Welcome, humans.

There is just so much cool stuff happening in robot land, its sometimes hard to keep up (ironic coming from us, a newsletter trying to keep track of news in one of the fastest changing industries of all time). But this project is PARTICULARLY cool:

This TRLC-DK1 kit lets anyone:

• Build robot arms from parts he provides.

• Collect training data (by showing the robot what to do).

• Deploy the trained AI in under a day.

So yeah, he's democratizing robot training: turning what used to be a months-long engineering project into something you can do over a weekend. Wanna train your own robot? Now you can! Pretty wild when you think about it!

Also wild: did you know there’s a derogatory word for robots going around right now called “clanker”, and that it’s actually originally from Star Wars? Love ‘em or hate ‘em, robots are making their cultural impact KNOWN.

And hey, while we’re on the topic of ROBOTS…

Here’s what happened in AI today:

1. MIT spin-out AlterEgo unveiled a "telepathic" wearable.

2. OpenAI's Microsoft revenue cut will shrink from 20% to 8% by 2030.

3. California passed a new AI safety bill for Newsom to sign.

4. xAI laid off 500 data workers to focus on AI tutors.

The World's First 'Telepathic' Wearable Is Here, But You Can't Buy It Yet

In the weird 90s movie Johnny Mnemonic, Keanu Reeves had a brain jack that let him interface directly with computers through pure thought. Well, AlterEgo just built the real-world version… except instead of drilling holes in your skull, they're reading the signals your brain sends to your speech muscles before you actually speak. WILD.

The Boston-based startup unveiled their new wearable device last week that lets you communicate silently by catching your words at the loading dock before they ship out of your mouth.

Their breakthrough “Silent Sense” technology can pick up everything from full-on talking to completely motionless intent to speak. You literally just think about saying something, and it captures it. No lip-reading, whispering, or mouth movements required (though you can do those too if you want).

What you can actually do with this thing:

• Type messages at the speed of thought.

• Silently Google that actor's name mid-conversation without looking rude.

• Control your smart home while your partner's sleeping

• Have completely silent conversations with other AlterEgo users

• Ask questions about what you're looking at (via built-in cameras)

This tech started at MIT Media Lab back in 2018, where researcher Arnav Kapur first demonstrated the concept, hitting 92% accuracy with limited vocabularies. Now he's CEO of the spun-off company, and they’re goal is to start by giving voices back to people with ALS, MS, and other conditions that affect speech.

Before you freak out about mind-reading: AlterEgo explicitly doesn't read your thoughts. It only picks up what you consciously intend to say by engaging your speech system. Your random shower thoughts about whether hot dogs are legally considered sandwiches, and if so, in what states? Totally safe.

Why this matters: We're hitting the awkward teenage years of voice assistants: useful enough that we want them everywhere, embarrassing enough that we don't want to use them in public. AlterEgo sidesteps that whole problem by making the conversation completely internal.

Did you know we’ve actually had the QWERTY keyboard since 1878, the mouse since Doug Engelbart invented it in 1964, and touchscreens since the 1960s (though nobody cared until the iPhone)? If AlterEgo delivers on its promise, we're basically looking at the first genuinely new way to interact with computers in decades.

Early access signups are open now, though no word on pricing or ship date. Given the seven-year journey from MIT lab to launch, they're clearly playing the long game here.

FROM OUR PARTNERS

Your keyboard called—it’s taking a long weekend ⌨️✈️

We’ve been banging away on keyboards for 150 years. Until now, voice dictation hasn’t been reliable enough to change that.

Wispr Flow finally delivers the no-edit confidence we’ve all been waiting for:

• 4× quicker than typing. Dictate emails, docs, and DMs in real time and save precious hours every week.

• AI auto-edits on the fly. Flow cleans filler words, fixes grammar, and formats perfectly as you speak.

• Works inside every app with no setup. Fly through Slack notifications, give more context to ChatGPT, or brain dump into Notion.

