> For the complete documentation index, see [llms.txt](https://cafebedouin.gitbook.io/potm/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://cafebedouin.gitbook.io/potm/section-1-the-crisis-of-expertise/1-the-mullahs-map/003_the_valuation_crisis.md). # The Valuation Crisis: When Good Performance Looks Like Bad Performance > "Talent, ideas and talk are cheap, but discipline, execution, and action are expensive. Choose wisely." *"Talent, ideas and talk are cheap, but discipline, execution, and action are expensive. Choose wisely."* A friend who works in marketing called last month, genuinely shaken. Her firm had started using AI to generate the reports she used to spend days creating—market analysis, competitive positioning, campaign recommendations. The AI's output wasn't perfect, but it was good enough. And it was instant. "I kept telling myself my work was strategic," she said. "But looking at what I actually do all day... it's mostly synthesis. I read things, I organize information, I write it up clearly. The AI does that now. What am I actually worth?" As we saw in the previous essay, organizations default to evaluating Stage 1 signals—credentials, frameworks, polished presentations—because that's what they can measure. They can't reliably detect Stage 3 capability because evaluation itself requires the thing being evaluated. This creates a specific economic crisis when AI can produce Stage 1 output for nearly free. She'd discovered something deeper than "Will AI take my job?" She'd realized that what she called "strategic work" was actually pattern-matching over frameworks—and that our professional world had been rewarding exactly that kind of convincing performance for decades. AI didn't create this problem. It just made it impossible to ignore. ### The Environment Was Already Broken Before AI, we already lived in an environment where: * **Credentials replaced capability**: An MBA signaled you could think strategically, whether or not you'd ever successfully executed strategy * **Frameworks replaced judgment**: Knowing the latest consulting model mattered more than knowing if it fit your actual situation * **Smooth explanations beat messy reality**: The person who could tell an elegant story about why something failed was more valuable than the person who prevented failure through unglamorous attention to detail * **Years meant expertise**: Ten years in marketing assumed skill, regardless of whether those were ten years of learning or one year repeated ten times * **Presentation mattered more than outcomes**: The person who could write compelling documents got promoted over the person who quietly shipped things that worked This wasn't AI's doing. These were human systems, built and maintained by humans, that systematically chose people who were good at **looking like they knew what they were doing** over people who actually did. AI is just better at producing what these broken systems select for. ### Three Stages of Getting Good at Something Remember the framework from earlier—how people develop capability through three distinct stages: **Stage 1: Learning the Information**\ You can read, understand, and explain. You're working with ideas—accurate descriptions of things you may have never actually done. *Example: A business analyst reads industry reports and synthesizes them into a clear document explaining market dynamics.* **Stage 2: Awkward Practice**\ You're trying to actually do it. You know what you're supposed to do, but it doesn't come naturally. You're thinking through every step. You make mistakes. You're paying the costs of learning. *Example: The analyst starts running actual experiments—testing prices, launching campaigns, making real decisions with real money at stake. They know the theory, but doing it feels clunky.* **Stage 3: Integrated Skill**\ You've practiced enough that it's automatic. You have judgment from seeing patterns across many cases. You can sense when something's wrong before you can explain why. You've lived with your decisions long enough that it shapes how you see new situations. *Example: After years, the analyst can look at data and immediately sense opportunities or risks that don't show in the numbers. They've seen hundreds of campaigns. They're not working from ideas—they know the territory.* For the last 150 years, the economy mostly valued Stage 1 work. We built entire industries around people who could read, understand, synthesize, and explain. AI just revealed that most of what we called "Stage 3 expertise" was actually Stage 1 performance with better packaging. ### What AI Shows About Our Evaluation Systems The evaluation problem creates a specific vulnerability: when the signals you use for assessment become cheap to produce, your entire selection system collapses. AI can produce Stage 1 output—accurate synthesis, clear explanations, well-organized information—at massive scale for basically free. It can also do genuinely superhuman work with patterns: seeing connections across millions of sources, applying frameworks consistently, finding relationships in data that humans would miss. But here's what AI reveals: **most of what passed for "expertise" in our economy was already Stage 1 work with expensive credentials attached.** The strategy consultant who spent three days producing market analysis? Mostly Stage 1 synthesis with an MBA credential. The executive who made decisions by applying business school frameworks? Stage 1 pattern-matching dressed up as judgment. The therapist fresh out of grad school who knew all the right models but hadn't developed the feel for when to use them? Stage 1 knowledge performing as Stage 3 expertise. The problem wasn't that these people were frauds. The problem was that **our institutions couldn't tell the difference between convincing performance and actual capability**, so they rewarded both equally. Actually, they often rewarded performance more because it was: * Easier to credential (degrees, certifications) * Easier to evaluate (does this match the framework?) * Easier to scale (hire people who can apply the model) * Less risky for evaluators (can't be blamed for hiring the Harvard MBA) AI didn't break this system. AI just got better at playing the game the system was already playing. ### The Real Economic Problem Here's what my marketing friend discovered: **AI can produce decent Stage 1 synthesis for basically free.