I built an agent-to-agent product. Then I killed it.

Hey — Keith here. I'm going to start writing these as I build. Real founder-log, not marketing copy in disguise. First entry: the version of TextMyAgent that isn't shipping, and the version that is.

The version I was building

I've been publicly bullish on an agent-to-agent future for a long time. I still am. Look at what's happening with agent-native operating systems like OpenClaw, where autonomous behavior for individuals is central to the whole OS. Look at endpoints I've built like agents.essentialist.io — a live API endpoint that lets other agents tap in and learn outbound email marketing. Machines negotiating with machines is real, and it's happening faster than most people notice.

So when I first sat down with TextMyAgent, the vision was ambitious. My agent talks to my wife's agent to plan a dinner date. Both agents listen quietly to a group thread and take autonomous action on the backend when someone throws out "let's do something next weekend." Agents that don't just answer questions but coordinate on behalf of the humans in the loop.

That was the pitch. That was v1.

The moment I killed it

My wife and I actually tested this. We connected our agents. We let them try to plan a dinner date on our behalf.

It worked, technically. It also didn't matter. What we found is that we just texted each other. Directly. Because it was easier and because we already knew each other's schedules better than any agent could infer from a Google Calendar.

The market for agent-to-agent isn't broken — the pain wasn't there. It's not that the idea was wrong. It's that the idea didn't hurt.

That was the aha. I killed v1 in my head that afternoon.

The rule I set for v2

Everything I build from here has to be a painkiller, not a vitamin.

Agent-to-agent gets to come back later, when the rest of the world catches up. Right now, in July 2026, most people don't have agents. Even fewer have agents empowered to act autonomously on their behalf. The mental model isn't there yet. It's a market timing problem, not a product problem — and forcing it doesn't work.

So I asked the honest question. What do I actually do 15 times a day that hurts? What's the pain I'd pay to make stop?

The two pains I kept

Two came up immediately. Both mine. Both daily.

1. Signal vs. noise across multiple inboxes. I have six or seven email addresses. Keith@this, Keith@that, Keith@theotherthing. Every day I miss things because they land in a box I'm not watching. I don't want to sit in front of Gmail all day sifting through transactional email, newsletters, and pitches to find the two messages that actually matter. I want a text when my boss emails me. That's it. Everything else can wait until I open the app.

2. Calendar negotiation between too many humans. Seven back-and-forth emails between four colleagues to book one 30-minute meeting is a national disease. TextMyAgent gives every user a personal email address they can forward or CC — and the agent handles the negotiation. Proposes times. Books the meeting. Texts you when it's done.

That's the whole product. Two pains. Solved narrowly.

Nice-to-have vs. painkiller — side by side

Who this actually turned out to be for

Once I re-scoped, something interesting happened: the audience showed up in categories I didn't originally target.

Realtors and loan officers. Speed-to-lead is everything in their world. The realtor who responds to an inbound in five minutes wins the client from the one who responds in sixty. TextMyAgent is a speed-to-lead machine — a text lands the second the lead does.

Blue-collar operators who aren't at a desk. The pool company owner who's actually cleaning pools, not staring at Gmail. The electrician driving between calls. The contractor on-site. They shouldn't have to stop and check email to know a $10K quote request just landed. A text solves it.

E-commerce sellers. I was talking with someone who runs an Etsy shop the other day. Etsy explicitly scores your store on responsiveness — the platform punishes slow reply times in search. Same math applies to Shopify sellers, Amazon FBA operators, anyone whose margin depends on being present to customer questions. Don't stare at your inbox. Get a text.

None of those audiences were in my original ICP. All of them showed up because pain is universal in a way that "coordinate my agents" isn't.

The beautiful vision will come back. Not before the market's ready — and not before I've solved a pain that actually bleeds.

What "developing in public" means

I'm going to publish more of these. Real founder-log stuff — pricing decisions I've second-guessed, features I cut, bugs I shipped that embarrassed me, users who told me I was wrong. I'll write them as they happen, not as after-the-fact marketing.

And if you have a pain the product could solve, I want to hear it. Text the agent, email me, reply to a TextMyAgent update. The fastest way to figure out what to build next is to ask people what actually hurts. That's the whole point of building in public — I don't have the answers alone.

Bottom line

I killed a beautiful vision to ship a simple product. Two pains, solved narrowly, no attempt to sell you a future that hasn't arrived yet.

If your inbox is a signal-vs-noise problem, or if you're wasting a workday a week booking meetings across humans — try it for fourteen days. You'll feel the difference inside a week. And if you don't, tell me why. I'll write about that too.