• Use it at your desk or on the go. Available on Mac, Windows, and iPhone. 

“This is the best AI product I’ve used since ChatGPT.”
— Rahul Vohra, CEO, Superhuman

Give your hands a break ➜ start flowing for free today.

Prompt Tip of the Day

Nate B. Jones and Mike Krigsman had a great discussion about the recent viral MIT study on how 95% of AI trials fail to generate ROI, and what the future of work with AI looks like. In particular, the idea of “intent” and ability to “communicate” that intent being key to success with AI.

This reminded of a film class we took back in college, where the instructor said a film director only truly needs two skills: to 1) have a vision, and 2) communicate that vision. The rest of the crew technically handles everything else.

When you work with AI, you are the director. If we apply Nate and Mike’s idea, it looks exactly like that filmmaking process in practice:

• Define success (your intent): What is the exact goal of this project?

• Communicate that intent: Use prompts, agents, and workflows to direct your AI.

• Accurately evaluate the result: Did you achieve your goal? Review the output and adjust your directions for the next take.

To make sure you’re communicating your intent well, here’s a simple trick: ask your AI to confirm it understands the assignment. Just add this to the end of your prompt:

Treats to Try

*Asterisk = from our partners (only the first one!). Advertise in The Neuron here.

1. *Luma AI transforms your text into videos and lets you completely restyle any video's background, characters, or setting in post-production.

2. Browserbase runs browsers in the cloud for you to automate agent browsing and workflow tasks, with Stagehand providing an AI-native automation framework that uses natural language commands, Director offering a visual workflow builder, and Browserbase MCP connecting it all to AI assistants like Claude for seamless browser control.

   1. P.S: They also have custom Evals for which AI is the best for this use; here’s Matt Berman and Founder Paul Klein breaking it down.

3. Google AI Edge Gallery lets you chat with AI, transcribe audio files, summarize text, and ask questions about uploaded images, all running locally on your phone without needing internet connection—free to try and open source (download from Play Store, search “Google AI Edge Gallery,” and select your preferred Gemma model size).

4. Math, Inc. introduced Gauss, an autoformalization agent that completed Terence Tao & Kontorovich’s Strong Prime Number Theorem project in Lean.

5. K2 Think is a 32-billion-parameter open-weights reasoning model (from G42 in the UAE) that the team says matches or beats much larger systems, with reported inference speeds around 2,000 tokens/second on Cerebras hardware—try it here.

6. Teton monitors your elderly care facility rooms with sensors and sends real-time alerts to staff when residents need help, like detecting if they're about to fall or having trouble sleeping.

Around the Horn

• OpenAI’s cut of revenue going to Microsoft will shrink from ~20% this year down to ~8% by 2030, which The Information says could be worth ~$50B; meanwhile, Microsoft will get ~one-third of the new company (worth ~$100-$166B, if you consider the current $500B valuation the official one).

• Penske Media sued Google over AI-generated search summaries, claiming they steal content and caused a 33% drop in affiliate revenue.

• Box announced Box Automate, new AI agents for enterprise automation in its “era of context” (read more from Aaron Levie of Box on X).

• xAI laid off 500 workers (33% of its data annotation team) to pivot from general data labeling to specialist AI tutors in STEM and medicine.

• Californiian lawmakers passed AI safety bill SB 53 and now awaits Governor Newsom's signature; if enacted, it would require major AI companies to disclose safety protocols and establish “CalCompute,” a public cloud resource to democratize AI development.

• Here are the top AI papers from last week according to Elvis Saravia (a fantastic AI creator and educator!).

• Speaking of robots, watch this awesome interview with Dwarkesh Patel and Sergey Levine of Physical Intelligence about how fully autonomous robots are a lot closer than we think (case in point: Physical Intelligence could be about to raise capital at a $5B valuation!)

Monday Meme

A Cat’s Commentary