Skip to content
Dalton + MichaelDalton + Michael

How to Build an MVP in the AI Coding Era

Dalton + Michael revisit the classic startup advice around building an MVP and update it for the AI coding era. When building features with Claude Code or Codex is easy, its easy to make something with a ton of features that is bloated and not what people want. A good MVP is just as much about what *not* to build as it is what to build. Remember: don't stop talking to users no matter how tempting it might be. Dalton + Michael is brought to you by @Standard_Cap Sign up for the StandardDB MVP that Dalton references here: https://www.standarddb.com/ Dalton Caldwell on X: https://x.com/daltonc Michael Seibel on X: https://x.com/mwseibel

Michael SeibelhostDalton Caldwellhost
May 17, 202612mWatch on YouTube ↗

At a glance

WHAT IT’S REALLY ABOUT

Building MVPs with AI means editing ruthlessly, not shipping more

  1. AI coding makes feature creation so fast that founders can accidentally build bloated MVPs in days, requiring aggressive cutting to regain clarity.
  2. The core MVP risk shifts from underbuilding to overbuilding: shipping too many features increases confusion and reduces adoption even if the product is “more complete.”
  3. User conversations matter more, but founders must avoid AI-enabled spam outreach and instead prioritize fewer, higher-quality, high-trust interviews that uncover real underlying needs.
  4. Democratized creation tools increase output (“vibe-coded” products and content) without increasing demand, so standing out comes from focus, taste, and execution rather than volume.
  5. Building in public still works, but AI-generated “slop” creates false positive signals; differentiated, idiosyncratic insights and measured iteration create durable advantage.

IDEAS WORTH REMEMBERING

5 ideas

Treat AI as a feature factory—and yourself as the editor.

Because features are now cheap, the constraint is no longer engineering time but product taste and focus; founders must proactively delete and narrow scope to preserve clarity.

More features can directly reduce adoption.

Dalton’s experience: early users were overwhelmed by a broad product, but after removing ~80% of functionality, users understood it and signed up immediately.

Never implement the user’s entire feature list just because you can.

AI can turn interview notes into a long backlog instantly, but that often encodes superficial requests rather than the underlying job-to-be-done, creating a “tar pit” of complexity.

Do fewer, deeper user conversations—especially now.

When both building and outreach can be automated, the competitive advantage shifts to intimate, high-context conversations that reveal what actually makes customers successful.

Avoid AI-enabled “talking to users” that is really just spam.

Buying leads and blasting messages can generate noisy signals and churn loops that look like traction; high-quality conversations and word-of-mouth flywheels beat brute-force volume.

WORDS WORTH SAVING

5 quotes

Number one, it is so much easier to build features that for me to launch that, I had to delete 80% of the features that I built in the MVP.

— Dalton Caldwell

You see people just vibe coding just crap.

— Dalton Caldwell

I think what I'm arguing is it's more tempting now than ever to not talk to users and to not launch with something small.

— Michael Seibel

You could take notes with Groena of you talking to a user and feed them to Codex, and Codex will build every feature the user asked for. And I'm saying that's really bad.

— Dalton Caldwell

It's become easier for a startup to basically look like it's going through a positive feedback loop when it's actually secretly going through a negative feedback loop.

— Michael Seibel

MVP definition in the AI eraFeature creep acceleration and editing disciplineResisting “build everything users ask for”High-signal user interviews vs automated spamComplexity reduction to increase activationGarageBand metaphor: democratization vs demandBuilding-in-public vs AI content slop and false signals

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

Add to Chrome