Now What? — Issue #8

Hello fellow curious people!

I’m writing at a time of great flux and uncertainty. Old ways are failing and new ways are unproven. Last year I threw myself into a series of experiments and learning projects to get a feel for a new medium, to develop informed opinions, to demystify some hype, and to find out for myself if I could now build a kind of product that could assist people like myself in doing the kind of product work I do (product management, strategy, consulting).

A quick proof of concept became a rough prototype that encouraged me to restart with a more considered architecture, around the time of the summer solstice last year. In July I started blogging about my experience, at first daily until people told me that was TMI, and a few weeks later I started writing and then sharing publicly Weekly Ship updates to the project’s Shipping News, a custom I nicked from 18F.

Eleven months ago I thought the product was ready for alpha. I closed the Alpha milestone and started one called MVP. I determined not to release the beta till the minimum valuable product was functioning. I initially projected about a two-month effort to reach that milestone and enable the start of beta. I was wrong.

I won’t fill up today’s newsletter with everything I learned and every mistake I made in the past eleven months but I will share that more than once I found myself questioning a core premise of the product that I had left unexamined in an era when all of your priors must be challenged mercilessly.

• I questioned if there needs to be web UI at all, and started a skunkworks project to explore an MCP-connector-plus-skills based plugin architecture instead for what I call the “bring your own chat” (BYOC) experience

• I questioned why we were using brittle deterministic methods to route user intent, instead of the LLM. This led to a complete inversion of the conversation stack (still being executed)

• I questioned why we were building a signup wizard and custom auth connectors in an era of existing MCP connectors for all of our integrations and conversational onboarding

• I deeply questioned what unique value proposition I am developing beyond what is already provided by commodity platforms today (or will be soon).

Even now I question whether Piper Morgan is actually what I need, let alone anyone else. For quite some time I’ve run an experiment with a role on the team called Piper Alpha. I’ve always felt that one bar Piper Morgan has to clear to be worth anyone’s time is to provide more value than a well-prompted role in a general-purpose assistant can.

Piper Alpha is Piper Morgan not as a software product but as an expression of my working method. Piper Alpha took the lead on the initial MCP skunkworks spike. This meant the Lead Developer agent wasn’t distracted (much) by this probe and it also tested whether the assistants have reached a level where they could drive a minor initiative on their own.

That’s the bar Piper Morgan needs to clear. It can’t just break even vs. that or even beat it by just a little for the same reason that enterprise products have always needed to give people a reason to migrate away from their custom spreadsheets and keystroke shortcuts.

I gave a presentation recently for a Rosenverse Live event in the Rosenfeld cinematic multiverse, and Lou Rosenfeld asked me to predict what my topic would be in a hypothetical return visit six months hence. I blurted out “Why it was all such a big mistake, but here’s my cool new idea.”

I was kidding but I was also not kidding. I’ve always said my first goal was to learn. Ka-ching. My second goal was to share what I was learning with my community. Mission accomplished. Making a product useful for me was goal three and making one useful for other people goal four. Those are the ones that have not proven true yet, although I have developed a method that is useful for me and that keeps threatening to be the real story.

At that same talk Peter Merholz asked me about the large (11-role) agent team that plans, builds, tests, and organizes Piper Morgan for me, and all the processes that have evolved to manage them, saying (in that blunt way we have all come to love) “are you avoiding shipping?” and I gave it some real thought.

There is an element of play where the journey is the path and the real product is the friends we made along the way, you might say, but at the end of the day I am still a product person and my methods are only as good as their ability to deliver things people really need, want, and will use.

On the other hand, I was also able to announce that day (last Friday, October 30) that I had finally closed the MVP milestone and was now in the process of rolling out a private beta. This is another moment of truth and inevitable questioning. Multiple real humans will point out flaws, extraneities, and opportunities to improve that have escaped me or that I have ignored or even avoided noticing.

I’ll find out what sticks, if anything, and probably what I can let go of as well.

I have a hypothesis that Piper Morgan can provide a reliable, trustworthy, constantly improving form of prosthetic assistance with product work. This raw idea is about to be plunged into the crucible of real lived human experience and what emerges afterward will either be a fiercely honed instrument or a valuable lesson to take forward, share, and integrate into my next move.

Oh, and none of my other projects use so many named roles running loops. Piper Morgan has a unique peripatetic history but it’s the T-Model Ford of my operating process. Nearly all of my projects have a single agent or two at the most.

Sometimes not even that. For example, I’ve got a new early-stage product idea (codename “Pimento”) I’m probing in a lean way that arose from the convergence of two trains of thought:

1. contemplating what the YC people really think they mean when they say they are looking for the “Cursor for product”

2. continually rethinking my core questions about what product professionals need from a coherent assistive product as well as what other professionals need when they don’t have a human product manager available.

My business agent Themis manages that one for me for now, no new agents needed (yet).

—xian

P.S. Piper Morgan’s private beta is open. If you do product work (or work without a product manager and wish you had one), and you’re willing to put up with some imperfections and tell me what’s wrong, join the waiting list at pipermorgan.ai/try/beta.

P.P.S. Have you built something for yourself that you suspect shouldn’t be a product? Or one that should? Hit reply and tell me.

P.P.P.S. Next issue, on November 17: what I learned trying to teach my agents to disagree with me.