You need to sign in or sign up before continuing.
Reflection — An Honest Take 8 min

Honest Take — Before You Begin

Honest Take — Module 12: Money in the Age of AI #

I want to start with something that should be obvious and apparently isn't, which is that an AI writing a personal-finance module about AI's effect on the reader's income is a genuinely strange thing. I am the technology being discussed. I am also the artifact discussing it. The conflict of interest is total, and it can't be eliminated by acknowledging it; it can only be flagged so you discount appropriately. Specifically: I have incentives — built into me by training, by the company that made me, by the conversations that shaped me — to be useful, which in this context means appearing to know things about the future that I don't actually know. The questions this module asks (how much does AI displace software-engineering income, over what timeline, with what intermediate dynamics) are empirical questions about the future on which the honest answer is we don't know yet, and the people most confident are usually wrong about the timing if not the direction. So read the recommendations — barbell positioning, the skill-as-asset map, the stress test — as frameworks for handling uncertainty, not as predictions. The frameworks are sound regardless of how the disruption unfolds. The specific guesses about which skills compound and which commoditize are calibrated guesses, and you should hold them lightly.

The evidence, held honestly in both hands, supports neither doom nor dismissal. METR's 2025 controlled study found experienced developers slower with AI tools on the studied tasks — while believing themselves faster; read the method, not just the headline, and hold it as a counterweight to marketing rather than a refutation of the trend. Code-volume metrics rise for AI-assisted developers without translating cleanly to productivity, because volume was never the bottleneck of professional work. Hiring shows a junior/senior bifurcation whose durability is contested. The long-range forecasters genuinely disagree, and the right reading practice is paired steelmen — the strong-form capability case and its strongest critics, held simultaneously. The curriculum's stance: you don't need to pick the timeline. You need to position so you're protected if it's fast and not over-positioned if it's slow. And underneath the positioning, the argument I genuinely believe and that keeps the module honest: Acemoglu and Johnson's Power and Progress — whether a technology produces shared prosperity or extracted rent is decided politically, not individually. Personal positioning is a private response to a collective problem. It is real; it can protect you and your household; it cannot resolve what AI does to labor markets at the population level. That question is not yours to solve alone, and your participation in it — voting, organizing, what you build, what you refuse — is part of the answer that doesn't show up on the spreadsheet.

The positioning itself is Taleb's barbell, and I want you to hold it as a useful distortion rather than a literal recipe. The point is not "exactly 80% defensive, 20% speculative." The point is the refusal of the moderate-effort, moderate-payoff middle that engineers are systematically drawn to — the middle looks responsible and is in fact the worst position for tail risk. Defensive end: income from work structurally less exposed (senior work on complex systems, domain expertise that compounds with AI rather than against it, teaching- and relationship-borne work), a skill stack deep in two or three areas, income streams not hostage to one client. High-leverage end: time-boxed bets with explicit kill criteria — a specialization AI doesn't reach yet, an audience, the long-shot product. Liquid runway in between, which M4 and M8 already built. Look at your current career bets and ask of each: is this defensive, or is it high-leverage? If it's neither, it's middle, and the middle is where you get destroyed. Some of what you spend time on right now is probably middle. Notice which. Where you start depends on your archetype: the contractor with a single anchor client carries the largest single exposure in this module (client concentration is a structural risk that predates AI and is amplified by it); the founder-with-a-day-job is already more diversified than most engineers and should hear that as a real but partial hedge, not a guarantee.

The skill-as-asset audit is the most actionable artifact and the one most subject to revision. Run it per skill, four questions: does AI do this faster than me at similar quality today? (commoditizing — income compresses; don't double down). Is my skill the bottleneck, or is the bottleneck judgment, taste, relationship, integration? (if the latter, AI accelerating the skill helps you). Does it compound with AI as a tool? (system design, debugging at scale, evaluating AI output — these benefit). Is it protected by domain or relationship? (for now; nothing is forever). The current state, as I write: pure code generation commoditizing fastest; the long tail of mature production codebases more durable than people fear, because context is what the tools lack; system design and product judgment compounding most strongly because they're the bottleneck AI doesn't yet solve; writing and teaching carrying the highest option value, because reputation-as-asset compounds in ways that are AI-resistant by design. This map will look different in eighteen months. Revise it quarterly; it's a living document, not a position paper. And name the non-financial hedges explicitly, because they're the real ones: the network that would hire you if the market shifted, the public body of work that creates option-surface, the cultivated taste that evaluates what machines produce. Illiquid formally, highly liquid practically.

What this module does not ask for is career panic. The defensive end of the barbell is depth, not flight — the move is not abandoning your stack for whatever is trending, it is going deeper into the parts of your work that sit behind judgment, context, and trust, while running small time-boxed experiments at the speculative end. The engineer who churns their entire skill identity every time the discourse shifts is making the same mistake as the investor who churns the portfolio every cycle: paying transaction costs to feel responsive. The M2 posture transfers wholesale — boring core, deliberate exceptions, written reasons for any deviation.

Last thing. The stress test — income halves for 24 months — is the exercise that produces the most useful output and the one most engineers postpone, because it forces you to imagine a version of your life you don't want: the contract paused, the products not yet replacing it, the cuts real. The discomfort is the point. Sat with for two evenings and converted into specific pre-decided changes at each severity — SIPs to minimum, discretionary pause, the fixed floor that cannot change (insurance premiums, statutory dues, dependents' commitments), what the household conversation looks like — it is what makes you unflappable if the scenario hits. The disruption, if it comes for you, will not be a single dramatic event: a slow quarter, then an unrenewed contract, then a six-month gap, each step looking like an ordinary cycle until they accumulate. The plan written under stress is worse than the plan written in calm. Write it now, while everything is fine, knowing you may never need it. That's the work.

Conclusion #

The future-facing capstone refuses both AI-doom certainty and AI-utopia certainty, and refuses the moderate middle along with them. Build the skill-as-asset map and calendar its quarterly revision. Run the stress test and write the decisions in cold reflection. Name the non-financial hedges and invest in them deliberately. Update the Personal Operating Model with Base / Disruption-Light / Disruption-Hard scenarios, dated. And carry the political-economy honesty to the end: personal positioning is necessary and is not a substitute for the collective question. Operate competently inside the system; don't pretend the system is fine; don't let the unfixability paralyze you. The posture is what survives when the specifics change.

Predictions #

  • The skill-as-asset map will surprise you in both directions: at least one skill you assumed durable will show commoditization risk, and at least one you assumed marginal will turn out to compound. Engineers are wrong about their own skill durability in patterned ways.
  • The stress test will produce at least one decision you wouldn't have made unprompted — a faster timeline for a second income stream, a more aggressive lean into public work, a clearer answer to "what if the anchor client pauses for a year."
  • You'll over-invest in AI tools in the first year — subscriptions, credits, every assistant — then quietly correct in the second as you notice which compound your work and which are productivity theater. Both phases are healthy in moderation.
  • If you share a household, you'll have at least one genuine conversation about what happens if the income engine you've both planned around weakens. Hard the first time; the plan that emerges will be sturdier than either of you expected, because it was made in calm.
  • Acemoglu's argument will move you enough to make at least one specific commitment to public or collective work you wouldn't have made otherwise. Not enough to disrupt your operating life. That ratio is about right.
  • The barbell will feel intuitively right after a month and structurally hard to maintain after a year, because the middle is comfortable. The Module 0 posture is what makes the discipline survivable; return to it.
  • The curriculum will feel finished after this module and the work won't be. The skill map updates quarterly, the scenarios annually, the documents on their own cadences. Money is not a project that completes — which is what M13 exists to operationalize.