👋 Howdy to the 1,054 new legends who joined since our last edition! You are now part of a 580,545-strong tribe outperforming the competition together.

LATEST POSTS 📚

If you’re new, not yet a subscriber, or just plain missed it, here are some of our recent editions.

🧸 Why AI Pivoted To Cute. How companies are pivoting their brands to be mascot-led.
🌊 Why Payroll Is Still Broken (And Fixing It). An interview with Ayush Sharma, Founder & CEO at Warp.
📔 The Fine Art Of Disagreeing & Committing. Why consensus is where good decisions go to die, and the seven-step fix.

PARTNERS 💫

Ship Framer sites with any AI. Framer MCP connects ChatGPT, Claude, Cursor, or Grok to Framer. Ask your agent for a landing page, a new blog post, or a full site, and it builds it on the canvas, fills your CMS, and publishes it live on your domain.

No editor open, no handoff to a designer. You review, tweak what you want, and ship. It works from wherever your AI tool runs, even your phone, and setup takes a minute.

Great product, thin content? HubSpot for Startups’ free Content Gap Finder scans your site, sorts everything you’ve published by buyer journey stage, and shows which stage is nearly empty.

You get five ranked ideas for what to publish next, each tied to the gap it closes. Free, and it takes about a minute.

_{Interested in sponsoring these emails? See our partnership options here.}

HOUSEKEEPING 📨

This photo below had a Muad’Dib level of prescience. Obviously, the results were a repudiation of the current administration, but they were also a shocking outcome for Elon. In terms of political donations, Elon now ranks #1 all time. And because people don’t like billionaires fucking with their elections, all it bought him was bad juju.

The Republicans have berated George Soros for years because of his donations; now Elon is Soros on steroids. It’s kinda insane that the U.S. works like this. Here’s hoping that some incoming left or right party takes the Citizens United ruling and flushes it down the toilet in the coming years. I think the U.S. would be much better for it.

INTERVIEW 🎙️

Jordan Tigani, Co-Founder & CEO at MotherDuck

Jordan Tigani is the Co-Founder & CEO (or, as his official title puts it, Chief Duck Herder) of MotherDuck, a Seattle-based cloud data warehouse built on the open-source DuckDB engine. Before going out on his own, Jordan was a founding engineer on Google BigQuery, wrote two books on it, and led first its engineering team and then its product team through its first $1B or so in revenue. He then became Chief Product Officer at SingleStore, where he helped turn the database company into a cloud-native SaaS business. MotherDuck emerged from stealth in November 2022 with $47.5M across a seed round led by Redpoint and a Series A led by Andreessen Horowitz, then raised a $52.5M Series B led by Felicis less than a year later at a $400M valuation, bringing its total funding to $100M.

A delicious irony sits at the heart of Jordan's story. He spent more than a decade building one of the world's largest data systems, only to notice that most users weren't running big queries at all. He turned that observation into an essay, Big Data Is Dead, and built MotherDuck on it. Jordan argues that most businesses need far less data infrastructure than they've been sold. Now AI is shaking up the data world all over again, and MotherDuck has moved quickly, shipping AI-built data visualizations (Dives) and agent-native data pipelines (Flights). Not bad for someone who told a conference panel a few months ago that AI wasn't really going to change data, before the models proved him wrong.

What problem does MotherDuck solve today?

When I started MotherDuck, the real problem I was trying to solve was that I'd just been fired from my last job and was trying to figure out what the hell I wanted to do next. I looked at places where I might find gainful employment and asked myself whether I really wanted to work there. A lot of them looked like they'd be crappy places to work, and I wanted somewhere I'd actually want to be. So how do you do that? Well, you can create your own. If you fail to build a good place to work, you have no one to blame but yourself.

As for the specific problem, I was originally just planning on hacking on something for a while. I'd helped create two SaaS analytics databases, first BigQuery and then SingleStore, where I helped build out their SaaS offering. Then I saw DuckDB, an open-source database that a couple of academics in Amsterdam had created. It was very cool, but it was also pretty early. I remember thinking that somebody should take it, put it in the cloud, and make it serverless. And then, well, why not me? I'd done this kind of thing before, although I'd never actually started a company.

So I was planning to hack it, and I got an intro to the founders of DuckDB Labs, the people who created it. I asked if they'd hire me to help build the SaaS service. They said no, as they wanted to stay focused on the core database engine, but they'd partner with me if I wanted to build something like that myself. Then a friend introduced me to a VC. I almost didn't take the call because I figured it would waste everyone's time and they wouldn't be interested. My friend told me not to worry, that they'd help me shape the idea and give me feedback. So I just described what I wanted to build. This was only about 72 hours after I'd first thought it might be fun to work on. The VC said it was a good idea and that he wanted to fund it. He led our seed round, and we were off to the races. Some founders start with a big technical problem they want to solve. This wasn't quite that. I wanted to solve how to build a great place to work. The technical problem came later.

What was hardest about taking MotherDuck from zero to one?

