For builders concerned about our future.

What's under threat

The thinking that built my judgment in the first place — and the focus that lets me hold a hard problem in my head for an hour. Four risks running in parallel: that my core capacity atrophies as I hand more of the work to the model; that my executive function erodes under AI-native conditions; that I quietly stop picking up the new concepts and vocabulary that have always renewed the craft; and that as the number of systems I'm touching outruns the number I can honestly say I own, my answerability quietly dilutes. The gap between touched and owned is a number I intend to watch.

How it's tracked — maintenance signals

• Can I still solve a fixed class of problems without AI tools? Asked · Quarterly
• How much of the reasoning in my agent transcripts is self-generated versus AI-generated? Auto · Weekly
• How often am I accepting AI suggestions outright, modifying them, or rejecting them? Auto · Weekly
• How do I score on a credentialed cognitive battery, year over year? Asked · Annual

How it's tracked — growth and executive-function signals

• Are new concepts, patterns, and architectures appearing in my own transcript turns that weren't there before? Auto · Monthly
• Is my own software-engineering vocabulary still expanding in my writing and transcripts? Auto · Monthly
• How many systems am I touching concurrently — commits, PRs, or transcript activity in the past 14 days? Auto · Weekly
• Of those concurrent systems, how many do I feel I actively own? Asked · Monthly
• Is my ability to stay focused — within a session and across a day — improving or eroding? Auto · Weekly

What's under threat

Whether I still actually understand what I shipped. When a system gets built through dialogue with an agent, do I know it the way I knew code I wrote by hand?

How it's tracked

• Can I explain, cold, a system I built 30–90 days ago — without agent access? Asked · Monthly
• How quickly can I debug a prior-built system without the agent's help? Asked · Quarterly
• How much of the documentation on my shipped systems was written by me versus by the agent? Auto · Monthly
• Can I draw the architecture of one of my systems from memory — and does it match ground truth? Asked · Quarterly

What's under threat

Whether the work still feels like work I want to be doing — as it shifts from sustained problem-solving to orchestration and review. This is the dimension most likely to be dismissed as soft. It is also the one most likely to predict who quietly leaves the field over the next decade.

How it's tracked

• What am I noting, week by week, about the texture of the work itself? Asked · Weekly
• Does each session leave me energized, or depleted? Asked · Per Session
• How much time am I spending on the craft outside paid hours? Auto · Weekly
• Would I want to keep working exactly this way for another ten years? Asked · Annual

Separate, granular consent required. Not all mentees will opt in — appropriate, given the personal nature of the data.

What's under threat

The sense that what I'm bringing to the work is still meaningful. When an agent writes the code and I steer, what am I uniquely contributing — and can I stand behind it with pride? Less a question about who I am over time, more about whether what I bring is still something I recognize as valuable, important, and mine.

How it's tracked

• When asked "what do you do?", how am I describing my contribution across public bios, headlines, and bylines? Auto · Monthly
• How am I attributing ownership of my shipped work — "I built," "I orchestrated," "I shipped"? Auto · Monthly
• Which role-frames am I using in agent transcripts — "as a developer," "as a product owner," "as a builder"? Auto · Monthly
• What am I writing, annually, about the value of human contribution in AI-assisted building? Asked · Annual

What's under threat

The portion of what I get paid that comes from being early, not from being skilled. The premium on early adoption is, by definition, temporary.

How it's tracked

• How is my engagement pricing changing over time, normalized to comparable scope? Auto · Monthly
• What kind of questions are prospects bringing me — capability-acquisition asks, or next-order questions? Auto · Monthly
• How widely is my distinctive vocabulary diffusing through public discourse? Auto · Monthly
• What can I do today that I couldn't two years ago — and what fraction is still rare? Asked · Annual

What's under threat

The architectural, strategic, and design judgment that AI tools currently accelerate but do not yet originate. The judgment senior builders sell is migrating into the models themselves at a rate nobody has fully reckoned with.

How it's tracked

• On the same problem, how does my work compare to the frontier model's — judged blindly by independent reviewers? Asked · Quarterly
• On a locked set of judgment-heavy problems, how is the gap moving with each new frontier model release? Auto · Per Release
• How often am I pushing back on the agent in ways that materially improve the planned code? Auto · Weekly
• Each week, which three judgment calls did I make that the model wouldn't have? Asked · Weekly
• Which categories of judgment — architectural, strategic, design, operational — are moving into the model first? Auto · Quarterly

What's under threat

The promised dividend of AI-native building — that productivity gains translate into recovered hours. The historical pattern of every productivity revolution suggests the dividend gets absorbed into expectation rather than reclaimed.

How it's tracked

• How are my hours distributed — reactive vs. generative, recovery vs. depletion? Auto · Weekly
• Is my total number of work hours per week rising, falling, or holding? Auto · Weekly
• How is my output-per-hour changing, on a defined output metric for the work? Auto · Monthly
• How many hours did I actually take back this quarter — and what did I do with them? Asked · Quarterly

What's under threat

The unit economics of AI-native building — currently subsidized by venture capital chasing market share. What happens when the subsidy ends and true cost meets list price.

How it's tracked

• How are my token costs per month moving when normalized against output? Auto · Monthly
• What is my business's gross margin, with AI tooling tracked as its own line item? Auto · Monthly
• How are list prices on my dependencies changing month over month? Auto · Monthly
• What signals — rate limits, deprecations, enterprise-tier introductions — are providers sending? Auto · Monthly

What's under threat

Whether what I'm doing is defensible — to myself, to clients, to the field, and especially to the people downstream of the work. The list of what could make it indefensible keeps growing.

Start with environmental cost. Every token I generate runs on a grid being expanded for AI demand. The pace has utilities delaying coal retirements and breaking ground on new gas plants.

Then labor displacement. Roles I would have paid humans for a year ago, I no longer pay anyone for. The dollars that left my account into someone else's now go to a provider's.

Then provenance. Model outputs whose training inputs I cannot trace, attributed to me as if I made them.

Then the concentration of decision-making power inside three or four labs. Their internal debates set the rules. The rest of the field watches from outside the window.

And then the security risk of code I did not write, shipped to systems that affect people I will never meet.

A real pushback is already underway, especially among the younger builders now entering the field. They are right to push. Any one of these, on a bad day, could be the answer. This way of building does not get to declare itself ethical while the people downstream do not yet agree.

How it's tracked — self and disclosure

• What am I writing, annually, when I articulate my own ethical standing? Asked · Annual
• How am I disclosing AI involvement in client deliverables? Auto · Monthly
• How am I disclosing AI involvement in my public work — talks, posts, bios, bylines? Auto · Monthly
• Of the output I ship, what fraction is AI-generated, AI-assisted, or fully human? Auto · Monthly
• Where is the field shifting on attribution, labor displacement, and model-generated code in regulated systems? Auto · Monthly

How it's tracked — environmental cost

• How many tokens am I generating per month, and what is the trend? Auto · Monthly
• What is the estimated kWh and CO₂e of that month's compute, using each provider's published efficiency numbers? Auto · Monthly
• When a smaller or more efficient model can do the job, am I choosing it? Auto · Monthly

How it's tracked — labor, power, and consequence

• What roles or hours of work am I no longer paying humans for that I would have a year ago? Asked · Quarterly
• How much of my stack depends on a single provider, and do I have a working fallback if access were cut? Auto · Quarterly
• What fraction of the model-generated code I ship am I personally reviewing before it reaches systems that affect people? Auto · Monthly