Nate B Jones opens with a Microsoft finding that 86% of people treat AI output as a starting point rather than a final answer, and that 58% of AI users (over 80% among advanced users) are producing work they could not have produced a year earlier. He treats these as good news but quickly pivots to the harder consequence: AI does not just make people more productive, it makes more people look productive. When a memo, prototype, resume, or project plan can be polished by AI, the finished artifact no longer proves that the person understood the situation well enough to make a good decision.
His thesis: this is not a resume problem, it is an evidence problem. The scarce, demonstrable thing in the age of AI is visible human judgment, and the clearest way to surface it is a whiteboard-style conversation where reasoning survives contact with another serious mind.
Before AI, production was hard enough that a finished artifact carried signal. If you shipped a roadmap, wrote a strategy doc, or delivered an analysis, the difficulty of producing it told a meaningful piece of the story about your expertise. AI breaks that link. Generation is largely solved, so a shiny portfolio no longer reliably indicates good judgment.
What we actually need to see is the human reasoning underneath:
Jones calls this "the age of the whiteboard." To know whether someone truly understands a problem, put them at a whiteboard with someone strong enough to push them. The problem should be real and the room should be serious. The person has to draw what they know, name what they don't, explain where the system is fragile, and say where the risk is. Then the other person pushes back.
A whiteboard conversation is valuable because it turns private judgment into visible human work before the work gets cleaned up. The person must think in the room, hold the situation in their head, respond to pressure, update when they learn something, and show where their confidence actually ends. That live reasoning is the evidence that matters now.
The standard career advice, "build a portfolio," is true as far as it goes but incomplete because it targets the part AI already made easy: producing things. Jones is explicit that you should still show you can ship. The addition is that you must also show:
A work sample alone is not enough because publishing it is downstream of all that thinking. What people need to see is the whiteboarding session: your understanding of the problem surviving contact with another serious human mind.
A good whiteboard conversation should showcase four things.
Write down what's happening. Who's involved? What's the system? What constraints matter? What facts do you have, and which are missing? Where is the pressure coming from, and why is it hard? Context is where judgment begins, so show the context.
What are the plausible paths? Which would you take, and which would you reject, and where does the decision sit? Good work involves rejecting genuinely plausible options in favor of what matters most. The rejected options matter because they show what you understood and refused to hand-wave away.
What could go wrong? What risk are you willing to take, trying to remove, or consciously accepting because the alternative is worse? Risk is one of the clearest ways to make invisible work visible, because good judgment done right often looks like nothing happened. Name the preventive wins: the bad launch that didn't happen, the customer who didn't churn, the model output that never reached production unreviewed.
If you make this decision, what's different? What gets clearer, safer, or faster? What work stops? What decision stops being relitigated? What does the team understand afterward that it didn't before? This keeps the exercise from becoming a diary. The point is to connect your judgment to a change in the work, not to record everything.
Part of the goal is to show how you learn. Do people get defensive when challenged? Do they update too quickly just to please the room? Do they hold a useful line when the argument is sound? The target is not perfect confidence or perfect recall. What everyone is trying to see is judgment under pressure.
Jones ties this to his Talent Board project, which started from the same problem. Because AI made building and polishing easy, generation is "solved" and portfolios have less value. The scarce thing is comprehension over generation: explanation as an artifact, a record of real work rather than just credentials.
A resume says you're qualified; a portfolio says what you've made. The better version says: here is the work, and here is the evidence that I understood it, made sense of it, and made good choices as a result. Whiteboarding is the live version of that; the Talent Board is where that evidence lives afterward, as a work sample, promotion note, hiring packet, or record.
Standard onboarding advice (listen, learn the org, meet stakeholders, get quick wins) is fine but incomplete. If judgment is the valuable work, starting strong means forming a point of view early and letting people watch it improve. That doesn't mean being loud; it means putting your early model in front of people who know more. A useful first-month move is to request a whiteboard session with a domain expert: state what you think the customer problem is, where the team is over-weighting, the technical constraint you don't yet understand, and the risk you want to validate. Then let them push back, write down corrections, and ask what evidence would settle disagreements. The goal is to learn in public without becoming mushy.
You don't need a physical board. A shared doc, a digital whiteboard, a Loom video, or an annotated prototype all work. The format matters less than the discipline of making the reasoning visible while it still feels alive, organically surfacing situation, decision, risk, and change.