First, I looked to people I respected, mentors and people who'd done this before, and asked them what they did. I talked to a bunch of people. One great thing about being a founder is that other founders tend to be generous with their time. They have a specific set of experiences you just can't get anywhere else except by doing it. There aren't many articles or blog posts about it, because most people in it are just doing it. So I'd ask big questions. How much should I raise? Who should I raise from? How should I build this company?

The hard part was learning to trust yourself, figure out what's right for you and your company, and then do it. There's no right answer, and nobody will tell you what it is. You just have to go. You don't get to wait around and weigh every option. You just have to do it.

Four years on, who is your ICP, and what exact problem are you solving for them?

We're building a cloud data warehouse based on open-source DuckDB. Our ICP is generally small and mid-sized businesses, up toward the light end of enterprise. These are people who feel like their data warehouse is overkill. It's too expensive, it's not fast enough, and they're not getting the bang for their buck.

Often, they're using Snowflake or Redshift, and we can be dramatically cheaper and faster. We like to say that people come for the cost and stay for the performance. The other big group is people building applications that need to show analytics, some sort of charts or dashboards, to their end users. We have a tenancy model that works really nicely for that. Each end user gets their own cloud-based DuckDB instance that can autoscale and scale down to zero when it's not being used. If you think about the shape of the problem, it ends up being a lot closer to the shape of the usage you actually have than a bigger, beefier data warehouse like BigQuery or Snowflake.

Was that always the ICP, or did it change over time?

One of the key things at the beginning was that I'd worked on BigQuery, which is big data, and I'd noticed that most of our customers didn't actually have a lot of data. What we were building was overkill for them, and because of how we built it, it wasn't serving what they actually needed. We were designing for scaling to multiple petabytes and giant queries, while they were trying to run a dashboard on five gigabytes of data and wondering why it took five seconds.

So I wanted to focus on that problem, on the things people were actually doing with their data warehouses. That leads you to smaller customers, because the message resonates more with them. But it's also true that most of what the giant customers did wasn't really big data either. At BigQuery, we had Walmart, Home Depot, HSBC, and Equifax, some of the largest customers in the world, and most of what they did was small. So the plan was to start with smaller organizations, prove we could handle their scale, and then take on larger and larger ones as we went.

Another thing is that the size and shape of the machines you can get have increased dramatically since we started BigQuery. Google popularized splitting problems across many inexpensive machines. That's how Google Search worked, but it's also how MapReduce, GFS, and all of those things worked. Dremel, the query engine BigQuery was based on, operated the same way. But if machines were as big and shaped as they are today, we wouldn't have needed to do that. Now, with simpler solutions that run on one machine at a time, you can actually handle very large scale. We're proving it works for one customer scale, then the next, and moving up and up from there.

I also learned a lesson at my last job at SingleStore. If you go after enterprise too early, the customer who can write the seven- or eight-figure check owns your roadmap. You have to do everything to keep them happy, and they'll ask you to do all kinds of weird stuff. You tell yourself every other enterprise customer is going to want it too. Half of that is true, and half of it isn't. The half that's true is useful, but what you built for them isn't really what makes your product compelling. You just end up building a bunch of table-stakes, checkbox features. The other half is stuff other companies aren't going to care about at all, but it does get you the check.

So we said we wanted to go really broad, really get product-market fit, and focus all of our energy on the things that make the product compelling. Then we'd worry about moving up the stack. That's been the goal since day one, and it hasn't changed. The only thing that has changed is that when we started, AI wasn't really applicable to the data world, and then all of that changed rapidly. So we've had to figure out what the future of data looks like in a world with really good LLMs that can write SQL and do most of the things you used to do with the data stack.

What has changed of late, and what do you expect over the next year or two?

About seven or eight months ago, I was on a panel, and someone asked how AI was impacting data. I said it really wasn't, and I didn't think it was going to. That was about two months before Claude 4.6 came out. I had a reason for thinking that wouldn't happen. But first, the models got better. And second, they started working differently. Instead of having to one-shot the exact right query, Claude does what a human would do: poke at your dataset. It runs a bunch of queries and figures things out, like which tables are joinable or which one hasn't been updated, and it ends up coming up with a really good answer.

So LLMs got really good at SQL, and natural language querying went from something that had been around the corner for ten years to something that actually works. At BigQuery, several brilliant engineers banged their heads against natural language queries, and it never really worked. Then, overnight, it just started working. The first thing we noticed was that when you ask Claude to show you your revenue over the last three quarters, it gets the answer and then builds a really nice visualization on top of it. Claude can build charts and graphs that are generally better than any BI tool and much more customizable.

That felt new and different. The reason you still needed a BI tool was that these were just static graphs, with the data encoded right into the graph. So we asked what would happen if we replaced that data with the SQL query. Claude builds the visualization from a TypeScript file, so we started hosting those files in our UI, letting you share them with your coworkers and run them against fresh data all the time. We called them Dives, and it took us about three weeks to build them.

