The Technology Wave Specialization Ladder
Ratchet down on a new technology while it's still weird — teardown, no-code, then fun build
- Difficulty
- Moderate
- Time to result
- ~months to results
- Steps
- 5
- Confidence
- 89%
Vo's method for getting ahead of a platform shift the way early PMs did with mobile: pick the emerging technology, specialize deliberately, and learn it through a three-rung ladder that works even if you can't write code — outside-in product teardowns of what exists, no-code/low-code stitching, and finally building something purely fun with the technology as a learning mechanism rather than a commercial project. Her own ChatPRD was rung three: a nights-and-weekends joy project that also taught her how to build non-deterministic AI products.
Origin
Claire Vo's own approach, explicitly modelled on the mobile wave — 'if you were a PM that jumped on mobile you had the pick of the litter' — and validated by her own ChatPRD build, which started as a prompt and became a paid standalone product built on nights and weekends.
Core principles
- 01Platform shifts create a window where specialists get the pick of the litter — mobile did it, AI is doing it now.
- 02You do not need to be an engineer to specialize; you need reps against the new technology.
- 03Build products WITH the technology, not just products about the technology — non-deterministic products are a genuinely different craft.
- 04Learning does not have to be commercially driven or work-related. Fun is a legitimate and effective learning mechanism.
- 05Prompt, instructions and context materially determine output quality — treat them as product surface, not incidental config.
- 06Do competitive analysis on your own build the way you would on any product: same input, compared across the alternatives.
How to run it
- 1
Do outside-in product teardowns
Take products already shipping on the new technology and crit them from the outside in: what is good about this, what is bad, how would I have written the PRD, what would I be measuring, how would I think about the error states, is this a great product, a good product or an okay product?
Pro tip This is the most accessible rung — it requires no code and immediately surfaces where your own skill gaps are.
- 2
Stitch something with no-code / low-code
Even if you cannot put hands on keyboard and write code, you can certainly stitch things together with no-code and low-code tools. Get a working artefact into the world so the technology's real constraints and failure modes become tangible.
- 3
Find where it's fun and build art out of it
Pick a space you're personally interested in and play. Vo points at how fun Midjourney and the creative tools are, and says to build art out of it as a mechanism for learning. It doesn't have to be commercial and it doesn't have to be part of work.
Pro tip Vo's own rule for ChatPRD: it has to be 100% fun for her — a pure bliss space, a hobby. She will not do anything to it that makes it not fun.
Watch out The moment you attach commercial pressure to a learning project you often kill the exploration that made it valuable.
- 4
Take the prompt seriously as product
When building on models, the prompt, the instructions and the context genuinely determine output quality. ChatPRD began as a single prompt lovingly crafted over several months. Then run competitive analysis on it like a PM: same input across the raw model, the GPT-store version, other tools in the category, and yours.
Pro tip Vo added per-user customization — an assistant that learns from each user's content, role and company — as the differentiator over the shared GPT version.
Watch out 'Prompt engineering isn't a real thing' is a fashionable claim and it is wrong in practice — prompt really does matter.
- 5
Ship it and set a joy-first success bar
Publish, put a reasonable price on it, and see what happens. Set a goal small enough that the project stays enjoyable — Vo's original stated goal was to buy one nice glass of wine a week.
Pro tip Small monetary goals let you honestly evaluate demand without turning a joy project into a venture obligation.
Watch out Monetization infrastructure for GPT-store-style products is still not solved out of the box; expect to build the standalone app if you want to charge.
In the wild
At a previous company, a critical, technical product needed a spec and there was no platform PM available. Vo raised her hand and, between the beginning and the end of the meeting, used ChatGPT and a prompt to produce a serviceable PRD. She refined that prompt over several months into a crafted 'product leader' persona, released it as a custom GPT for her team, hit the GPT store's monetization and access wall, and stood up a standalone app on the assistants API with per-user customized assistants and document generation.
→ Thousands of daily users; roughly 60% generate a PRD from an idea, 30% improve an existing spec/strategy/roadmap, the rest brainstorm and do internal PM work. It went from a 'glass of wine a week' goal to real money, while staying her hobby and joy space.
Vo's reasoning by analogy: when mobile happened, PMs who jumped on it early had the pick of the litter when it came to jobs at very interesting startups. She sees AI as the same moment — ratchet down, specialize, learn the technology, and you can get into a very interesting position.
→ Her own dual motivation for ChatPRD: understand how AI impacts the function she leads, AND learn how to build great products on underlying technologies fundamentally different from anything she'd built on before.
Common mistakes
Waiting until you can code to start
The ladder is deliberately designed so the first two rungs need no engineering. Teardowns and no-code stitching produce real learning; treating 'I'm not technical' as a blocker forfeits the window.
Dismissing the prompt as incidental
The industry cycled through 'prompt engineering is a thing / isn't a thing.' In practice the prompt, the instructions and the context drive output quality — ChatPRD IS a prompt, lovingly crafted over months.
Only learning through commercially-justified projects
Requiring a business case for every experiment kills the exploration. Fun is a legitimate learning mechanism, and Vo protects ChatPRD's fun explicitly because that is what keeps her building.
Is it for you?
Best for
PMs and product leaders — including non-engineers — who want to get ahead of the AI platform shift and are willing to spend nights and weekends building
Not ideal for
People with no discretionary time or energy for a side project, or those seeking an immediate venture-scale business rather than a learning vehicle
From the transcript
“back to like when mobile happened if you were a PM that jump jumped on mobile you had the pick of the litter when it…”
“I love these this idea of doing outside in product teir Downs like what is good about this what is bad about this how would…”
“find where there's something fun and build art out of it as a as a mechanism for learning”
“she may just be a prompt but she is my prompt this is lovingly crafted over several months”
From the episode
Bending the universe in your favor
Claire Vo (LaunchDarkly, Color, Optimizely, ChatPRD)