Inside Ai4 2026

Hey there, I'm Emil Protalinski and this is Hard Coded, a newsletter where I share my firm opinions and knowledge about the most compelling tech news of the week. In Hard Coded’s inaugural issue, I’m breaking the format right out of the gate to bring you a snapshot of Ai4, which grew from 8,000 attendees last year to 12,000 attendees this year. Aside from the brutal registration line, the conference ran without the usual growing pains you’d expect from 50% YoY growth.

Because AI is such a horizontal technology impacting every industry, there wasn’t a clear throughline at the conference. Also, I'm told the recordings of my panels won't be ready for a while. So, here is a snapshot of what more than a dozen CEOs, CTOs, and CMOs told me in 1:1 interviews:

What are you worried about?

“The misuse of AI. For example, I read that Palo Alto Networks was doing some experimentation and asked Anthropic and OpenAI to hack into systems and those models said no. But DeepSeek happily did it.” — Kognitos CEO Binny Gill

“Being a new company. You want everything to fire on all cylinders at all times.” — Actualyze AI CEO Rafi Khardalian

“There is always the fear of being made obsolete.” — Airia CEO Kevin Kiley

“The funding environment is more complex than I thought it would be. There's a lot of money going into a small set of companies.” — Axonis CEO Todd Barr

“There are a few players in the market that have a very incorrect, immature perspective on what really matters when you're building and scaling infrastructure. These players did not exist two years ago. If they did exist, they existed for renting bare metal GPUs for Bitcoin mining.” — Vultr CMO Kevin Cochrane

“Security concerns. We started as a remote monitoring company. We have millions of devices connected to our systems. I'm always saying to every new employee in the company: We are a gold mine for hackers.” — Atera CTO Oshri Moyal

“I get stressed by potential. We're sitting at the precipice of something incredible. We really are on the edge of being able to figure out how to program biology and human biology.” — Generate:Biomedicines CTO Gevorg Grigoryan

“I have a lot of pride in the people on board. We have really incredible, smart, and humble people. If I can make them succeed, I will be very happy.” — Taktile CEO Maik Taro Wehmeyer

“Big Tech companies collecting all my data. I’m a privacy nut. I use Tor. On my keychain, I have a little YubiKey. I keep Bluetooth off on my phone.” — Modulate CTO Carter Huffman

“Can you run fast enough? That’s the only thing in front of us.” — Vinci CEO Hardik Kabaria

“How do we build something that is reliably harness agnostic, gateway agnostic, and model agnostic so that we can onboard a lot of these models and basically remove the integration friction piece of the equation?” — Not Diamond CEO Tomás Hernando Kofman

“How do we get more and more of an organization's token traffic through the gateway and make this governable for the organization?” — TrueFoundry CEO Nikunj Bajaj

“Failing to hit that window because I don't know how many more windows like this are going to exist in my lifetime to be able to terraform a broken industry.” — Akido CEO Prashanth Samant

What are you excited about?

“Neurosymbolic AI and context graphs are real.” — Kognitos CEO Binny Gill

“We're going to see the cost per token go down dramatically.” — Actualyze AI CTO Sean Lynch

“We’re all following the interesting debate about open models. I think a very healthy discussion is happening around tokenomics that is 180 degrees from where we were even six months ago.” — Airia CEO Kevin Kiley

“I’m running a 122-billion-parameter model in my house and that blows my mind that I can do that and get great performance: 55 tokens per second.” — Axonis CEO Todd Barr

“The upleveling of platform engineering as a function and discipline within an infrastructure operations and cloud services team. Having platform engineers become the driver of redefining how enterprises build and deploy infrastructure to support agentic AI.” — Vultr CMO Kevin Cochrane

“The goal loop. It's a concept where you can take very large projects and let the AI work on them continuously until the goal is achieved by creating loops of work. So, you can take a really big project, break it down to a lot of small tasks and make sure that they are happening.” — Atera CTO Oshri Moyal

“The availability of robotics is super interesting.” — Akido CEO Prashanth Samant

“The physics-based view of the molecular world and the machine-learning-empirical-learn-from-database view… I really see a path in which they're starting to converge.” — Generate:Biomedicines CTO Gevorg Grigoryan

“The harness of agents. How can you actually make sure agents operate under some constraints?” — Taktile CEO Maik Taro Wehmeyer

“The goal is to be able to talk to something just like you talk to Computer from Star Trek or talk to another person. Hierarchical models are the best architecture that I've seen to solve that problem.” — Modulate CTO Carter Huffman

“Mixture-of-Kittens. We're looking at interesting ways to scale models such that they're really efficient but also can hold large contexts. People are looking at mixture of experts and doing a lot of innovation there, that's pretty interesting to us.” — Vinci CTO Sarah Osentoski

“Research out of Ramp called PorTAL. Basically, reusable adapters for fine-tuning.” — Not Diamond CEO Tomás Hernando Kofman

“Automated routing layers. There was a recent paper that talked about how you can break it down into multiple stages of automated routing. We are taking a lot of inspiration from that paper and implementing our own routing technique.” — TrueFoundry CEO Nikunj Bajaj

Feel free to reply to this email and let me know which interview you’re most interested in seeing as a Q&A.

(Disclosure: Ai4 covered my travel to the event.)