Kimi's AI Cuts Its Own Teaser

PLUS: OpenAI's screenless AI speaker fuels an Apple lawsuit, and DeepMind's Hassabis pushes for a Wall Street-style AI watchdog

Author

Gen AI Art
July 18, 2026 • Estimated Reading Time: 3 minutes

Sponsored by

Moonshot AI just released the largest open-weight model anyone has built — and it didn’t stop at benchmarks. Kimi K3 spent its debut editing its own teaser trailer from 56 raw clips, syncing every cut to the beat.

Chinese labs keep closing gaps analysts said wouldn’t close until 2027, and this time the proof is a machine that can direct video as well as it writes code. Does the U.S. still hold the technical lead it assumes it does?

Today in AI:

  • Moonshot’s Kimi K3 becomes the largest open model

  • OpenAI’s screenless AI speaker fuels Apple’s lawsuit

  • DeepMind’s Hassabis pitches a Wall Street-style AI watchdog

How Jennifer Aniston’s LolaVie brand grew sales 40% with CTV ads

The DTC beauty category is crowded. To break through, Jennifer Aniston’s brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category.

What’s new? Moonshot AI unveiled Kimi K3, a 2.8-trillion-parameter open-weight model the company calls the world’s first open “3T-class” AI — one that spent its own launch editing a 56-clip teaser video with frame-accurate beat syncing.

What matters?

  • The model runs on a new Kimi Delta Attention architecture paired with “Attention Residuals,” which Moonshot says delivers roughly 2.5x better scaling efficiency than its predecessor, Kimi K2.

  • Beyond code, K3 handles motion design and animation — building a 3Blue1Brown-style motion-graphics explainer of its own architecture and generating playable 3D scenes from a single image.

  • Fortune reports K3 performs competitively with Anthropic’s Fable 5 and beats GPT 5.6 Sol on several benchmarks — and it’s already free to try today at kimi.com, with full weights landing July 27.

Why it matters?

A model this large usually means a coding tool, not a creative collaborator — K3 blurs that line by treating video, motion graphics, and code as one skill. Full weights arrive in under two weeks, so anyone curious can put its editing chops to the test well before then.

GUIDE

What’s new? OpenAI is building a screenless, mobile smart speaker designed to move on its own and act as a “physical manifestation” of ChatGPT, TechCrunch reports — a device now caught in a trade-secrets lawsuit from Apple.

What matters?

  • The speaker has no screen, includes moving mechanical parts meant to make it “feel alive,” and can draw on a user’s emails to get proactive.

  • Apple’s lawsuit accuses OpenAI of building its hardware business on “misappropriated trade secrets” from ex-Apple staff; OpenAI says it has “no interest in other companies’ trade secrets.”

  • Former Apple design chief Jony Ive sold his studio io to OpenAI for $6.5 billion and has been developing the device with the company for roughly two years.

Why it matters?

Design pedigree alone won’t settle whether people want an always-listening speaker with a camera and email access sitting in their kitchen. The lawsuit now adds a second open question: whether OpenAI’s first hardware product can even ship on schedule.

SPONSORED BY ATTIO

Introducing The First Agentic CRM

Get revenue agents, workflows, and automations across every stage of your motion. Access customer data in real time through Attio's web app, MCP, API, and SDK.

Then Ask Attio anything about your business and get instant answers.

It's the CRM that runs the work behind every win.

What’s new? Google DeepMind CEO Demis Hassabis called for a U.S.-led independent AI standards body modeled on Wall Street’s FINRA, arguing frontier AI needs a formal safety check before the next wave of models ships.

What matters?

  • Under the proposal, frontier labs would submit new models to the body 30 days before release for testing on cyber, biological, and “deception” risks.

  • The board would be industry-funded but majority-independent, stacked with Turing Award winners alongside government and open-source representatives.

  • Hassabis wants the rules to apply to every frontier-class model, regardless of country of origin or whether it’s open or closed.

Why it matters?

A voluntary review board sounds modest, but Hassabis says formal requirements “could quickly follow” once the process proves itself. His timeline — a working body before year’s end — leaves little room to referee the kind of frontier leap Kimi K3 just made days later.

Everything else in AI

Google renamed NotebookLM to Gemini Notebook, adding a secure cloud code-execution environment for Pro users and syncing notebooks across the Gemini app and Search.

ChatGPT shows how to turn your calendar, emails, and docs into a daily work brief — one manual pass, then a scheduled task to keep it running every morning.

God of Prompt packages prompt libraries, templates, and tutorials into one bundle for turning ChatGPT and Midjourney skills into a side income.

Essential AI Guides - Reading List:

Let us know!

Work with us

Reach 100k+ engaged Tech Professionals, Engineers, Managers and decision makers. Join brands like MorningBrew, HubSpot, Prezi, Nike, Ahref, Roku, 1440, Superhuman, and others in showcasing your product to our audience. Get in touch now →