OP #321: Energy From Building, Synthetic Audience Research, and More

A full OP today which covers a wide range of topics... please make sure to take the synthetic audience research survey.

Author

Darren Herman
March 25, 2026

Happy Wednesday.

I’ve come to the conclusion that things are just better when your [college] basketball team is good.

I didn’t really have that growing up. My alma mater never had much of a team, so March Madness always felt like something happening in the background and not something I actually cared about. Frankly, I just “donated” money to my friends who were running the brackets and pools.

That’s changed.

My son’s heading to High Point this fall (class of 2030), and somewhere along the way they became our families team. Last year they made the tournament for the first time and ran into Purdue. My wife, Sherri, took my son to the game in Rhode Island and they had a blast.

This year was different.

They beat Wisconsin by one in the first round which was a total rollercoaster. Then played Arkansas and it was one of those games where you keep thinking “there’s no way this pace holds”… and it just did (I got tired from watching!). They lost, but honestly, it doesn’t matter. To me, they showed up and got to the second round. Even John Calipari, the legend coach of Arkansas, had nice things to say about HPU.

I have been all-in on rooting for them. It’s the first time I can remember actually being excited watching college basketball.

(Also—Knicks on a 7-game win streak. Not unrelated to my mood.)

On another note, I forgot how much I love building. Going back to my IGA, VMM, or even Mozilla Content Services days… starting something from scratch, putting it out there, and then seeing if anyone shows up. And when they do… there’s just nothing like it. I imagine many of you who are software engineers feel the same energy. Many of you have been with me on those journeys.

I’ve been back in that mode lately burning the midnight oil… I even purchased a new Macbook Pro M5 to help me:

  • PlayWellDoGood — ticketing platform for Silicon Alley Sports. Already pushed through a decent amount of volume.

  • DudleyGym — AI personal trainer. Small group of us are using it now and it’s kind of wild how good it’s getting.

  • MKTGDPT — basically a 10x marketer layer for our portfolio companies. Still early, but you can feel where it’s going.

Working on these has brought me back to building games with my late friend Josh. We launched PetsofWar on Student.com and watched tens of thousands of people start using it, commenting, breaking it, loving it.

That feeling stuck with me and I haven’t had that energy in a while.

But you are not here to read about how I’m feeling… but moreso, learn about why I have that feeling. I learn by doing and there’s no better way to learn AI than actually building and implementing yourself. I’ve been taking my HTML/CSS skills from the 90s and using some of those learning loops today.

It’s a little chaotic and a lotta scrappy. You don’t fully know what you’re doing (even if you kind of do). But it’s a ton of fun… and I’m learning a lot that’s directly applicable to what I do in my private equity role. No, I am not leaving PE - but I am trying to learn as much as I can, to bring these skills, partners, and thinking to my role… which, is one of the most-AI impacted functions. I’ve done this my entire career - deep immersion into a space - and it’s dramatically helped my thinking.

Thanks again for subscribing. Oh- please read the Synthetic Audience Research note below. I could use your input.

Be well, do good.
Darren

Synthetic Audience Research

I’ve been spending a lot of time lately on synthetic audience research.

You’ve probably heard some version of this already. It’s where you use LLMs to create personas that behave like real customers. You can ask them questions, pressure test ideas, campaign concepts and more.

Some people think this is the future of market research. I’m not sure it replaces it—but it’s definitely becoming part of the stack.

I’ve been going pretty deep here, working with teams at AARU, Evidenza, Simile, AuxoAI, and a few others. My guess is a lot of you are starting to poke at this too, or at least thinking about it.

So… I put together a quick survey that takes ~4 minutes to complete.

If you’ve used tools in this space, I’d love to hear how.
If you haven’t, I’m honestly just as interested in why not.

Both sides matter.

My intention is to share what comes out of it in a future OP. The more participation, the better. Thank you in advance.

It's a Skill, Not a Pill (Jeremy Utley Blog)
Only 3% of the workforce are using AI to achieve meaningful productivity gains despite a proven 40% quality boost being available to anyone, revealing AI proficiency is a learnable skill rather than an innate advantage.

Staying Human in the Age of AI (Rishad Tobaccowala's Substack)
Tobaccowala reframes the AI era with the metaphor "we are living in a world of aliens and replicants”. And we are humans emphasizing the fundamental distinction between computational and emotional intelligence as critical for business strategy.

One Billion Lost Packages (tomtunguz.com)
Eighty-five million packages arrived damaged in the U.S. in 2024, up 30% from the prior year, costing businesses $4 billion, creating opportunity for AI-powered operations solutions.

Building Next-Horizon AI-Native Experiences (McKinsey)
Most organizations fail with AI adoption because they design tools for yesterday's workflows rather than reimagining systems for AI-native collaboration; success requires shifting from command-and-execute interfaces to iterate-and-refine models.

Ari Paparo framed the current moment as one of rapid experimentation where AI and agentic systems could rewire consumer behavior and programmatic advertising mechanics.

Investors emphasize a fundamental shift from 'AI curiosity' to practical application and scale, the fastest companies aren't building the largest LLMs but solving domain-specific problems; VCs are moving from backing 'AI startups' to seeking exceptional founders leveraging AI to 10x traditional market efficiency.

All Shopify stores will appear inside ChatGPT by default in late March 2026 through agentic storefronts, with Shopify Catalog surfacing merchant products in responses when users search for products.

A critical analysis of competing agentic AI standards (AdCP vs. ARTF), arguing that industry confusion around fragmented standards is obscuring what marketers actually need at a time when clarity is paramount.

YC's Spring 2026 RFS covers 8 categories and signals that AI-native companies with software margins can now be built by tiny teams, faster and cheaper than ever—relevant for operators optimizing team structure and capital efficiency.

Performance improvement represents 33% of the AI in sports market ($2.9B by 2030), driven by personalized training programs, athlete workload monitoring, and integration with wearables and biometric data.

Successful founders operate in motion, prioritize warm networks over cold outreach, and demonstrate that startup performance ultimately reflects founder psychology; particularly their ability to stay calm under pressure and maintain transparency.

Thank you for reading OP #321. Feel free to reach out with any questions or feedback… I’m all ears.

What did you think of today's OP letter?

Login or Subscribe to participate in polls.