Howdy Founders,

Let's talk about the thing nobody puts on LinkedIn: shutting down.

If you've ever killed a company, or you're quietly wondering if it's time, I want you to hear this first. It's normal. Like, statistically boring normal. The founders who shut down aren't the exception, they're most of the room.

But there's a new way startups are dying right now, and it's worth paying attention to before you pick your next idea.

The numbers nobody brags about

Here's what the data actually says:

• 966 US startups on Carta shut down in 2024, up about 26% from 769 the year before. And that's only Carta customers, so the real number is higher.

• Q1 2024 alone saw 254 shutdowns, the worst quarter of the decade. (Carta)

• Up to 75% of venture-backed companies never return cash to investors, per Harvard Business School research on 2,000+ startups that raised at least $1M. They returned nothing. Zero.

• 62% of startups in Carta's Class of 2018 were gone within 7 years. Only 1.3% become unicorns. (SaaStr)

So call it roughly 800 to 1,000 venture-backed shutdowns a year on one cap table platform, and climbing. Shutting down isn't failing at startups. It's mostly just... doing startups.

But here's the part I find more interesting. SimpleClosure, a company that literally helps startups shut down, put out its 2025 report and called it the “first major wave of AI company shutdowns.” Series A shutdowns jumped from about 6% to 14% of all closures. These aren't half-baked weekend projects. These are funded teams with customers.

Their read on what survives? Companies with proprietary data, real unit economics, and deep workflow integration. Not tools that sit on top of someone else's platform.

Which brings me to Relay.

Relay: the right idea, a little too early

If you never used Relay.app, it was a Zapier competitor (with the perfect name) with one really smart twist. Instead of chaining automations end to end and praying, Relay let you drop a human approval step right in the middle. A draft waits for your review. A refund waits for a manager. An email sits until someone says go.

That's a genuinely good insight. It's basically the exact thing everyone wants from AI agents today: do the work, but check with me first.

The pedigree was there too. Founder Jacob Bank sold his last startup (Timeful) to Google, then ran product for Gmail and Google Calendar. Relay raised $8.1M from Khosla Ventures and a16z.

In July, customers got the email. Free accounts were deleted August 15. Paid accounts went dark September 14. Today the homepage just says Relay has shut down.

Relay never said why. But a month later, Bank announced he was rejoining Google as VP of Product for Chrome, bringing several Relay teammates with him, with "really ambitious plans" for working with AI in the browser.

TechCrunch's take was blunt: as OpenAI, Google, and the other giants built similar automation directly into their own products, Relay struggled to justify existing as a standalone product. And look at what exists now. Claude Cowork, ChatGPT's agents, Gemini in Chrome. The "AI does the work, human approves" loop isn't a product anymore. It's a feature of the thing you already pay for.

Great idea. Great team. Great investors. And the idea got so good that the biggest companies on earth made it a default setting.

Huxe: dead in 24 hours

Relay at least got a slow fade. Huxe got a Friday.

Huxe was started in late 2024 by some former Google folks behind NotebookLM (yes, the thing that made AI podcasts go viral). The app let you type a prompt and get a full podcast, or a whole series, about anything you wanted to learn.

They raised $4.6M from Conviction, Genius Ventures, Figma CEO Dylan Field, and Jeff Dean. If you're picking a team to build AI audio, it's hard to pick a better one. They basically built the category back at Google.

On May 21, Spotify shipped a personal podcast feature that does basically the same thing. On May 22, Huxe announced it was shutting down. App pulled from the stores, data deleted a week later. THE NEXT DAY…..

And Spotify wasn't even the first. Adobe, Amazon, ElevenLabs, and Meta had all shipped their own version. The product was a feature in five other companies' roadmaps, and once it shipped in the one app people already open every day, the standalone app had nowhere to stand.

The pattern

One more quick one. Clockwise was an AI calendar assistant that protected your focus time and shuffled meetings around so your day didn't look like confetti. It had been around since 2016 and had real traction in tech companies.

In March, the team announced it was joining Salesforce to work on Agentforce, and the product shut down about a week later. (Carly) Meanwhile, Google Calendar already gives Workspace users Focus Time for free.

See the pattern?

1. Smart team spots a real problem

2. Builds a genuinely good AI product on top of someone else's platform

3. The platform notices the problem too

4. The platform ships it as a feature, for free, to a billion users

5. The startup either shuts down or the team gets absorbed into the giant

None of these founders did anything wrong. They weren't lazy, they weren't dumb, their ideas weren't bad. Their ideas were so good that the companies with infinite money and distribution built them too.

That's the new risk. It used to be "what if nobody wants this?" Now it's "what if everybody wants this, including OpenAI?"

Build something they can't ship in a Tuesday update

Here's my honest takeaway. If you're picking your next thing, ask one question before anything else: could Anthropic, OpenAI, or Google ship this as a feature next quarter?

If the answer is yes, you're not building a company. You're building their roadmap and paying for it yourself.

And if you're one of the founders who already shut something down this year, please don't read that as "you should have known." Nobody knew. Relay's founder ran Gmail. Huxe's founders built NotebookLM. If they couldn't see it coming, you weren't supposed to either. Shutting down is just the tuition.

The stuff that's genuinely hard to wipe out right now is boring and physical. No model is going to pull up to a cracked driveway in Leander and seal an expansion joint. I started Hill Country Slabs in February and sold it about 20 weeks later, and no feature update was ever going to compete with it.

That's also why I built Groundwork. It's a community for tech operators who want to start a future proof trades business from zero, on the side, and grow it until it replaces their tech income. You learn to quote, sell, and manage jobs while experienced subcontractors do the actual work, so you don't need to learn a trade or hire a crew.

You get the community, the playbook, the same software Hill Country Slabs ran on, and me working directly with you on your quotes, pricing, and sales calls. Cohort #1 starts November 1st and it's capped at 25 people. Several spots are still open. Apply at groundworksociety.com or just hit reply and tell me you’re interested in learning more.

💡 Business Idea Of The Week: SaaS Shutdown Migration Service

• Every time a tool like Relay or Clockwise dies, thousands of businesses get a few weeks' notice to move everything

• Relay's export was JSON that no other platform could import, so customers had to rebuild their workflows by hand

• Offer done-for-you migrations: audit what the dead tool was actually doing, rebuild it on n8n, Make, Zapier, or an AI agent, and hand it back working

• Flat fee per workflow, rush pricing when the deadline is close

• Lead gen is free: monitor shutdown announcements and the "[Tool] alternatives" posts that pop up the same day

• Upsell a monthly "platform risk" retainer that keeps exports current and a backup plan ready

• With the shutdown numbers above, the supply of new customers isn't slowing down

THE END

As always, thank you for reading. I promise to continue writing these without any sort of regular cadence 😀 

I love you