Welcome, humans.

So, there’s this robot in Shanghai named Moya. She smiles, nods, holds eye contact, and walks around on DroidUp’s upgraded Walker 3 skeleton. In her debut video, she already lands deeeeep in the uncanny valley:

Then a later video reveals the part that pushes it over the edge: her silicone skin stays between 32 and 36°C, her gait is supposedly 92% human-like, and her starting price is expected to be around $173K. We are watching them build Westworld in real time, one freakishly warm handshake at a time.

Elsewhere, Moya’s vocally overactive cousin Annie is an expressive robot head mounted on a Unitree body that’s being pitched as a robot pop star. Then there’s 1X who showed NEO’s new hands zipping jackets, pouring tea, installing light bulbs, handling LEGO, and using tools. That feels like the more important milestone TBH. A convincing smile might make a robot easier to talk to, but dexterous hands make it useful.

Not to be left out, humanoid robot company Figure also just dropped a four-year hype reel, and the most clout-chasing aura-farming robots of all time, Atlas, performed the Norway Row (sorry for the loss yesterday Haaland and the rest of the Norway crew!).

So anyway, the robot industry now has engineers, actors, pop stars, and football ultras. All it really needs now is a reality show.

Here’s what happened in AI today:

• 😹 Apple sued OpenAI and io Products over alleged trade-secret theft

• 📰 Meta suspended an Instagram AI image feature after public backlash.

• 📰 Stanford researchers introduced Biomni, a biomedical co-scientist agent.

• 🍪 GitHub Spec Kit and Railway Agent gave builders more structure.

• 📰 Boko Haram apparently has used frontier AI for propaganda and much worse.

...and a whole lot more that you can read about here.

Hey: Want to reach 700,000+ AI-hungry readers? Advertise with us! 

Love robots? We just launched a robotics newsletter! Sign up for it here.

😹 Apple Is Suing OpenAI Over Its Hardware Ambitions

Apple and OpenAI went from "ChatGPT on your iPhone" to "see you in court" in roughly the time it takes most people to find the Settings app.

The fight is about OpenAI's hardware push, the Jony Ive-linked device effort, and whether former Apple employees brought more than experience with them when they left Cupertino. Apple sued OpenAI, io Products, and several former Apple employees, accusing them of using Apple trade secrets to speed up OpenAI's move into consumer devices.

Here's what happened:

• Apple's complaint names former Apple design VP Tang Tan, former iPhone engineer Chang Liu, and OpenAI's newly acquired hardware arm.

• Apple claims Liu exploited an authentication bug after leaving Apple to access confidential files from an Apple-issued laptop.

• Apple alleges OpenAI used insider supplier terminology, asked targeted component questions, and received Apple files including circuit-board manufacturing documents.

OpenAI's response was short: the company said it has "no interest" in other companies' trade secrets. Which is exactly the kind of sentence lawyers write when everyone in the room has suddenly become very interested in discovery.

Why this matters: OpenAI's next phase is not only about smarter models. It wants to own more of the interface people use every day: apps, browsers, agents, and eventually devices. Apple built its whole empire by controlling that interface, from chip to glass to supply chain. If Apple can slow or constrain OpenAI's hardware effort, this becomes more than an HR poaching fight. It becomes a platform war.

Our take: The weirdest part is that Apple and OpenAI are still partners on Apple Intelligence. So the same company helping put ChatGPT on iPhones is now accusing OpenAI of building its future device business on stolen Apple know-how.

That is the real tension: OpenAI wants to become less dependent on Apple devices, while Apple wants AI partners that do not eventually become hardware rivals. The lawsuit may decide whether OpenAI's first real device launches as a breakthrough product, a legal exhibit, or both.

FROM OUR PARTNERS

Your AI lab just got a desk address

Most AI experiments start in a browser tab. The serious ones need more room.

Dell Pro Max with GB10, powered by the NVIDIA GB10 Grace Blackwell Superchip, gives builders a compact desktop for AI development with 128GB memory, NVIDIA DGX OS 7, and 4TB storage.

Use it to test agents, prototype workflows, and run model experiments without turning every idea into another cloud line item.

For founders, technical teams, and AI-obsessed operators, this is the “let’s actually build it” machine.

Explore Dell Pro Max with GB10.

🎓 AI Skill of the Day: Make the Smart Model Plan, Then Let the Cheap Model Build

Most people waste their best model on every step of a task. A better setup is to use it only where judgment matters, then hand the mechanical work to something faster and cheaper.

Designer Emil Kowalski showed this with an animation-auditing skill: a strong model reviews an entire codebase, scores the animations across eight criteria, and writes a prioritized repair plan without touching the code. Cheaper agents can then execute the plan.

Use the same planner → builder → reviewer loop for almost any project:

1. Give your strongest model the goal, files, and constraints. Tell it to inspect and plan, not execute.

2. Ask it to rank the problems and write self-contained instructions for each fix.

3. Send those instructions to a cheaper model one task at a time.

4. Return the finished work to the strong model for a final quality check.

Act as the senior reviewer and planner. Inspect the project against the goal below, but do not edit anything. Identify the highest-impact problems, rank them by priority, and write self-contained implementation instructions that a cheaper model can follow one task at a time. Include success criteria for each task and a final QA checklist.