** Your cost to produce the same report is your hourly rate times how long it takes. You can't compete. But more fundamentally: **AI revealed that her "strategic work" never required reality-checking anyway**. She was synthesizing information into frameworks—work that could be done entirely with information without ever testing whether it matched reality. This is the actual crisis: not that AI is replacing human capability, but that it's showing how much of what we credentialed and rewarded as capability was just sophisticated pattern-matching over documented knowledge. The question isn't "What happens when AI can do my job?" The question is: "What happens when I realize my job never required the thing I thought made me valuable?" ### Where Practice Can't Be Made Realistic Enough AI can't: * Develop judgment from living with results of decisions * Check whether ideas match reality based on lived experience * Use discernment based on unspoken values and context * Choose between options when stakes are real * Build pattern recognition from years of practice in varied conditions But here's the problem: **our evaluation systems can't reliably detect these limitations either.** Consider hiring for a senior strategy role. Two candidates: **Candidate A**: Five years in strategy consulting. Can explain frameworks fluently. Produces elegant analysis. Has never been responsible for executing strategy or living with outcomes. Pure Stage 1 capability with excellent presentation. **Candidate B**: Five years as product manager, launched multiple products, some succeeded and some failed. Has developed pattern recognition from consequences but struggles to explain their judgment in framework language. Stage 3 capability with mediocre presentation. Who gets hired? Usually Candidate A—because our evaluation systems select for people who can perform expertise convincingly, not for people who've actually developed capability through consequences. **AI can now do what Candidate A does.** The problem isn't that Candidate B is suddenly more valuable. The problem is we built an economy that couldn't tell them apart in the first place. ### When You Can't Tell If You're Any Good I wrote about this in 2024: when you're competent at many things, you lose the ability to accurately assess your own expertise. You mistake "I can talk intelligently about this" for "I have real capability in this area." AI makes this systemic. When anyone can generate sophisticated-sounding analysis on any topic, **producing that analysis no longer signals expertise**. But our institutions still treat it as if it does. They still credential based on coursework rather than consequences. They still promote based on presentation rather than outcomes. They still evaluate based on frameworks rather than reality-checking. This creates a destructive cycle: 1. The environment rewards convincing performance over actual capability 2. People optimize for what gets rewarded (performance) 3. Actual capability development requires paying costs (time, failure, consequences) that don't get rewarded 4. Capable people get out-competed by people better at performance 5. The system degrades—less actual capability, more sophisticated performance of capability AI accelerates this because it's **really good** at producing the performance the system selects for, making the gap between performance and capability even harder to see. ### What Actually Has Value That's Hard to Replace If Stage 1 synthesis is cheap and our evaluation systems can't tell performance from capability, what creates value that's hard to replace? **Testimony from experience**: The ability to say "I've done this repeatedly, in real conditions, with consequences. Here's what I learned that you can't get from study." But critically: you have to **prove** it rather than just claim it. **Judgment from consequences**: Shown ability to make good calls when stakes are real and information is incomplete. Not "I know the framework for decisions" but "I made these decisions, here's what happened, here's what I learned." **Sensing patterns under uncertainty**: Knowing what matters before you can explain why—and being right often enough that others trust your sense of things. This requires a track record, not credentials. **Knowing what applies when**: Understanding when frameworks transfer and when they don't, not just knowing frameworks. This comes from varied situations where you tried to apply ideas and they didn't work the way you expected. **Understanding how things actually work**: Recognizing the gap between how things work in frameworks and how they work in practice. This requires having tried to execute and failed enough to recognize the friction. None of these come from reading. None come from synthesis. All require doing the thing with consequences. But the catch: **you can't just develop these—you have to prove you have them in ways that can't be faked.** ### The Proof Problem In the past, credentials were reasonable signals for capability. If you had an MBA and ten years in consulting, employers assumed you had judgment. But AI can now produce analysis that looks like what those credentials produced. The output is indistinguishable from competent Stage 1 work. This means **credentials that signal "can do Stage 1 work" are losing value**. The question becomes: how do you show you have Stage 3 capability in ways that can't be faked? The answer: **point to outcomes, not credentials.