Then we realized this was basically a BI tool, even though we'd started out explicitly saying it wasn't one. Customers were using it as a BI tool and telling us they could turn theirs off. In fact, in 13 days our own BI contract is up, and we're not renewing it because we're using Dives instead. It turned out to be really powerful. Think about how the modern data stack used to look. You had data ingestion with Fivetran, transformation with dbt, your query engine with Snowflake, and visualization with something like Looker. We now have a solution for the two on the right. We have a really good data warehouse, and now we can do BI and build rich, interesting visualizations that go beyond what you'd do in a dashboard, just with natural language, using Claude or Gemini or whatever.

But the stuff on the left is actually pretty vibe-codable as well. If you want to get your data out of HubSpot, you can point Claude Code at the HubSpot API, and it'll figure out how to get that data out. There are also a bunch of open-source connectors that are quite good. dlt has really good connectors. So we asked whether we could do the same thing there. Recently we launched Flights, which covers data ingestion, pipelines, and transformation. You can use natural language to build these pipelines, run them on a schedule or continuously, and then go straight to visualization. Once you do that, you have the whole thing end to end. The one missing piece is context, and that's coming too.

That's a long-winded answer to how things are changing. Parts of the data stack are highly vibe-codable, and Claude can get you an alternative. Other parts aren't. Claude isn't going to vibe code a query engine or the infrastructure underneath it. Even if it could, you'd want to host it somewhere, make sure it's scaled to the right size, and have permissions and all those things. Those are things we can provide. We've spent all of our energy building that part, which I think is the hard part, and from there we can expand into everything else.

So over the next 12 to 18 months, the modern data stack, which had settled into a stable equilibrium over the last decade or so, is going to be blown up. The silos and swim lanes we used to have will be obliterated, and everyone will be going after the same thing. Everybody recognizes that this is happening, and I think it's going to be exciting. The data stack got kind of boring for a while. It felt like the winners had been decided. Now it's anybody's ballgame.

How are you using AI operationally across the company?

I like to think of our tech stack as a wheel. At the center of the wheel, velocity is pretty small, and at the edge, it's much larger. We have core teams like our storage team, core engine team, and hybrid execution team. Those move slowly, and they don't use a whole lot of AI, because it's got to be absolutely right. We want to know exactly what's going into the code and how everything fits together, and we design those pieces very carefully. At the edge, we have the UI team and the AI team, and they're basically running multiple Claude or Codex agents and moving as fast as possible.

It's been amazing to watch all of our intuitions about what takes a long time and what doesn't get totally upended. Then there are the teams in the middle, like infrastructure and platform, that need to move a little more slowly.

We need to keep things reliable and supportable, but they can still use a lot of AI. So we're using a lot of it in the product. Our HR team is building HR apps, which is awesome. We have one and a half people doing HR, and they can build these really nice, branded apps. And we use the heck out of our own product. Pretty much everyone builds Dives, data visualizations in the product. The sales team uses them to figure out which customers to talk to today, how their book is doing, and where they're at on attainment. The support team is building all sorts of them too. That's been super rewarding. We get to eat our own duck food.

As CEO, do you have any cadences where AI helps you day to day?

I don't use it very efficiently. I use our product a lot. I used to be the number one SQL writer at the company, and now I don't write it at all. I use Claude to answer random questions, and I use Claude Code to build stuff and hack on things every once in a while. But I know a lot of people have much better routines where AI does things for them. I don't.

How do you get the best out of yourself?

One of the things I've noticed over the years is that if I don't get enough sleep, I'm just an asshole. It's one thing to be an asshole in normal life, but when you're an asshole as a CEO, bad things start to happen. So I have to make sure I get enough sleep.

The thing that works against that is exercise. If I don't get enough of it, I kind of go nuts. I go stir-crazy. So I row eights, fours, singles, quads, and so on, and I also have my own single. But we often row at 5 a.m., and getting up to row at 5 a.m. and trying to get enough sleep don't always work well together. Sometimes it just means I have to go to bed early and take some time out. I'm not good at disconnecting. But as long as I get exercise and good sleep, I can be a reasonable human being.

Extra reading

• Introducing Flights: Agent-Native Ingest in MotherDuck - June, 2026

• Duck, Dive, and Answer - February, 2026

• Big Data is Dead - February, 2023

And that’s it! You can follow Jordan on LinkedIn or check out MotherDuck on their website to keep up with what they’re building!

BRAIN FOOD 🧠 

TWEETS OF THE WEEK 🐣 

TOOLS WE RECOMMEND 🛠️

Every week, we highlight tools we like and those we actually use inside our business and give them an honest review. Today, we are highlighting Airwallex—Open multi-currency business accounts, collect payments globally, save time and cost, and automate bill pay.

See the full set of tools we use inside Athyna & Open Source CEO here.

HOW I CAN HELP 🥳

P.S. Want to work together?

1. Hiring global talent: If you’re hiring tech, business, or ops talent and want to do it 80% less, check out my startup, Athyna. 🌏

2. See my tech stack: Find our suite of tools & resources for both this newsletter and Athyna here. 🧰 

3. Reach an audience of tech leaders: Advertise with us if you want to get in front of founders, investors, and leaders in tech. 👀