Surfacing and staking is an AI governance framework for keeping human judgment in charge: you stake a position, let AI surface the counterarguments, then re-stake and own the call.
Learning isn’t broken, we’re just teaching it incorrectly. AI has reformed the landscape of tech from two sides of the career ladder. On the one hand, you have employees working inside of companies that feel lost, disincentivized, jaded about management. According to Glassdoor, employees are increasingly distrustful of management and increasingly negative when companies mention AI.
[The] rapid increase in discussion of AI is turning negative. Workers are concerned by leaders demanding workers use AI while simultaneously touting AI as a reason for layoffs and reduced hiring.
— Glassdoor Worklife Trends 2026: Midyear Check-in
On the other end of the career ladder, students are increasingly anxious about what opportunities will be available to enter into a workforce that’s rapidly adopting AI.
As one graduate put it, “AI appears to be a skill you teach yourself at your own risk. When I was in school, my professors were strongly against it, but now employers expect you to have already mastered it.”
— AI and the workforce ahead, Handshake
In June I published two companion pieces on Substack and Linkedin that addressed this trend in anxiety by both speaking to it and recommending a framework to solve it.
- The Test Was Already Broken – An examination of the reintroduction of exam proctoring at Princeton after 133 in part because AI made cheating so rampant.
- A Framework for Solving Problems, Not a Subscription, Not a Software Solution – An exploration of the surfacing and staking protocol as a way to give basic governance to AI.
Keep Human Deciding
The surfacing and staking framework is deceptively simple. The premise is most people use AI like they do a search engine — input and then output. What the framework says is that decision is play to strengths. At any point in a decision framework whether learning something new, writing a paper, or doing work inside of a team can take two modes:
-
Surfacing – unearthing possibilities
-
Staking – Forming opinions
Our framework supposes that while both AI and humans are good at surfacing, AI is deceptive in that it appears to stake opinions but has no accountability.
The idea is to build an AI governance framework for people who don’t know what that means.
Don’t ask AI for the answer. Hand it your draft to attack.
Steps:
- Stake – Write the rough claim yourself. Be wrong on purpose.
- Surface – Ask AI to undermine it — counterarguments, gaps, what you missed.
- Re-Stake – Take the critique. Make the call. You own the final version.
Try it out on GitHub
I’ve published the framework on github. By putting it out into the wild, I want to make it extensible. Students can use it as a way to force thinking through a paper assignment, or a team can use it as an ideation tool to guide sensible outcomes without outsourcing the thinking.
Try it out yourself the github link is below:
https://github.com/natecooper/surfacing-and-staking
https://github.com/natecooper/surfacing-and-staking/blob/main/SKILL.md
