Welcome to the AI For Humans newsletter!

Did you know that the same companies that spent 20 years gluing our faces to screens now want to free us from them?

This shift represents either the greatest act of tech-industry atonement since... well, ever, or perhaps the long-awaited gateway to a more natural, less intrusive way to interact with technology.

The big news this week is that OpenAI is going all in on audio (again).

According to The Information, the company has unified several engineering, product, and research teams over the past two months specifically to overhaul its audio models.

The goal? A much better audio experience for the long awaited Jony Ive-designed audio-first personal device. Is it going to be a pen? Headphones? A puck you wear around your neck or something else?

Who knows. But it is coming.

Ive himself has said he wants to make reducing device addiction a priority and sees audio-first design as a chance to "right the wrongs" of past consumer gadgets. This def feels weird considering it's coming from one of the main people responsible for this addictive black rectangle in my hand right now.

But the bigger question to ask is: what does OpenAI have to gain from this?.

Despite all its promise, no one has really cracked the best use case for AI audio yet.

And there's good reason for that.

First, you actually have to talk. We're used to talking to humans, not computers. And if you do it in public? People think you're weird.

Then there's the response problem. It has to feel real without trying to feel real. Do you get annoyed at the "um, oh yeah" stuff from OpenAI's Advanced Voice? That's them trying too hard. To be fair, no one has really nailed this.

Unlock a focused set of AI strategies built to streamline your work and maximize impact. This guide delivers the practical tactics and tools marketers need to start seeing results right away:

Stay ahead of the curve with these top strategies AI helped develop for marketers, built for real-world results.

Download the Free Report

In our work on AndThen, we've found that grounding the AI in an actual personality helps. But an assistant that's just... you, extended? That's harder.

The real prize here is the one-on-one AI assistant with growing knowledge of your life. That's the killer product. But we're still missing the memory and the privacy infrastructure to make it work.

A device that listens to you all day is a step in that direction. And if OpenAI can pull it off, they cement their place in the AI stack.

Big if, though.

We've been here before.

The Humane AI Pin burned through hundreds of millions before becoming one of tech's most high-profile failures. The Friend pendant, a necklace that records your life and offers "companionship," has sparked equal parts privacy panic and existential dread.

And now there are at least two companies building AI rings, including one from Pebble co-founder Eric Migicovsky, that let you literally talk to your hand.

Love this newsletter? Forward it to one curious friend. They can join in one click.

The form factors keep changing. But the thesis remains the same: audio is the interface of the future. Your home, your car, your face, everything becomes a listening post.

To me, one of the most interesting things about this shift is what it means for how we interact with AI.

OpenAI's new audio model (slated for early 2026) is supposedly going to sound more natural, handle interruptions like an actual conversation partner, and even speak while you're talking.

That last part is huge. Current models can't do real conversational turn-taking. It's always call-and-response. This would be actual dialogue.

This would maybe finally be the version of Her that Sam teased so long ago.

For anyone who's worked with voice AI, this is the dream.

The real-time APIs right now are good, but they still feel like talking to a very smart answering machine. The models are getting better at not just understanding what you're saying but how you're saying it. The cadence, the interruptions, the half-finished thoughts that are just part of how humans actually communicate.

For now, I think the opportunity for creators is obvious. Audio experiences like podcasts, interactive voice games, AI companions, ambient storytelling... all of it is about to get a lot more interesting. The infrastructure is finally catching up to the imagination.

And if the big tech companies are right about where this is going, the screen might not be the center of our digital lives for much longer.

Whether that's liberation or just a more subtle prison, well... we're all about to find out.

That’s all for this week and we’ll be back on Thursday with a new episode!

-Gavin

We pride ourselves here on trying all the latest models and being able to tell you what’s good and what’s not so good but sometimes there is such a blast of new AI stuff that it’s hard to get to everything.

Over the holiday break, Claude Code and Anthropic’s latest model Opus 4.5 had a real rush of positive, organic press from the coding enthusiasts of the world and I realized that I had not spent nearly enough time trying it out myself.

Reader: It is SO good.

If you’re even remotely interested in how far you can push frontier LLMs, I suggest you go and give it a shot. You’ll have to sign up for one month of a Pro account ($20) in order to get access but it’s useful for a LOT, not just high end coding tasks.

Here’s a link to gift it to someone or get it yourself (not an affiliate link just the easiest way to get you to the ‘pay-for-it’ page).

I’m working on a novel right now (yes, I write weird stuff outside of this) and, while I’m pretty opposed to AI writing any actual words people read, it can be super useful for organizing thoughts and large files if the system doesn’t screw it up.

Mostly, they still screw it up.

Opus 4.5 is the first LLM system that does not. It was able to take in my large document (60k words), give me a beat-by-beat synopsis and then work with me to help restructure the chapters in a way to make the book more compelling. I was also able to use it to brainstorm potential ideas for character arcs and the like.

One big downside: The Pro plan gets eaten up fast. It kind of sucks that the $20 a month plan doesn’t get you a ton of Opus 4.5 outputs but it is what it is.

What’s even crazier is that 2026 will have even better AI models and it’s hard to even image what this world looks like going forward. Get ready y’all!