Welcome, humans.

ICYMI: We just put together a recap of our livestream with Vercel CPO Tom Occhino for a live demo of v0, their vibe-coding tool.

In just 20 minutes, Tom built a fully deployed, interactive financial portfolio tracker from a single prompt. We even built a custom Neuron app live on stream just to prove it works for non-engineers. No wrestling with AWS, just pure vibe-coding. Plus, Tom is such a good hang. We could have talked to him all day TBH.

Check out the full recap to see how Vercel is making building fun again (and learn how to do it yourself).

Here’s what happened in AI today:

• Physical Intelligence launched Robot Olympics for household chores.

• China advanced its AI “Manhattan Project” for making its own AI chips.

• Neurable raised $35M for wearable "brain AI" interfaces.

• Salesforce added 6K enterprise AI customers in one quarter.

Don’t forget: Check out our podcast, The Neuron: AI Explained on Spotify, Apple Podcasts, and YouTube — new episodes air every week on Tuesdays after 2pm PST!

Robot Olympics?! PI’s π0.6 robot takes on doors, socks, keys, and a greasy pan

When you’re scanning social media for robot news every day, you start to get to know the typical robotics comment-section script by heart:

• Robot does something flashy → everyone says “cool, now wash the dishes.”

• Robot tries to wash the dishes → everyone says “not like that.”

Physical Intelligence decided to stop arguing in circles and turn the whole thing into a test. Their new Robot Olympics is basically the Olympics of household chores: same tasks, same constraints, lots of ways to fail, and a painfully honest scorecard for how close we actually are to useful robots.

The whole “Olympics” framing was first pitched in Benjie Holson’s Humanoid Olympic Games (he literally sub-titled the post “gauntlet thrown”) and PI said that part out loud when it launched; gauntlet accepted!

Here’s what PI tested with its π0.6 “generalist” model (a vision-language-action policy—think “LLM for robots,” where vision + instructions turn into motor actions):

• Door entry: navigate a self-closing lever-handle door without getting bodied by the door itself.

• Textiles: turn a sock right-side-out (and admit the gripper is too wide for shirt sleeves).

• Tool use: put a tiny key into a lock and turn it; aka “precision, torque, and no second chances.”

• Cleaning: wash a frying pan with soap and water like a real person who doesn’t want to live in filth.

• And of course, Deformables: open a thin plastic dog poop bag (which conveniently blinds wrist cameras at the worst possible time).

PI says all videos are autonomous; meaning the robot is decomposing the task, making contact-rich moves, and recovering when things go sideways, without a human nudging the sticks between cuts.

Why this matters: Because PI is trying to merge two worlds that usually don’t talk to each other:

• Benchmarks that look like real life (doors, doggy bags, laundry) instead of clean lab puzzles.

• Foundation-model scaling (train big once, then fine-tune for new tasks) instead of bespoke policies for every new object.

That connects directly to PI’s latest research on human-to-robot transfer. The claim is that once you pretrain VLAs (vision language action model) on enough diverse robot experience, they start to “line up” human egocentric video with robot behavior in representation space. Then, you can teach robots from cheap human footage without a ton of explicit alignment tricks.

In the paper, PI reports roughly ~2× improvements on a set of “human-only” generalization scenarios when adding human video during fine-tuning. That’s an early hint that the next data firehose for robots might not be more robot hours… but more humans living their lives on camera.

Of course, there’s a big caveat: “impressive run” is not the same thing as “reliable product.” A few reality checks to keep your expectations from doing parkour:

• These are still brittle, contact-rich tasks where success can swing on lighting, object placement, or a slightly-too-wet sponge.

• Some failures are just hardware: a too-wide gripper loses the shirt-sleeve event no matter how smart the policy is.

• Benchmarks are the beginning, not the end—what matters is repeatability across many trials, in many kitchens.

What to watch next: whether this “foundation model + fine-tune + real-life eval” loop starts to compound the way it did for language models. If it does, the practical timeline gets less sci-fi and more boringly inevitable. If you want the full technical thread in a clean format, PI’s write-up is mirrored as an arXiv HTML paper page.

FROM OUR PARTNERS

You need to add subtitles, resize for Instagram, and clean up background noise. In traditional editors, that's three different menus, seven clicks, and a YouTube tutorial to remember which export settings to use.

Here's what's broken: video editing shouldn't require memorizing 50 buttons and workflows.

Veed lets you edit by typing commands. "Add subtitles in brand colors" → done. "Resize for TikTok" → instant. "Remove background noise" → clean audio in seconds.

