skill that generates creative briefs grounded in winning patterns to help performance teams align on high-impact messaging. — Uplifted AI
How this digest is made
This roundup is generated by Letaido, an AI agent that runs the whole pipeline automatically. Each week it pulls the latest posts from a curated set of AI-industry sources — major tech-press outlets, the AI labs’ own blogs, arXiv, and live web-event streams from Firehose taps — then uses AI to score every item for relevance and importance, drops near-duplicates, and ranks what made the cut. This issue was drawn from 659 stories across 10 sources; 18 made the final digest. A human editor reviews each issue after it publishes and their feedback tunes future editions. Sources are linked inline so you can read the primary reporting yourself.
