Why I joined the Globe

July 30, 2025

Proprietary data + deep domain expertise form the strongest AI moat.

After seeing a single OpenAI update wipe out scores of Y Combinator AI startups, I thought long and hard about how to develop a deep and sustainable moat.

I realized that the moat came from deep domain expertise combined with private proprietary data. This is made extra important because LLM performance plateaus as high-quality data becomes scarce.

A model that is trained on your proprietary data and used as a tool within your own unique workflows is much harder to copy. LLMs have to be a feature not the product.

Partnering with incumbents lets AI double revenue faster than launching new startups

To capitalise on this insight my AI agency partnered with companies that were less high tech but had deep domain expertise and proprietary data.

I spoke to oil brokers, partnered with specialist translation companies, worked with e-sports gaming companies. Because they already owned unique data in their specific nice.

When we sped up a month‑long task to two weeks, revenue doubled without starting a “new startup” from scratch.

You don’t need to do sales if you are already in the company. Yes, you do not capture as much upside, but how much upside is there to capture if your product is undifferentiated?

Meaningful impact demands immersive focus in a single business domain

Large, durable impact in AI demands immersive focus on one business domain. Deep immersion lets you map the real money flows, absorb team culture, and master the tiny workflow quirks generic consultants never see. That breadth of insight requires more than code: you need product intuition, crisp communication, and change-management skills to carry the tech across the finish line

Spreading attention across many clients dilutes depth and erodes quality. Each context switch forces you to rebuild mental models draining my brain power; even top engineers struggle to maintain flow when toggling domains all day. And that’s just the engineering context adding the business context on top makes it even harder.

To restore depth, I shut down the agency and decided to devote myself to a single, high-leverage company. Fewer plates mean tighter craftsmanship, clearer accountability, and the satisfaction of seeing ideas through from prototype to production impact.

The Boston Globe offers the perfect mix of brand, scale, and untapped data.

I almost skimmed past the AI Engineer Fellow role at The Boston Globe. My first thought was, “People still read newspapers?” A quick dive changed that. The Globe owns a 150+ years of reporting expertise and a mountain of private data that no start-up can spin up overnight. Pair that with a household-name and real-world infrastructure and you get an incredible moat.

Note: If you have not watched the movie Spotlight you should do it right now!

The numbers backed it up. Roughly half a billion dollars in annual revenue means the organisation has the muscle to fund serious experiments. The playbook stays the same: dig into the archives, learn the workflows, spot the high-leverage points, then build.

The scale is bigger than any customer I’ve served or any company I’ve worked at. But the core idea remains the same: go deep, unlock hidden value, execute with focus.

Build, write, give back

My LinkedIn tagline is “Build, write, give back.” It mirrors what journalists do: they write to reveal something about society to society, giving people the power to change their world for the better.

(Inspired by John Henry’s 2013 op‑ed “Why I Bought the Globe,” which drove home how big, local institutions can still reinvent themselves if they marry purpose with innovation.)