Since its inception, Knowm Inc has had a genuinely funny problem:

Everyone thinks we’re much larger than we actually are.

That’s probably par for the course given the nature of our work, but I’ll be honest—it’s been irritating in a very specific way. Why we’ve remained stubbornly small for so many years is a story for another blog post, one involving the CIA, the NSA, the Air Force Research Lab, and a generous helping of politics. Let’s save that for another article and just say this: we were too early, and our tech attracted the wrong kind of attention.

We could take lots of money and burn out, or go into hibernation like extremophile bacteria on the outside of the International Space Station, waiting for the inevitable AI asteroid to make the world a little more habitable for us. Hibernation it was.

Since that decision, our competitors have come and gone, burning out in predictable VC-funded fireballs. And yet here we are—still standing, still getting written up in industry reports, still allegedly controlling billions of dollars of the growing memristor market (lol), still sitting on some of the most powerful technology in the AI space and watching the world go bonkers.

When it became obvious our technology was still years ahead of market demand—and that we’d jumped the gun—I made a deliberate decision to go into hibernation. During that time, I continued packaging memristors for researchers around the world (see: memristor basics), but otherwise kept busy with unrelated work, quietly waiting for the future to catch up and laughing every time I saw a report about Knowm’s mythical market dominance. But one thing was really not that funny: our phone line.

Because public perception and reality were so wildly misaligned, our phone number became a beacon for solicitors, spammers, phishers, and other assorted undesirables. Day and night. Dozens of calls every day. Maybe one of those calls was legitimate. Eventually, it got so bad that answering the phone became impossible.

Enter: A “Fun Side Project”

About a year ago, Tim and I decided we wanted to work on something together again. We wanted hands‑on experience with modern AI programming tools, and I was watching the rapid emergence of genuinely capable voice AI systems. The idea hit me:

What if a multilingual voice AI answered Knowm’s phone line? It seemed practical and fun. So we built it. And then something unexpected happened.

Oops, We Built a SaaS!

Fast‑forward a bit and suddenly Tim and I had built an entirely new SaaS product: AI Receptionist.

It worked immediately for Knowm. The spam stopped. Real callers still got through, and many got their questions answered without my involvement—which was glorious. The phone line became usable again for the first time in years.

My aunt, who operates a mobile pet clinic, was our next customer. It worked brilliantly for her, saving many hours on the phone and dealing with truly stressful calls. We realized we had a real product: an AI receptionist for small businesses, a voice AI phone system that blocks spam calls, answers FAQs, and books appointments.

So we decided to go for it and scale it up. Why not? Read about our first year in review: AI Receptionist: A Year in Review (2025). Now we’re onboarding new clients daily.

“Boring Problems” Can Have Fun Solutions

Applying modern AI to boring problems has turned out to be far more fun than I expected. Building an AI Receptionist now running at ai-receptionist.com, didn’t just validate how powerful today’s AI has become—it validated the new programming tools themselves. The velocity and leverage modern “agentic AI” coding provides is incredible. As someone who has spent decades programming, it feels like we’ve crossed a threshold in practical AI development. If you’re tracking the broader AI safety and standards landscape, the NIST AI program is a good anchor point.

Maybe It’s Time to Wake Knowm Back Up?

When Knowm started, machine learning couldn’t reliably tell a cat from a dog. A “universal cortical algorithm” hadn’t been discovered. Power efficiency in AI—an obvious, looming problem—was largely ignored as a “solution in search of a problem.” Now power efficiency is front and center. Hardware matters again. And suddenly, a lot of the things Knowm has been quietly obsessed with for years look relevant.

So maybe it’s time for Knowm Inc to come out of hibernation. Maybe it’s time to show people how to build LLMs with memristors. It honestly does not seem that hard to me—a bit boring, if you will—since the problem is essentially one of mapping very large matrix multiplies to memristor architectures efficiently and with high fidelity.

You know what?

I think we might just start publishing again.