The AI handles audio cleanup, subtitle generation, and platform optimization automatically. 10,000+ marketing teams cut editing time by 60% using Veed's command-based workflow.

Start editing with AI free →

Prompt Tip of the Day

Holiday messages fail in two ways: they’re either generic (“Warm wishes…” 🥱) or accidentally too familiar (“Love ya!” 😬). This prompt works because it forces range (12 options) and guardrails (no clichés, no religious language, short length).

The “Merry Message Generator” (cards/texts/emails) 

Write 12 different holiday messages I can send to: coworkers, close friends, extended family, neighbors. 
Mix tones: funny, heartfelt, minimal, and professional. 
No clichés. No religious language unless I ask. 
Keep each under 240 characters.

BONUS: Label each message by audience + tone (e.g., Coworker / Professional), and include 2 that reference a specific gratitude or shared moment using placeholders.

Treats to Try

*Asterisk = from our partners (only the first one!). Advertise to 600K readers here!

1. Lemon Slice-2 turns one photo into an interactive video avatar for your voice agents that can gesture with hands, screen-share to show where to click, or model clothes while guiding shoppers.

2. BU-30B-A3B-Preview from BrowserUse automates browser tasks, clicking buttons, filling forms, navigating sites, and runs hundreds of actions for $1 of compute (open-source).

3. OpenTinker trains reinforcement learning models for LLMs from your laptop—configure your GPU cluster once, then develop experiments locally without reconfiguring infrastructure for each test (code, open-source).

4. Firecrawl /agent searches and navigates websites to gather any data you describe—tell it "find all YC companies from 2024" and it clicks through pages to build datasets that other APIs can't reach (research preview).

5. Principles of Building AI Agents is a book from Mastra that teaches you how to build AI agents, covering prompts, tools, memory, breaking complex tasks into workflows, and connecting agents to knowledge bases with RAG.

   1. Sidenote: I’ve been digging the Mastra AI Agents Hour show lately; they do a great job of rounding up tools (for example, this roundup was heavily influenced by this episode) and news for builders in a weekly show format… Abhi and Shane, if you’re reading, come on The Neuron in 2026!

6. Mercury & Mercury Coder diffusion-based LLMs let you generate text and code via an API. You can paste a file, ask for a refactor, and get an updated version back (raised $50M).

7. EgoX turns a single third-person video into a first-person POV version, so you can take a clip of someone walking and generate what it would look like through their eyes.

8. Autumn sits between Stripe and your app so you can set pricing, usage limits, and entitlements without wrestling with webhooks.

9. Vercel’s AI SDK 6 includes MCP + structured tool loops + reranking so you can ship agentic flows with fewer glue layers.

Around the Horn

1. China’s AI “Manhattan Project” advanced after Reuters reported a prototype EUV lithography system built using parts from older tools and former ASML engineers working under fake names (goal: working chips by 2028).

2. Child exploitation reports tied to genAI “skyrocketed,” as investigators warned AI is accelerating sextortion and synthetic abuse material at scale.

   1. Meanwhile, Child chatbot policies argued teen chatbot use is outrunning governance, producing a patchwork where safety rules depend on your ZIP code.

3. Neurable raised $35M for wearable “brain AI,” pitching hands-free interfaces from intent/attention signals.

4. Meta rolled out “Conversation Focus” for smart glasses and a Spotify feature that picks music based on what you’re looking at.

5. Salesforce quietly added 6K enterprise AI customers in a quarter (48% growth), pushing Agentforce past 18.5K customers and $540M+ in agentic ARR.

FROM OUR PARTNERS

We learned all about Guru's AI source of truth… here's what stood out.

Guru connects your scattered knowledge (Slack, Drive, CRM) into one permission-aware search layer. The killer feature IMO: when info gets outdated, you flag it once and updates sync everywhere instantly. No more AI hallucinations from stale internal data (we wish the AI search tools would implement a similar feature… just saying).

Watch our full walkthrough

Midweek Wisdom

1. Addy Osmani’s workflow argued AI coding works when you do boring engineering first (specs, decomposition, tests) and feed models the exact context they’d otherwise hallucinate.

2. Thomas Wolf’s essay argued execution is easiest to automate, but judgment and agency are the real job moats, and they barely show up in today’s evals.

3. Palona argued reliable agents require vertical focus (workflow data + multimodal inputs + strict guardrails), because one hallucinated deal can torch trust fast.

4. FYI, Claude confirmed it still doesn’t support changing the email on your account (cancel/delete/recreate is the official workaround).

A Cat’s Commentary

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

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