Two-Question New Technology Adoption Test
Before adopting any new AI tool, ask: how big is the gain, and how painful is the exit?
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 3
- Confidence
- 90%
A two-question decision gate for whether to adopt an unproven new technology (a new framework, protocol, or tool). Use it any time you're tempted to bet on the latest thing. It weighs the realistic upside against lock-in risk, and heavily discounts anything not yet battle-tested.
Origin
Huyen describes the questions she asks people who are torn between competing new technologies, using the MCP-versus-agents-protocol debate as the recurring example.
Core principles
- 01Upside must be measured as optimal-vs-nonoptimal delta, not raw excitement.
- 02Switching cost is a first-class input, not an afterthought.
- 03Unbattle-tested tech carries hidden lock-in risk that should raise the bar for adoption.
How to run it
- 1
Quantify the upside
Ask how much of the improvement you could get from the optimal solution versus a non-optimal one. If the gap is small, the decision barely matters.
- 2
Price the exit
Ask how hard it would be to swap this technology out for another later. High switch cost plus low maturity is a red flag.
Pro tip If the honest answer is 'a lot of work to switch out,' treat adoption as a near-permanent commitment.
Watch out A tool that isn't widely tested yet may leave you stuck with it forever.
- 3
Bias toward the battle-tested
Think twice before over-committing to anything that hasn't been proven by a lot of users.
In the wild
Asked which protocol was 'better,' Huyen ran both questions: the performance delta was often small, and switching later would be a lot of work, making an early bet on the unproven option unattractive.
→ Adoption deferred until the technology is battle-tested.
Common mistakes
Adopting on novelty
Choosing the newest technology because it's exciting, without pricing the switching cost or the real gain.
Is it for you?
Best for
Engineering leads and founders deciding whether to build on an emerging AI tool or protocol.
Not ideal for
Throwaway prototypes where lock-in is irrelevant and speed of trial is the only goal.
From the transcript
“if you adopt a new technology like how hard it would be to switch that out to another”
“It hasn't been tested by a lot of people. And if you adopt it, you would be like stuck with it forever.”
From the episode
Al Engineering 101 with Chip Huyen (Nvidia, Stanford, Netflix)