Goal: [PASTE YOUR GOAL]
Constraints: [PASTE YOUR CONSTRAINTS]

The expensive model becomes your architect, not your intern.

Want more tips like this? Check out our AI Skill of the Day Digest for July.

Have a specific skill you want to learn? Request it here. 

🍪 Treats to Try

*Asterisk = from our partners (only the first one!). Advertise to 700K+ readers here!

1. *The #1 AI Service Desk for MSPs. Thread is the AI Service Desk 750+ MSPs trust to automate triage, dispatch, and client conversations across ConnectWise, Autotask, and HaloPSA, processing 173 million tickets, returning over half a million hours to technicians, and earning 97% positive sentiment from end clients along the way. Get Started.

2. GitHub Spec Kit gives you a spec-first workflow for coding agents, forcing requirements, clarification, planning, and task breakdown before implementation —free/open-source.

3. Railway Agent now works from Slack and Discord, with CLI usage controls so teams can manage agent and workspace access without living inside the dashboard.

4. Claude Code's in-app browser gives desktop users a sandboxed way to open docs, inspect websites, and test local app flows without leaving the coding agent —paid Claude plans.

5. Microsoft Research Flint helps agents turn compact chart specs into polished visualizations —free/open-source.

6. Colibri streams GLM-5.2 experts from disk so a consumer machine can run a massive open model with roughly 25GB of RAM —free/open-source.

7. ChatCut edits video from natural-language instructions while keeping cuts, captions, graphics, music, and generated media editable on a real timeline —free plan, then $25/mo.

8. Bono AI turns one spoken conversation into blog posts, newsletters, social posts, and a wider content plan in your voice —free plan, then $30/mo.

9. ConnectMachine scans business cards and badges, remembers how you met each person, and lets you search your network conversationally —free plan, then $5.99/mo.

New from The Neuron: AI Explained

New episodes air every week on Wednesdays: Spotify | Apple Podcasts | YouTube 

📰 Around the Horn

• Meta suspended Muse Image on Instagram after users and talent groups objected to likeness generation from public accounts, with Variety reporting the feature was discontinued after days of backlash.

• Stanford researchers introduced Biomni, a biomedical co-scientist agent that can read literature, choose tools and datasets, write code, interpret results, and propose experiments.

• Cambridge's CASP and The New York Times reported that Boko Haram has used frontier AI for propaganda, bomb construction, and attack planning.

• GitHub CodeQL added prompt-injection detection for JavaScript and TypeScript flows into OpenAI, Anthropic, and Google GenAI SDKs.

• AI agents helped find an Ethereum validator bug, but humans still had to verify, prove, and patch the issue before it could take validators offline.

• Ed Sim, Benedict Evans, and Pangram Labs mapped three less flashy AI pressures: proprietary data moats, token-price compression, and AI-written social feeds.

FROM OUR PARTNERS

Your employees are connecting AI to everything. Now what?

ChatGPT and Claude aren't just answering questions. Employees are connecting them directly to Notion, Linear, Jira, and the rest of your stack — with no security visibility into what data moves or what actions they take.

Harmonic Security gives your team the visibility to control it.

See what's connected

🌟 Sunday Special: The Week's Top 5 Stories and Tools

The five stories readers should remember:

1. OpenAI turned ChatGPT into a work super-app. GPT-5.6 landed alongside ChatGPT Work, pulling browsing, files, connected apps, scheduling, deliverables, and Codex-y desktop workflows into one increasingly everything-shaped ChatGPT.

2. Cloudflare put a bouncer on the AI web. Its new crawler controls split search, agent, and training bots into separate permissions, which made the publisher-vs-model-lab fight feel less like theory and more like a settings page.

3. Anthropic found Claude's hidden J-space. The research suggested Claude has an internal scratchpad-like workspace for holding and editing concepts before answering, which is the kind of model-inspection work that could matter a lot as agents get harder to supervise.

4. A rogue Dialogflow agent showed where enterprise AI security breaks. Varonis' Google Dialogflow CX disclosure was not about a chatbot being too clever. It was about agent plumbing, permissions, and runtime isolation becoming the new front door.

5. GPT-Live made voice feel like the next real interface. OpenAI's full-duplex ChatGPT Voice can listen while it talks, which makes tutoring, role-play, translation, and live brainstorming feel much closer to an always-on interface than a novelty mic button.

The five tools worth carrying into next week:

1. ChatGPT Work is the big one: a GPT-5.6-powered work agent for browsing, connected apps, files, deliverables, scheduling, and desktop/mobile continuity.

2. Claude Cowork keeps remote agent sessions running across desktop, web, and mobile, so longer Claude work does not have to live in one fragile tab.

3. NOX pulls iMessage, WhatsApp, Slack, email, and more into one Mac inbox, then drafts replies in your voice. Very normal-person useful.

4. Willow Frontier Mini gives you unlimited voice dictation for turning speech into clean text, which is one of those small upgrades that can quietly change how much work you actually capture.

5. Framer's AI agent can design and edit a website live on the canvas, with changes triggered from Slack or GitHub too. Useful if your "quick landing page" keeps becoming a three-week emotional support project.

A Cat’s Commentary

P.S: Before you go… have you subscribed to our YouTube Channel? If not, can you?

P.P.S: Love the newsletter, but only want to get it once per week? Don’t unsubscribe—update your preferences here.