Stamp Thinks Your Email Client Should Think For You

The Macro: Email Refuses to Die and Nobody Can Fix It

Email is the cockroach of productivity software. People have been predicting its death for 15 years and it keeps growing. There are over 4 billion email users worldwide. Slack didn’t kill it. Teams didn’t kill it. Discord didn’t even dent it for professional use. Email is permanent infrastructure, like plumbing, and about as pleasant to deal with.

The email client space has seen waves of challengers. Mailbox got acquired by Dropbox and shut down. Inbox by Google got killed. Spark has a loyal following but never broke through to mass adoption. Superhuman proved you could charge $30/month for email and built a real business around speed and keyboard shortcuts. Hey from Basecamp took a philosophical stand about email workflows and found a niche. Newton, Airmail, Canary, Edison. The list goes on.

Now AI is the new angle. Every email client is adding AI features. Superhuman has AI-powered summaries and drafting. Shortwave rebuilt around AI search and summaries. Even Gmail has Gemini baked in now (though it’s clunky and most people ignore it). The question for any new entrant is: what does “AI native” mean beyond marketing copy? Is there a version of an email client where AI isn’t a feature but the foundation, and does that produce something people actually want?

The Micro: One Founder, One Inbox, One Big Bet

Stamp is built by Archit Mehta, a solo founder out of San Francisco. That’s it. One person. He’s part of YC’s Winter 2025 batch, working with Jared Friedman, who has a track record of backing developer tools and productivity plays.

The product bills itself as “the first AI Native inbox.” The core features: it composes emails in your personal voice (not generic AI voice, but yours, based on your writing patterns), organizes your inbox using plain English filters (like “put all newsletters in a folder and summarize them weekly”), and extracts action items from incoming messages automatically.

The AI-as-copilot comparison is intentional. Stamp frames the experience like AI coding assistants: inline completions as you type, the ability to highlight text and ask the AI to revise it, contextual suggestions based on the thread. If you’ve used Cursor or GitHub Copilot for code, imagine that interaction model applied to email composition.

Stamp is hiring a founding engineer specializing in email and messaging, offering $90K-$120K and 5-10% equity. That’s a serious equity stake, which tells you two things: the team is genuinely tiny and the right early hire would be building core infrastructure, not polishing edges. Early signups get a year free.

The solo founder dynamic is worth noting. Building an email client is an enormous engineering challenge. Email protocols are old, messy, and full of edge cases. IMAP alone is a nightmare. Add AI on top and you’re building two hard things simultaneously. One person can prototype this. Shipping a production email client that handles the full complexity of real-world email is a different story.

The Verdict

I think the “AI native” framing is the right one for email. The bolt-on approach (here’s your regular inbox, plus an AI button) feels incremental. If you rebuild the client around AI from the ground up, you can do things like automatic action item tracking, voice-matched drafting, and intelligent filtering that would be awkward retrofits in a traditional client.

The competition is fierce and well-funded. Superhuman has been iterating on AI features for over a year. Shortwave raised $9 million and has a strong AI-first approach. Both have teams, users, and momentum. Stamp is one person with a prototype.

But email clients have a history of rewarding focused, opinionated products. Superhuman proved that a small team could build a premium email experience that people pay real money for. The market is large enough for multiple players if the product is differentiated enough.

In 30 days, I want to see how accurate the voice-matching is. If it writes emails that sound like me after a week of learning, that’s compelling. If it sounds like generic ChatGPT output, it’s dead on arrival. In 60 days, the question is reliability: does it handle weird email edge cases (forwarded chains, HTML newsletters, calendar invites embedded in threads) without breaking? In 90 days, I want to know if users are actually sending AI-drafted emails without editing them. The gap between “AI suggested something” and “AI handled it” is where the real value lives. Stamp has the right thesis. Now it needs about four more engineers and a year of brutal email protocol debugging.