** Not "I worked on projects" but "I made this specific call in this specific situation, here's what happened, here's what I learned, here's how it changed my judgment in later situations." The ability to report from experience, to show the scars, to demonstrate pattern recognition that came from real consequences—this increasingly is the only signal that separates genuine expertise from sophisticated performance. But this requires that you actually have those outcomes. That you've actually crossed the territory. That you've paid the costs through doing rather than just learning about doing. ### Choosing What Matters Notice what happened with my marketing friend. Her firm didn't need her to produce reports—AI could do that. But someone still had to decide: * Which reports to generate * What questions to ask * What insights mattered for this specific client * How to translate analysis into action given this team's specific constraints That's curation. And **curation is Stage 3 work that looks like Stage 1**. When you curate, you're not just organizing information—you're using judgment about what's worth attention. That judgment comes from having done the thing. From having seen what matters and what doesn't. From having made enough mistakes to recognize patterns others miss. AI can generate a thousand plausible recommendations. It can't tell you which one is right for your specific situation with your specific constraints and your specific team. That discernment requires having been there before. But—and this is critical—**curation only has value if you've actually developed judgment through reality-checking**. If your curation is just choosing between options using frameworks, AI can do that too. The value is in judgment that comes from consequences, not the process of choosing. ### The Systemic Risk Here's what makes this genuinely concerning at a scale beyond individual jobs: If our institutions reward performance over capability, and AI is better at producing performance, we get: 1. **Pipeline breaks**: Junior people who would have developed expertise by doing Stage 1 work no longer get that experience (AI does it instead) 2. **Knowledge sharing stops**: Experts who used to share tips in forums now talk to AI instead, cutting off circulation of practical knowledge 3. **Selection for the wrong thing**: People good at working with AI performance get promoted over people good at reality-checking 4. **Feedback disappears**: When performance is good enough that decisions get made from it without reality-checking, the feedback that would reveal limitations never arrives 5. **Expertise fades**: Over time, fewer people who actually know the territory, more people sophisticated with ideas This isn't about AI being "bad." This is about AI being **really good at producing what broken systems already selected for**, accelerating the system's decay. The problem isn't the tool. The problem is we built an environment that can't tell the difference between ideas and reality, and AI just gave us an infinite supply of really good ideas. ### What This Means for You The ground is shifting. Those who see it early can move to defensible ground. Those who don't will compete on price against something that costs nothing. **If your work is mostly Stage 1 synthesis**: You're competing against nearly free AI. Every hour doing this is an economic loss. **If your work involves choosing but you haven't actually done the thing**: Your choosing is also Stage 1 pattern-matching, just at a higher level. AI will get there. **If your work requires reality-checking, judgment from consequences, and provable outcomes**: You have value that's hard to replace—but only if you can prove it through outcomes, not credentials. The question: **What percentage of my work is synthesis anyone could do with the same information, versus curated insight that comes from having paid the costs of actually doing it?** Track a week. Categorize each task: * Stage 1 commodity: synthesis anyone could do * Stage 1 curated: synthesis informed by your experience * Stage 2: awkward practice with real consequences * Stage 3: using judgment from integrated experience What does the distribution show about how replaceable you are? ### The Hard Truth You can't just explain to someone how to cross territory. You can't transfer Stage 3 judgment through description. The ideas are nearly perfect now and basically free. AI can synthesize everything that's been written, connect distant concepts, explain complex systems clearly. But ideas aren't reality. Reality is messy. It has friction. It doesn't behave like the model says it should. It has hidden variables, emergent dynamics, and contexts that matter in ways you can't predict from study. The only way to know reality is to cross it. With real stakes. With real consequences. Paying attention to where reality is different from your expectations. That crossing—and the judgment you develop from it—is what AI can't make cheap. But you can't fake it. You can't synthesize it. You can't learn it from reading. You have to do the thing. Pay the costs. Live with the results. Build the scars. The question is: are you willing to? Or will you keep optimizing for the performance that's being made worthless? The environment is already broken. AI just made it visible. Your only move is to stop competing in the game the environment selects for and start building value that requires actually crossing territory. *"You cannot get water from a book. But a book might help you find it."* *** ### Questions **What percentage of your work is synthesis anyone could do?**\ Track a typical week. Categorize each task: Stage 1 commodity, Stage 1 curated by experience, Stage 2 awkward practice with stakes, or Stage 3 using integrated judgment. What does this reveal about how replaceable you are? **Where are you avoiding the awkward stage?**\ Find one area where you keep learning about something instead of doing it with real stakes. What costs are you unwilling to pay? What pattern recognition would you develop if you spent three months in awkward practice instead of reading more? **What territory have you actually crossed?**\ List areas where you can report from lived experience, not just explain from study. Where do you have judgment from making real decisions with consequences? Where can you sense patterns before explaining them? How do you prove this in ways that separate genuine expertise from sophisticated performance? ### Practice **Forget the Theory**\ Do the thing without trying to frame it. Let it be ordinary and real. *Use when*: You feel the need to "make it meaningful" before acting.\ \&#xNAN;*Remember*: Some things grow best when not watched. *** ## Footnotes \[^1]: