Build for the Future Model
Design product ideas for the model capability that is coming, not the one you have today
- Difficulty
- Advanced
- Time to result
- ~ongoing to results
- Steps
- 3
- Confidence
- 82%
A strategic posture for building on top of fast-improving AI: choose product ideas whose value unlocks as models get better, so that by the time capability arrives the product works beautifully. Rather than being blocked by today's model limits, you position ahead of the capability curve. Nguyen cites Canvas/artifacts as ideas that predated the model quality needed to make them shine.
Origin
Karina Nguyen's stated design philosophy, illustrated with the arc of Claude artifacts / early Canvas concepts that predated ChatGPT and only became excellent once model coding quality caught up.
Core principles
- 01Model capability is a rising tide — build the boat now so it floats when the tide comes in.
- 02It doesn't matter whether the model is good enough right now; it matters that the idea works well when it is.
- 03Fast-iterating teams that invent new training methods and listen to users ride the capability curve best.
- 04Form factor should follow the function the future model will enable, not just today's constraint.
How to run it
- 1
Identify a product idea gated by current model quality
Find a valuable experience that today's model can only do partially — the limitation is model capability, not the concept. Early coding-in-Canvas and Claude artifacts were like this before models could make high-quality edits.
- 2
Design the form factor for the capability that's coming
Choose a form factor that will scale gracefully as the model gains new capabilities (better search, more creative writing, rendering apps). Tasks was designed as a general feature that 'scales very nicely' as models improve.
Pro tip Pick familiar symbols (a doc like Google Docs, notifications, reminders) as the shell so that when the 'magical' model capability lands, adoption is frictionless.
- 3
Iterate fast and let improving models fill the gap
Ship, listen to users, and keep iterating so that as model quality rises your product is already positioned to capture it. Speed of iteration is what lets you exploit the curve.
Watch out Being too early with a rigid form factor can leave an idea dormant for years — Nguyen notes some ideas sit unrealized for ~2 years until the model catches up. Keep the idea alive and cheap until capability arrives.
In the wild
Nguyen notes early Canvas-like coding ideas existed back around 2022 before ChatGPT, but Claude 1.3 couldn't make high-quality edits yet. The concept was right; the model wasn't there.
→ When model coding quality improved, the pre-designed form factor (Canvas, artifacts) became powerful — the idea was ready to capture the capability.
Nguyen points to startups like Character iterating extremely fast, inventing new ways to train models and listening to a massive user distribution.
→ By moving fast and staying ahead of the curve, they capture model improvements as they land.
Common mistakes
Judging an idea only by whether today's model can do it
Killing a product idea because the current model does it poorly ignores that models improve monthly. The right question is whether the idea works well once the model is good — many winning products looked mediocre at launch-time capability.
Shipping a rigid form factor too early and abandoning it
Being early with an inflexible design can strand an idea for years. Keep the concept alive and cheap to revive rather than fully building and discarding it before the model catches up.
Is it for you?
Best for
Founders and product leaders building on top of frontier LLMs who must place bets that pay off over a 1-3 year capability horizon
Not ideal for
Teams needing near-term revenue or reliability from a shipped product today, who can't afford to wait for capability to arrive
From the transcript
“you kind of want to like build for the future”
“it doesn't necessarily matter whether the model is good or not good right now, but you can build product ideas such that like by the…”
“I see tasks as like this like um general kind of like feature that will scale very nicely as the models would develop like new…”
From the episode
OpenAI researcher on why soft skills are the future of work
Karina Nguyen (Research at OpenAI, ex-Anthropic)