Welcome, humans.

If the word “Napster” still makes you picture MP3s and early-2000s internet chaos, same. Which makes the company’s new business extremely weird in the best way: Napster is now building visual AI agents for businesses.

Napster CEO John Akunto calls the idea “streaming intelligence.” Instead of a customer poking around a website alone, Napster wants an AI agent to talk with them, learn what they actually want, and help the business keep that intelligence for itself. Yes. That Napster.

Here’s our favorite parts:

• (04:26) So what does Napster actually sell now? John explains how businesses can train a lifelike agent to become an expert on their products and have one-on-one customer conversations.

• (05:25) The four-word operating system: His framework is “capture, learn, act, own.” First collect better customer signals, then learn from them, act on them, and keep the resulting intelligence under your control.

• (14:51) Why give AI a face at all? John’s answer is simple: humans are audiovisual. He thinks talking to a visible agent will feel more natural than typing into another little box.

• (20:14) The penny-per-minute claim: John says Napster can price its agents around one cent per minute because more of the work happens at the edge instead of entirely in the cloud.

• (41:02) What happens when your AI knows you for ten years? Corey asks the portability question, and John argues that the intelligence should belong to you and move with you if a vendor disappears.

The interesting part is the ownership thesis underneath “Look ma, Napster made an avatar!”, which is: AI can become the front door to your business, but the customer relationship should not automatically belong to the model provider.

Why watch this? Because this is a surprisingly useful look at what comes after the chatbox. John connects visual agents, memory, customer data, edge compute, and eventually robots into one bigger idea: businesses may need an AI interface they actually control.

Watch and/or Listen now: YouTube | Spotify | Apple Podcasts

P.S. At 25:30, John compares today’s AI interface problem to the jump from DOS to Windows. The models may be powerful, but normal people still need a layer that makes them usable.

Keep scrolling for a plain-English breakdown of Napster’s “capture, learn, act, own” framework, our November 18 San Francisco event, and the last two podcast episodes.

Special shout out to Outshift for sponsoring this episode.

Additional Resources: What “capture, learn, act, own” actually means

John’s whole pitch can sound abstract until you map it onto one normal website visit. Say you show up looking for a shirt.

• Capture: Instead of only logging clicks, an agent can ask why you came, what you need, what you like, and when you need it.

• Learn: The business starts seeing patterns across those conversations. Maybe customers keep asking for a size, feature, color, or delivery option the site never surfaced well.

• Act: The agent can help complete the task, recommend the right product, or eventually continue the interaction in a physical location.

• Own: This is the part John cares about most. He argues the business should retain the customer data, models, and intelligence created by those conversations instead of handing the whole relationship to an outside AI provider.

The bigger bet is that this becomes an interface layer for more than websites. John thinks the same human-to-machine interaction eventually connects to physical devices and robots. See what Napster is building here.

UPCOMING EVENT: The Neuron IRL in San Francisco

The Neuron is going IRL in San Francisco on November 18 starting at 4:30 PM PT. If you’re in town, come hang out with us for drinks, bites, and a live recording of The Neuron podcast. We’ll have The Neuron community together in one room, plus some special guests and plenty of AI conversation.

Huge thanks to Alumni Ventures, Slack, and Sonar for helping us bring the night to life. It’s free, but space is limited, so RSVP and save your spot.

🎙️ In Case You Missed It…

Two recent interviews and episodes we think you’ll love.

1. For anyone who wants robots they can actually teach

TL;DW: Standard Bots CEO Evan Beard explains how robots are learning jobs by watching people demonstrate them, rather than following brittle, step-by-step instructions. His team is combining demonstrations, computer vision, and everyday language so factories can teach machines new tasks without constantly calling in robotics specialists.

Why you should watch: Evan gets practical about what happens when a part shifts or an unexpected object shows up on a conveyor belt. If you want to know when AI-driven robots are useful in an actual factory, rather than a flashy demo, start here.

Watch / Listen: YouTube | Spotify | Apple Podcasts

2. Can living brain cells make your AI videos faster?

TL;DW: Neurosurgeon-scientists Alex Ksendzovsky and Jon Pomeraniec are studying roughly 150,000 living brain cells on a 4,096-electrode array to learn how biological brains process information. They say the software inspired by those experiments can make AI video generation around 5x faster at roughly 80% lower inference cost. The software still runs on ordinary GPUs, not living tissue.

Why you should watch: You'll see the difference between using neurons to discover smarter computing techniques and actually powering a computer with brain cells. They explain how they read electrical activity from the cells, and why that research could make familiar AI tools cheaper and faster.

Watch / Listen: YouTube | Spotify | Apple Podcasts

Have a topic you want to learn about? Request it here!

Subscribe to our YouTube Channel for more!

Subscribe on YouTube to help us bring in more builders, researchers, and guests who can teach you something useful about AI every week.

Stay curious,

The Neuron Team