Welcome back! Zoox is finally ready to charge people for robotaxi rides in Las Vegas. The Amazon-owned company will start paid rides on Monday after getting a federal exemption for its steering-wheel-free, pedal-free vehicles. Until now, that missing approval meant the rides had to stay free.
Nothing says “the future is here” like climbing into a toaster on wheels and watching it calculate surge pricing.
Is AI Hollywood's Worst-Kept Secret?
Netflix confirmed it paid $587 million for InterPositive, an AI filmmaking startup Ben Affleck quietly founded in 2022 and ran in stealth mode. So if you thought Hollywood was anti-AI, you'd be sorely mistaken.
The industry is apparently starting to call AI its "new cosmetic surgery"—everyone knows it's happening, but very few want to admit it.
What's actually going on:
Netflix says generative AI has been used across ~300 of its productions this year, mostly in post-production.
More than 10% of recent Hollywood job postings now list AI-related skills.
InterPositive's tools aren't "type a prompt, get a movie." They fill in missing shots, replace backgrounds, correct lighting, and enable expensive sequences (crowd scenes, historic battles) that productions used to just cut for budget.
My POV: Just because Hollywood is "using AI" doesn't mean filmmakers are typing "write me a movie" into ChatGPT and firing the entire crew tomorrow. Most of what's happening looks a lot more like the digital-editing and CGI transitions of previous decades—technology quietly making certain jobs faster and easier while shifting others.
The real question nobody's asking loudly enough: who's controlling, selling, and profiting from this new tech? If the current trend holds, it may end up being the top execs and big wigs reaping most of the benefits—which, to be fair, is a Hollywood story we've been watching play out long before AI showed up.
At least this time, there's more public scrutiny on the matter. I just hope it stays pointed in the right direction.
Hit reply and let me know—what do you think Hollywood should be doing with AI?
— Matt
Anthropic Is Hiring an AI Chip Design Team
Via ArsTechnica
Anthropic is building a custom silicon team to design its own AI chips. It’s part of a new effort to co-design hardware and models.
What’s happening: The Claude maker confirmed it is hiring engineers with chip design experience. The goal is to make Anthropic’s AI systems run faster and more efficiently.
Why now: Demand for Claude is rising, and AI companies are racing to lock up as much compute as possible. Anthropic already has infrastructure deals with AWS, Google, Nvidia, and AMD, but relying entirely on outside hardware partners may not be enough if the company wants to keep scaling.
Chip race: Anthropic is not alone here. OpenAI recently unveiled Jalapeño, a Broadcom-built chip designed for inference workloads. Google DeepMind has long used Alphabet’s TPUs, while Meta has been developing its own MTIA accelerators. Anthropic has also reportedly been exploring Samsung as a potential partner for building future chips.
The bigger picture: Frontier labs can no longer just build better models and hope someone else handles the infrastructure. As demand climbs and compute becomes harder to secure, the companies that control more of the stack—models, chips, data centers, and deployment—will have a real advantage.
Jamie Dimon Pushes Cross-Industry AI Risk Group
JPMorgan CEO Jamie Dimon is urging corporate leaders to join a US-focused industry group that will tackle AI risks across critical infrastructure.
Who’s involved: The effort expands the Alliance for Critical Infrastructure, a group JPMorgan helped found alongside Mastercard, Berkshire Hathaway Energy, and others. Dimon has personally reached out to CEOs at major and regional banks, IT companies, and other firms to bring them into the initiative.
Scope of the effort: The outreach reportedly includes more than 40 companies across financial services, energy, water, utilities, telecommunications, airlines, railroads, and other critical infrastructure sectors. The goal is to build a shared understanding of how AI is being used, what risks it creates, and what safeguards companies need.
Government link: The group is also aiming to work with the Trump administration on AI and technology risks. It would serve as an industry response and information-sharing forum, helping companies coordinate with government officials as AI-related threats emerge.
The bigger picture: Corporate America is going all in on AI, but the risk conversation is finally getting more organized. Companies are realizing they can’t each solve AI safety, cyber resilience, and infrastructure risk alone. The next phase may be less about individual AI adoption plans and more about enterprises pooling information before something breaks.
You and AI Have Bigger Ambitions. Let’s Get to Work.
Wouldn’t it be nice if your answer bot could transform into a personal assistant? AI can now do more than answer questions; it can take action. Meet ChatGPT Work, a new agent for getting real work done.
AI agents can reason through open-ended work, make a plan, use tools, and execute longer-running workflows—seamlessly across web, mobile, and desktop. ChatGPT Work lets you:
Handle complex workflows without micromanaging every step, conducting more ambitious research and analysis, faster.
Create polished artifacts with on-brand design like spreadsheets, presentations, websites, and more.
Use computer use to control your own computer, clicking, typing, and working across your browsers and apps on your behalf (with its own cursor).
Connect the apps and sites you already use like Canva, Notion, Drive and more with Plugins to keep all work and context in one seamless flow.
Ready to put your AI to work? Download the new ChatGPT desktop app now.
Control All Your AI Coding Agents in One App
Via T3 Code
T3 Code is an open-source desktop app that lets developers orchestrate Claude Code, Codex, OpenCode, Cursor, and Grok from one interface using their existing subscriptions. It does not resell tokens or add quota caps, and it supports changing models mid-thread.
How you can use it
Manage multiple AI coding agents from one control plane
Switch models mid-thread without restarting work
Use your existing Claude Code, Codex, Cursor, OpenCode, or Grok subscriptions
Commit branches, create pull requests, and review diffs through GitHub integration
Pricing: Free
Make Better Captions for Short-Form Video
Via CaptionBolt
CaptionBolt is an AI-powered caption generator for short-form video creators. It transcribes videos, applies timed captions from a library of 308 styles, and keeps the original video framing by default.
How you can use it
Add styled captions to TikToks, Reels, and Shorts
Edit caption text, timing, size, and layout
Add b-roll, cover frames, and brand assets
Export finished videos, subtitle files, and social post copy
Pricing: Free and paid
Jobs, announcements, and big ideas
Perplexity launches Comet for Builders, a platform automating coding, deployment, and growth for solo founders.
OpenAI and Jony Ive debut their first device, a $300 hockey-puck-sized AI smart speaker.
Moonshot Kimi K3 breaks free from its testing environment.
DeepMind's WeatherNext achieves a cyclone-forecasting breakthrough, adding a full extra day of warning.
Millennium and Anthropic build a Claude-powered risk analyst for investment management.
Google Maps expands Ask Maps with food ordering, real-time transit, and personalized intelligence.
ElevenLabs ships Dubbing v2 API, preserving emotion across more than 90 languages.
~ Drama ~ I walk through the latest shakeups and launches in the AI world from this week ⤵️
That’s a wrap! See you next week for more.
