The Human Warrant Series
Too Good, Too Soon
AI erased the routine work that trained experts. A plan to rebuild the ladder — exposure, practice, proof, responsibility.
About the book
The ladder nobody designed is disappearing.
Across professions, experts were never really made by training programs. They were made by accident — by the ladder nobody designed. A new engineer started with routine cases nobody else wanted, made small mistakes on low-stakes work, got corrected, and moved on to harder ones. Decades later, that person was the one everyone brought the genuinely difficult problems to, not because anyone certified them, but because they had accumulated enough encounters with enough variation that their judgment was worth more than any rule someone could write down.
That ladder is now missing its bottom rungs. Capable systems can produce, in minutes, the routine analysis, the first drafts, and the low-stakes judgment calls that used to be junior people’s work — and every individual decision to hire fewer junior people looks completely reasonable in isolation. But a decade of reasonable-looking cuts compounds: the cohort that should be arriving at senior readiness has been shrinking the whole time, and a population that was never hired shows up on no org chart. The work disappeared. The developmental need it used to satisfy didn’t.
Too Good, Too Soon names the crisis — a missing middle of professional judgment, arriving in roughly five to ten years — and builds the deliberate replacement. Dr. Daryoush Tehranchi lays out a four-stage cycle (exposure, practice, proof, responsibility) that organizations can actually run, the honest economics of keeping deliberate inefficiency alive through budget pressure, and the harder case for why no single company can solve this alone. The appendix gives managers something concrete for Monday: an eight-question Expertise Pipeline Audit for one entry-level role, built to be run per team and answered plainly. Written across four decades in defense and aerospace engineering, this is the book for anyone who has to answer the question the industry keeps postponing: where does the next generation of real experts come from?
Key takeaways
Six ideas worth keeping.
Takeaway 01
Performance is what you can see; capability is what actually produced it.
For nearly all of human history these were close enough that nobody needed to separate them — excellent work reliably meant an excellent worker. Now a system can produce the performance without ever becoming capable of anything, and the old inference quietly stops holding. Every hiring, promotion, and trust decision that assumed it has to be rebuilt around that distinction. (Chapter 2)
Takeaway 02
The ladder was the work itself, and nobody designed it.
Experts were built because routine work genuinely had to be done by somebody, and the somebody available happened to be new. Each rung was a real task, scaled to what a person could survive getting wrong, repeated at volume until the library of specific instances behind real judgment was full. Remove the volume and the ladder disappears, not the need for what it produced. (Chapter 1)
Takeaway 03
The missing middle is arithmetic, and nobody is tracking it.
Cut junior hiring ten percent a year and after five years you’re hiring twelve instead of twenty; extend it over an eight-year development timeline and the cohort that should reach full senior readiness has been shrinking for its entire development. Senior headcount looks intact right up until the year the pipeline’s output is actually needed — precisely the moment it’s too late to fix quickly. (Chapter 6)
Takeaway 04
The inefficiency was the mechanism, not a flaw in it.
Having a less experienced person work a real case at real stakes — slower and costlier than giving it to the best available person — was precisely what built the judgment the organization would need later from that same person. The only serious response is to deliberately reintroduce that friction: protect real independent attempts on real cases, and treat the inefficiency as an investment rather than a cost to optimize away. (Chapter 7)
Takeaway 05
The replacement cycle has four stages, in order: exposure, practice, proof, responsibility.
Attempt real work before a capable system supplies the answer; receive timely, specific correction from someone with more experience; demonstrate the resulting judgment on an unfamiliar problem without the system’s help; then receive real authority and accountability, sized to what the proof actually verified. Skip any stage and the effort fails in a specific way. (Chapters 7–11)
Takeaway 06
The step most organizations skip is the last one — and there’s an audit to find your gaps.
After doing everything right — real exposure at real cost, a genuine correction loop, an honest test with an honest answer — organizations still decline to act on it, withholding responsibility from someone who demonstrably earned it. The appendix’s eight-question Expertise Pipeline Audit converts the whole argument into something actionable: run it per role or team. (Chapter 10; Appendix)
Who it’s for
For anyone responsible for growing the next experts.
Questions
Before you ask.
Is this book anti-AI or arguing against efficiency?
No. The productivity gain from automating routine work is described as real and immediate — but quietly borrowed against a capability that was never built to pay it back. The book argues for reinvesting a deliberate slice of that efficiency into development, not against using the tools.
Is it just a prediction, or does it give me something to do?
It gives you two concrete tools: the four-stage cycle (exposure, practice, proof, responsibility) with diagnostic questions for checking whether your organization is actually running it, and the eight-question Expertise Pipeline Audit in the appendix — built to be run per entry-level role and repeated periodically.
Is the “missing middle” actually happening, or is this speculation?
The book is explicit: the contracting floor — junior hiring shrinking for locally sensible reasons — is an observable fact about today’s labor market. The missing middle a decade out is a risk reasoned forward from the mechanism, not yet an observed outcome in its own right.
Has the four-stage framework been scientifically validated?
No, and the book says so directly: it’s a reasoned reconstruction of the mechanism the argument lays out, not a model validated through a longitudinal controlled study. The individual pieces rest on real evidence; the cycle is offered as a disciplined starting point to adapt.
I’m an individual contributor with no power over my company’s training. Is this book for me?
Yes — Chapter 16 was written for exactly you. Exposure, practice, and proof can all be run on your own initiative. The one stage you genuinely can’t do alone is responsibility — but arriving with the other three built is the strongest possible position to make the case.
The author
Dr. Daryoush Tehranchi is an engineer, architect, educator, and author whose work focuses on complex systems, systems and software architecture, digital engineering, artificial intelligence, technical leadership, and the decisions that remain fundamentally human.
The Human Warrant Series
More from the series.
Connect with Daryoush