The AGI Conference is the longest-standing conference series dedicated specifically to artificial general intelligence, and remains the primary venue where the field's foundational researchers present peer-reviewed work. 

This year's program spans four days of keynotes, paper presentations, workshops, and technical demonstrations. Speakers include Ben Goertzel, Karl J. Friston, Alison Gopnik, Neil Gershenfeld, Alexander Lerchner, Catherine McMillan and many more!

Here's what the partnership means for you:

The AGI Society has prepared customized sponsorship packages exclusively for the AI Collective community. If you want your company in front of the researchers, founders, and investors shaping this field, please reach out to our team and we'll connect you directly.

If your work touches agents, alignment, or the architecture questions behind general intelligence, this is the strongest concentration of that thinking in one venue this year. A dedicated Investor Day on July 30 closes the conference, focused on the commercial landscape forming around the field.

One safety story a week, pressure-tested for what's actually happening underneath the surface.

By Lindsay Gross

Nowadays, when someone takes your picture, their arm lifts a phone to eye level with the lens pointed at your face. You know this means someone is taking a photo of you. Nobody ever taught you that.

Meta's AI glasses shrink that gesture down to a fingertip. Taking a photo means pressing a button on the temple or saying so out loud. One of those you might catch. The other looks like a person adjusting their sunglasses. What's left is a small LED on the frame, and that light is the only notice anyone gets who isn't wearing them.

Meta's own European regulator asked about that light five years ago. In 2021 the Irish Data Protection Commission, which oversees Meta's privacy compliance for the whole EU, and Italy's Garante pointed out that a phone is "visible as the device by which recording is happening, thereby putting those captured in the recordings on notice," while the glasses offer "a very small indicator light." They asked Meta to demonstrate the light actually worked. Nobody had field-tested it.

You can also just remove it. On June 5, NBC News' Joanna Stern paid a stranger in a New Jersey garage to disable hers. Her team searched Facebook Marketplace for the service and found listings in 30 states.

Eighteen days later, Meta took the Ray-Ban name off three new pairs. Asked about privacy that day, CTO Andrew Bosworth reached for history. "I'm old enough to remember when there was controversy about phones having cameras," he told Gizmodo. "So, there is this social learning thing that has to happen." No camera cover shipped. No fix for the people disabling the light. Then Meta updated its second-generation glasses so you can't turn the light off without turning the camera off.

Now the Financial Times reports Meta is testing "super-sensing" prototypes that shoot a photo every few seconds and record audio continuously, with executives discussing leaving the light off. Per the FT, the raw footage wouldn't be stored by Meta or be available to the wearer. What's left is whatever the model took from it.

Bosworth is right that we adjusted. We did. There are rules now, and most people follow them. You don't film in a locker room. You ask before you post a picture of someone's face. Somebody says put that away, and you put it away.

But every one of those rules rests on the same thing: you can tell it's happening. You could see it, and so could everyone else in the room. Shared seeing is what makes rules enforceable in the first place. A norm needs a witness. That's the mechanism, and it's the thing being engineered out.

So watch one specific thing, because it's binary and it's close: whether Meta's super-sensing AI glasses ship with the light on.

The full version is on Hidden Layer.

Seven teams took the stage at Florida Atlantic University for a demo night run with the FAU Entrepreneurship Institute and hosted by Fernando Koch and Kevin Cox. The lineup skipped slideware for working systems. Drones that detect landmines, agents that run regulated compliance workflows, and a platform that compresses tokens to cut inference cost.

Patriot Logistics Group, led by Paul Murray, won Best Overall, and Nyquest AI, led by Michael Simpson, won Best Tech, both decided by live audience judging. Attendee Aman Sharma summed up the night as real deployment over theory, built around infrastructure, autonomy, and use cases teams can ship now. That's the bar we want every chapter demo to clear.

The AI Collective's online Lunch & Hack series ran a session called Trust Is the New Moat, led by Adelina Martiniuc with Vik Ghai and Michael Luu. It opened on a question every enterprise AI founder eventually runs into. What does it actually take to build systems a large company will put into production?

Martiniuc put the stakes plainly. "Every enterprise AI deal has a hidden gate. Before a customer signs or an investor writes a check, they're asking the same question: Can I trust this where it matters?" With Ghai, a three-time founder who now backs enterprise AI at G2C Ventures, and Luu, the group worked through what earns that trust in practice instead of on a slide. The full recording is on YouTube, linked from Adelina's post.

Our partners and sponsors get exclusive placements across the newsletter and access to AIC's in-person network — demo nights, dinners, hackathons, and forums across 180+ chapters.

For all inquiries, send us a note at [email protected].

The AI Collective is built by volunteers across 180+ chapters in 40 countries.

Thank you to the thousands of volunteers around the world who make this work possible. We truly could not do this without you.