WTF is a Software Moat in 2026?
The Weekend Windup #30 - Cool reads, events, links, and more
I’m back in Salt Lake City, cat-napping and hiking the hills and trying to shake off the final dregs of jet lag from Asia. It’s been a good week to clear my head, get outside, and think about where our industry is actually heading right now.
Earlier this week, I was catching up with a VC friend, and we had a great chat about moats and defensibility in the AI era. Historically, software moats were built on a few reliable pillars: feature velocity, great UX, and a massive engineering team capable of executing flawlessly. When coding agents can ship features around the clock for pennies on the token, that entire model gets inverted. If the barrier to writing and shipping code drops to near-zero, what exactly is a moat anymore?
Borrowing a page from Charlie Munger, “invert, always invert”. Instead of asking what a moat is, let’s look at what is actively dying in the market right now, and what isn’t a moat anymore. I bucket this into a couple of things. The biggest ones are thin wrappers around foundation models. I can’t tell you how many founders pitch me products that, within 10 seconds of looking at them, are clearly just a UI duct-taped around Claude or GPT. If your entire product is a harness around a public model, I wouldn’t call that a moat or an interesting product with any sense of longevity in the marketplace. You are also doing unpaid R&D for big foundation model companies. The moment their next foundation model update rolls out, your “special workflow” gets absorbed, and you’re back to square one.
Feature velocity isn’t a moat anymore. It used to be that large teams pumping out code had advantages over smaller teams. Shipping code fast used to be a massive competitive advantage. Now, everyone can ship fast. When speed is commoditized by AI, it ceases to be a differentiator.
What’s defensible these days? If you strip away the hype, you start to realize that moats exist wherever there is high friction. If an LLM can do it easily, it’s not defensible. Here are a few things I’m seeing that actually survive, at least today.
Systems of Record and Deep Infrastructure. Technologies that embed themselves deeply into mission-critical workflows are incredibly sticky. You aren’t going to “vibe code” your way out of using DuckDB or Postgres for serious workloads anytime soon. The same goes for enterprise giants like SAP or Oracle, whose tentacles are deeply embedded in many enterprises, many of which have tried to migrate away from these giants to little avail. Once you are the system of record, your defensibility is immense.
Proprietary Data and Custom Workflows. If you have unique telemetry on a business, a key workflow, or important data that the hyperscalers don’t have access to, you have a moat. This is exactly why improving your data architecture and models is now a business survival tactic, as AI amplifies the effects of poor architecture and data models. If your data is a mess, you can’t train or fine-tune models that actually give you a competitive edge. LLMs will query garbage. And if you can afford it, consider using private AI models instead of the public ones from the foundation model companies, who are making a big push into enterprise for obvious reasons: money and getting your data and workflows.
Deep Expertise, Distribution, and Brand. AI can aggregate existing knowledge, but it can’t create net-new frameworks from lived experience. Building something fundamentally new and tying it to a trusted personal reputation is something an LLM can’t replicate. You can’t prompt or vibe code your way into a trusted brand. Distribution and awareness are another area I get a ton of inbound. It seems almost every company has a great product, but their biggest obstacle is getting noticed. I expect distribution and awareness to get far more challenging soon, as the amount of noise continues to rise.
I’m sure there are others I’m missing, but hopefully you get the point. This shift will force us to rewrite the startup playbooks. Even go-to-market strategies are changing in real-time. I question whether the Lean Startup methodology still applies, but that’s for another rant. We’re already seeing the death of “per-seat” SaaS pricing in favor of per-action or token-plus-margin models.
It’s easy to think the sky is falling and cry for the old world of SaaS, but the ground shifting is what makes this fun. We get to invent new models of monetization, as we had to at other inflection points.
I’m hitting the road again next week. I’ll be in San Francisco on Tuesday, April 28th, for the Dev+AI event with DeepLearning.AI, speaking at the Agentic Analytics Summit (virtual) on April 29th, and then off to Stockholm the following week for the Data Innovation Summit. After that, I’ll be at AI Council in SF, Confluent Current in London, and some cool stuff with the Data in the D crew in Detroit. If you’re going to be at any of those, come say hi.
