Vibe First, Validate Fast, Verify Fit
Getting proof-of-life, before playing at scale.
Welcome to product discovery in the era of vibe coding—AI-assisted prototyping that looks real but may not be—agentic AI tools, and stakeholder theater—where the wrong prototype costs more than the wrong roadmap.
"The most expensive way to test your idea is to build production-quality software." ~ Jeff Patton, Dual Track Development
Every prototype tells a story. Some whisper feasibility. Some scream, “LOOK AT THIS FIGMA!” Others lie … beautifully, confidently, and often. And your execs believe them. As a product manager, you don’t need to scale yet … but you do need to stop guessing what’s worth building.
Let's fix that. But first, let's reset what we even mean by "prototype" in the year 2026 (that's not a typo, peeps).
🔍 Reset: The Proof-of-Life Probe
Back in 2015, I was stuck in feature hostage negotiations with execs who couldn't tell a prototype from a production build. So I invented a new kind of artifact:
The PoL Probe — Proof of Life
Lightweight
Disposable
Narrow in scope
Brutally honest
Tiny & Focused
Think of them as reconnaissance missions. Not MVPs. Never MVPS!. These tiny acts of discovery (TADs) exist to de-risk decisions … not justify the latest idea bellowed by the HiPPO or swoop-n-pooped by a vibe coding Seagull Manager.
🧠 Choose Wisely (and Cheaply)
Not every problem is a nail in need of a hammer. Meaning, just because vibe code makes it look real doesn't mean it is real. And not everything deserves to be vibe-coded.
So, before you hand over dev hours like Halloween candy, ask:
What risk are we trying to de-risk?
And what's the cheapest, fastest way to hear the harshest truth?
🍦 The Flavors of Prototypes (Now with AI Toppings)
1. Feasibility Checks
Can we even build this?
1–2 day spike-and-delete tests. Think: GenAI prompt chains, LMMs Evals, data integrity sweeps, third-party API sniff tests, or evaluating tool fit. This is about surfacing technical risk—not building to impress.
2. Task-Focused Tests
Can users complete this one job without friction?
Validate the make-or-break moments: a field label, a decision point, a drop-off. Tools like Optimal Workshop and UsabilityHub make this dead simple.
3. Narrative Prototypes
Does this workflow earn a 'hell yes'?
Think Loom walkthroughs, explainer videos, or slideware storyboards. It's not "build vs buy." It's "tell vs test." Tell the story, then see who buys in.
Text-to-video tools like Sora, Synthesia, and Veo3 let you create fast, low-effort narratives—without writing code or crafting clickable prototypes.
4. Synthetic Data Simulations
Can we model this without burning prod?
Use tools like Synthea to simulate users. Leverage systems like DataStax LangFlow to test prompt logic, edge cases, or surface unknowns. It's like a wind tunnel—only cheaper than a postmortem.
5. Vibe-Coded PoL Probes (fka "Hybrid Prototypes")
Will this Frankensoft survive contact with users?
ChatGPT prompts wired through Canvas screens, deployed with a dash of Replit and Airtable? Boom: fake frontend, semi-plausible backend, and user feedback in 48 hours. This isn’t product—it’s just enough illusion to catch real signals. If they use it, great. If they ghost it? Even better—you just dodged a six-month roadmap detour. Just don’t confuse momentum with maturity.
“The biggest risk is building something nobody wants. That’s why we need to validate our assumptions before we invest in building the full solution.”
~ Melissa Perri, Escaping the Build Trap
🧪 Rule of Thumb
Use the cheapest prototype that tells the harshest truth.
If it doesn't sting, it's probably just theater.
🙌 Hat Tip to a Prototype OG
This post owes a debt to Marty Cagan's 2014 blog post Flavors of Prototypes and Chapter 43 of his book Inspired: How to Create Tech Products Customers Love.
While his examples may be dated in today's world of generative assistants and cheap prototyping tools …
... his message is still relevant …
… we need intentional proof-of-life validations, not prototype theater.
🎯 Your Turn
What's your go-to PoL probe when you need to gut-check risk before devs throw laptops?
Drop your favorite flavor—or your biggest fail—in the comments.
Let's validate smarter. Prototype cleaner. And dodge the next swoop-and-poop.