Shovels
Shovels
Get Started
Company
How Shovels embraces AI tooling and what it means for the future of data startups

February 2026 Newsletter

Company
Ryan Buckley

Ryan Buckley

CEO

This newsletter is going to be about AI at Shovels: how we use it, how we think about it, and how all this AI tooling is going to change the data landscape for startups like us.

And don't worry, this is not another "Look what I vibe coded while the barista made my latte!" kind of post.

Instead, I want to give a sober assessment of what it means to be an entrepreneur today. In short: it's scary.

Scary awesome, but also scary scary.

Let's go!

Shovels → AI

I'm sure I've written this before, but I'll write it again. Our domain was a happy accident. The first name of this project was ChooseAGC (as in, choose a general contractor). Not sexy. At all.

Then one night I went on a journey to the outer bounds of GoDaddy. I needed a new name. Something about construction. Something about digging. You get the picture, and then… Shovels. But of course the .com, .net, .org, .everything was all taken. Everything but shovels.ai.

Are we AI? No, not really, I thought. But maybe someday.

So I bought it. ChatGPT dropped exactly three months later!

("True fact," as my kids used to say.)

Fast-forward to today and we are definitely an AI company, but so is everybody else, with the exception of maybe my dentist.

We have made it a point to embrace all of it, to not be afraid to break our routines and mental models of what it means to be a startup, a data company, or to have venture capital investors. It feels like we're running through a funhouse sometimes, mistaking mirrors for hallways, bouncing off the walls as we go.

This is the nature of entrepreneurship today, but honestly to me it feels like there's no other way. We could embrace it cautiously like everyone else, but then we'd look and feel like… everyone else. That's not us. We're curious and willing to find the outer boundaries and if that means a few bumps and bruises, we'll take them. This is the Shovels way.

I'll give you a few examples.

  • A few weeks ago we got every non-engineer using Claude Code. Claude Code lives in the command line. We onboarded teammates who had never opened Terminal before, completely new to shell scripts. Just showing them how to install Homebrew was an eye-opener. And then… magic. They fully embraced the simplicity of command line tools. It's not without hiccups, and we may still ditch this for Claude Cowork, but the point is we tried and good things happened.
  • Morgan, our VP of Marketing, isn't a coder. Well, he wasn't a coder when he joined us. Unfortunately for him, our marketing website is built on Pelican, a Python-based command-line tool for static websites. What to do? We got him going on Cursor and later we switched to Claude Code. And now he can do it all, way more efficiently than if we were on Framer or Webflow like… everyone else.
  • Most recently, we've begun to experiment with OpenClaw. It's been all rage in the AI internets and the punchline finally came over the weekend when OpenAI scooped it up. We named our OpenClaw agent Libby and now we can chat with her on Slack. She can help us with almost anything, from permit counts to city council downloads. I don't know exactly what OpenClaw means for Shovels. This one feels different. It's dangerous. It feels like a new best friend who's just a little too perfect and that makes me suspicious. We're being careful, but boy oh boy does this one have potential!
  • And then there's Luka, my co-founder, who has built more bespoke AI tooling for Shovels than I can even start to tell you about. It's fantastic. We have agents and sub-agents with more personas and backstories than a Russian novel. Our entire codebase is AI-optimized now, primed for the kind of tooling that gives us 10X leverage. Yet it feels like we're just getting started.

Are we an AI company? You're darn right.

Everyone → AI

We are not alone. Every entrepreneur, every startup today is using AI. I think the most likely to succeed will remain on the cutting edge, pushing the limits of security. Higher risk, higher reward.

This is what worries me. If we get beat, we'll get beat because someone else levered this technology more than we did, and that's scary. It should scare every entrepreneur. This game has become a race to see who can find the right mix of AI tools before the next guy does. It means companies will get sloppy, and they may start to look the same.

This is the problem when everyone's on AI. If we all write emails, LinkedIn posts, and newsletters with Gemini, Claude, and ChatGPT, we all sound the same.

Similarly, if we all write code with Gemini, Claude, and ChatGPT, we all build the same stuff, with the same quality, and the same outputs.

I think about this a lot. We want to use AI, but we don't want to be the same. How does that work?

I think the answer lies in the dependency relationship. People new to coding will "vibe code" — let the AI do it all. This will always lead to average outcomes. But when they start to inject opinions, thoughtfulness, and actual understanding of customer problems, the outcomes diverge.

Suffice to say, that latter point is what we do at Shovels. We leverage AI A LOT — but we have been fine-tuning the dependency now for months and months. We're comfortable with where it stands. It's our Goldilocks approach: not too hot, not too cold, not too hard, not too soft. Don't go to extremes.

I'll wrap up this section with a few random thoughts I've been bottling up over the last few months.

  • Software is going to be a bloodsport. To win at software, you need to have better distribution, a better brand, something else extraordinarily special. You'll be competing against every vibe coder on the planet if you try to launch a software product now.
  • We're all benefiting from subsidized AI. The foundation model companies are betting on economies of scale that haven't materialized yet. There's a scenario where the prices we're paying now are below the long-term variable cost. In other words, get it while it lasts.
  • I'm super excited about projects like Ollama and the brilliant silent heroes who will figure out how to shrink the models and compute requirements so LLMs can run locally on your phone. We already can run them on MacBooks.
  • I'm obviously super bullish on high quality bespoke data. There's a massive opportunity for data companies like us. The market appetite seems insatiable.

So I'll conclude this newsletter with the most obvious thing: a story about data that was missing in the market.

Things to do with permits: predict rezoning projects

ICYMI, we bought ReZone a couple of months ago. We're finally stitching the permits and decisions together and the results are super cool!

Here are two cases where similar permit patterns foretold a rezone was underway.

Case Study 1: Chicago (Jackie Robinson Park, 10540 S Morgan St)

The permit history tells a pretty clear story of decline: real investment in 2011–2013 (over $33K in renovation permits), a couple basic security upgrades in 2018, and then… nothing.

Our permit data captured the first signs of revival: a $625 fire alarm permit in November 2024, followed by a $175 modification permit in January 2025. These weren't just routine permits — they were the first tangible steps toward an 18,000-square-foot fieldhouse that will serve the historic Jackie Robinson West Little League community.

In December 2025, the city approved $2 million in TIF funding to complement a $15 million state grant. We saw this in the ReZone decision above.

The permit timeline dropped some hints that a rezone was going to happen!

Case Study 2: Raleigh (2806 Gresham Lake Rd)

From 2010–2016, the permits read like an industrial greatest-hits album: abatement work, waste management, grading for gravel lots, and a rotating cast of tenants (with sign permits as the breadcrumbs).

The first hint of change? A 2023 residential HVAC permit in an industrial zone, followed by attempts to build a carport.

A very recent decision, linked above, shows the site is approved for Industrial Mixed Use-5 stories (IX-5) zoning, paving the way for a $15 million, 42-court pickleball facility. WRAL reports the facility will open in 2026, transforming a 7-acre warehouse into a recreation destination.

The pattern

Both cases demonstrate how permit data can signal impending zoning changes:

  • The Quiet Period: Extended gaps in traditional use permits (6 years for Chicago, 3+ for Raleigh)
  • The Transition: Out-of-character permits (residential HVAC in an industrial zone)
  • The Surge: Quick succession of new permits once zoning changes, often starting with fire/safety systems

These patterns emerge months or years before official announcements!

Perhaps we'll build a prediction engine based on these data and offer up a "rezone likelihood" score. Endless opportunity!