Also, a quick reminder, I’m still collecting responses for the April 2026 Data Modeling Pulse Survey. If you haven’t filled it out yet, take 90 seconds to share what you’re actually seeing on the ground.
Have a great weekend,
Joe
Here’s the YouTube/travel vlog version of this, live from the foothills of Salt Lake City, Utah.
Here’s this week’s Freestyle Friday podcast. Available on Spotify, Apple, and wherever else you get your podcasts. Please support the show with a review. It means a lot.
Modern data modelers need to live at the intersection of business and tech. Ellie.ai allows you to collaborate effectively with business while maintaining credibility with the Tech team. Get contextual support from AI, reverse engineer anything building a repository of sources with synthetic AI generated contextual metadata while delivering insights via an MCP Server and integrating anything with full blown API support.
The full stack data modeling future is here today!
Thanks to Ellie.ai for partnering on this newsletter.
Awesome Upcoming Events
Here are a few things I’m up to. Much more to come, so stay tuned.
Agentic Analytics Summit 2026
A LOT is happening right now with AI + analytics + agents. The Agentic Analytics Summit will have lots of great speakers and the latest updates on this fast-moving space.
Definitely register for the event. It’s free and will be awesome!
When: Wednesday, April 29. Starts at 9am PT
Where: Virtual
Data Innovation Summit 2026 - Nordics
🇸🇪 Sweden! See you at the Data Innovation Summit in Stockholm.
I’m doing a keynote and workshop on Mixed Model Arts: Data Modeling in the Age of AI.
May 7 - keynote
May 8 - workshop
Here’s 10% off: SD10OFF (good for the event. Workshop is not included)
Register here
AI Council The technical conference for humans who ship
For 10+ years, AI Council has been gathering the world’s top AI infrastructure minds to share what’s working (and where we’re headed next). You’re invited for 3 days of high-quality discourse with 1,200+ technical experts, including office hours, small groups, and zero marketing keynotes. Speakers include: The co-inventor of ChatGPT, Creator of DuckDB and Codex, Engineers behind ClickHouse, Databricks, Datadog, and LangChain.
Join us May 12-14 in San Francisco.
I’ll be there too, but not speaking. If you want to grab a beer with me and see amazing talks from amazing people, register now.
Use code JRSUB20 for 20% off, valid thru 5/4
Disclosure: They’re giving me a ticket to attend. AI Council is one of the last indie data/AI events around, and they do great work. I support non-vendor events as much as I can, and so should you.
Cool Videos and Reads
Are vendors trying to lock down your data? In this episode, George Fraser (CEO of Fivetran) breaks down why the "modern data stack" has evolved into "open data infrastructure". We discuss why data gravity is the most overrated concept in data management, how egress charges are often misunderstood due to poorly designed pipelines, and why companies must insist on having a true replica of their own data.
Here are some things I read this week that you might enjoy.
Inside GitHub’s Fake Star Economy | Awesome Agents
The Distribution Singularity: Why Speed, Story, and Surface Area Now Decide Who Wins in AI
Pricing Pressure Will Crush You
Drunk Post: Things I’ve Learned as a Senior Engineer
Project Deal: our Claude-run marketplace experiment | Anthropic
Find My Other Content Here
📺 YouTube - Interviews, tutorials, product reviews, rants, and more.
🎙️ Podcasts - Listen on Spotify or wherever you get your podcasts
📝 Practical Data Modeling - This is where I’m writing my upcoming book, Mixed Model Arts, mostly in public. Free and paid content.
The Practical Data Community
The Practical Data Community is a place for candid, vendor-free conversations about all things tech, data, and AI. We host regular events such as book clubs, lunch-and-learns, Data Therapy, and more.
Awareness is going to be awful.
deep fake software is sprouting like mushrooms.
Using local AI with local models is definitely worth thinking about for some industries.