The Work That Built My Judgment Is the First Work We Automate.

In December 1996 I started my first full-time software developer job, at FlexiInternational. I was 21. I had done tech support and some development consulting, and I had finished the computer science classes for an associate's degree at a technical college, then dropped out when Flexi offered me the position. I moved out of my dad's house and bought a car the same month, on a salary of $30,000 a year, which felt like a fortune.

Every egg I had was in that basket. I was not afraid of any of it. At 21 there was no fear of anything, although there should have been.

I had very little database background. A senior database engineer took me under his wing; his name has long since passed from myth into legend. He pushed me to tune my own queries, and he let me struggle through them when he could have rewritten them in minutes. A few days of that struggle shaped my whole career. I came out thinking in rowsets instead of single rows, and SQL is still one of my strongest skills.

Flexi was not running a charity. On that $30,000 salary I delivered systems worth more than $100,000, with minimal support from anyone else. The company developed my career, and it was better off because I was there.

At a dinner in Manhattan last month, the conversation among a table of technology leaders turned to what happens to entry-level engineers now. I said that coding has been getting cheaper my whole career, and judgment is the piece that cannot yet be replaced. The grunt work we used to hand juniors was how they built that judgment. I left with a question I still cannot answer cleanly: if the work that built judgment is the first work we automate, where does the next generation of senior engineers come from?

The question is personal for me. I grew up with a single mom, barely above the poverty line, in a house with a strong focus on education. I spent weekends with my dad, who had started another family, and he did the best he could too. Those circumstances shaped who I am. My sons do not face that kind of insecurity, so I have had to look for other ways to shape them. The industry is in the same position now.

Leaders answer the entry-level question in one of two ways. One camp cuts junior hiring, since agents now do the grunt work. The other hands juniors a copilot and calls them productive from day one, because the tool fills the gap experience used to fill. Either way, senior engineers' time is finally freed from teaching. Almost every senior engineer can name the person who let them struggle on purpose, and very few can name the junior they have done that for this year.

Grunt work did two jobs at once. It was the business case, because a junior cost less than the value they produced. It was also the curriculum, because the struggle through that work built the judgment that turned a junior into the company's next senior.

Automate the grunt work first and both jobs go. The business case goes, because the agents now do the work that made juniors profitable. The curriculum goes even for the juniors still hired, because the tool rewrites the query in seconds, before anyone has struggled long enough to see the pattern underneath it. Nobody decides to remove the rung. It disappears because the faster path is always there.

The cost arrives late. Agents bring no judgment of their own, so the scarce resource becomes people who can judge what the agents produce, and that bench is built out of today's juniors. In five to ten years, organizations that removed the rung will either buy that judgment from outside at senior prices or promote people who never had the struggle that builds it.

The instinct that removes the rung lives in the senior engineer, not the junior. I named it the High-Performer Trap: every intervention that accelerates an outcome in the moment lengthens the team's path to independence. Rewriting the junior's query in minutes is that trap, and an agent now offers everyone the same shortcut at no cost.

What has worked with my sons is structure: religious education, scouting, and modeling the behavior I expect from them, inside a strong family. Scouting solves this problem by design. A Scout does not earn a rank because circumstances made life hard. They earn it by meeting defined requirements and demonstrating each skill to an adult who signs off only on the demonstration, and older Scouts teach younger ones.

Engineering teams can build the same thing. Define the rungs: the specific judgments a junior has to demonstrate, such as finding the row-by-row loop in an agent's query and rewriting it as a set operation before anything ships. Make sign-off depend on the demonstration, have seniors withhold the answer the way mine did, and have last year's juniors teach this year's. The business case returns with it: juniors used to pay for themselves by producing cheaply, and now they pay for themselves by verifying, which is the bottleneck every AI-heavy team already has.

Name the most junior engineer on your team. Write down the last three things they struggled through to the end without someone, or something, finishing it for them, and next to each, the judgment it built. If you cannot fill three rows, or the rows are tasks an agent finished, the rung is gone. Then turn the question on yourself: when did you last withhold an answer on purpose, the way someone once did for you?

Quiet Confidence goes deeper into the instinct that removes the rung: why the leader who steps in fastest is the one whose team learns slowest, and how to choose restraint when the fix would take you minutes: http://TheQuietConfidenceBook.com