Case Study: Raj's Capstone Plan — The AI-Native Developer
Raj's Starting Point
By the time Raj completes his Personal AI Mastery Plan, he's been one of the more advanced AI practitioners he knows for almost two years. He developed the capability testing battery that other developers on his team now use. He built and maintained the team AI coding standards that transformed how code review works. He's been thoughtful and deliberate about what he's done with AI and what he's kept doing without it.
His capstone challenge is different from Alex's. He doesn't need to restart a stagnant practice. He needs to decide what comes next for someone who's already in the top tier — and to think honestly about whether his current ceiling is the right ceiling.
His Six-Dimension Assessment
Mental Models & Trust Calibration: 5
Raj's trust calibration is precise and regularly updated. He has a documented map — high trust, medium trust, low trust, minimal trust — across all the task types his team works on. He reviews and updates it quarterly. He updates it when evidence contradicts it.
Prompting: 4
Raj is an excellent prompter for coding tasks. His system prompts and context-setting are among the best he's seen. Where he gives himself a 4 rather than a 5: he's less skilled at prompting for non-technical tasks. His prompts for written communication and strategic thinking are good but not expert-level.
Platform Knowledge: 4
He has deep knowledge of the AI coding tools in his stack and solid knowledge of the underlying model families. He's less familiar with non-coding AI tools, which matters because some of what he needs (documentation quality review, technical blog posts, stakeholder communications) involves non-coding AI use.
Workflow Integration: 4
AI coding tools are deeply integrated into his development workflow. The automated quality pipeline is running and generating value. He's built custom configurations that extend standard capabilities.
Critical Thinking & Ethics: 5
His verification habits are among the best documented in this book. He's thought through the ethics of AI-assisted code thoroughly: the explainability requirement, the security review requirements, the testing standards. He's also thought through the professional development implications — the deliberate no-AI practice for junior developers.
Advanced Skills: 4
He has sophisticated automated workflows, API-level integration knowledge, and a measurement practice that generates real team-level data. The gap: he's never formally presented what he's built. The knowledge lives in his head and his team's documentation, but he hasn't contributed to the broader community of practice.
Total: 26/30 — Advanced
His Primary Growth Path
Builder primary, moving toward Expert.
Raj is already deeply competent as a Builder. What he's less developed at: the Expert dimension of breadth and contribution. His knowledge is deep and narrow — excellent for his team, less useful for the broader community of practitioners who could learn from what he's built.
His one-year goal is to begin contributing to the broader practitioner community — not abandoning depth, but adding breadth and contribution.
The 30-Day Sprint
Goal: Complete the documentation of his capability testing battery as a publishable resource — not just team documentation but something he could post publicly or present at a developer conference.
Writing for a public audience forces a different level of clarity and completeness than internal documentation. The battery is genuinely useful; making it available beyond his team is the right next step.
The habit: One deliberate "no-AI" session per week on a hard problem. Not because AI wouldn't help — because keeping the debugging instinct sharp is worth the time cost.
Quick win: Draft the first section of the battery documentation this week. Get feedback from one team member.
The 90-Day Plan
Primary skill investment: Non-technical prompting.
Raj has been strong on technical AI use and weak on non-technical use. In 90 days, he'll develop a working prompt library for his top five non-technical tasks: technical blog posts, stakeholder communications, performance review feedback, onboarding documentation, and technical job descriptions.
Why this matters: about 25% of his work week is non-technical writing. Improving his AI use on these tasks would recover significant time with no quality cost.
New capability: Agentic coding workflows.
The automated quality pipeline Raj built is a form of agentic AI — AI that takes actions in a workflow without constant human supervision. In 90 days, he'll extend the pipeline to include a security pre-scan that flags code in high-risk categories before human review. This is the next step in his Builder path.
Learning investment: Submit a talk proposal to one developer conference on "AI Quality Standards for Development Teams." Preparing the talk will force him to articulate what he's built clearly and expose it to the broader community's critique.
The One-Year Vision
In one year, Raj will have moved from "advanced practitioner" to "community contributor." The shift:
His capability testing battery will be publicly documented and available as an open resource. At least one other team outside his organization will be using it. The conference talk on AI quality standards will have been delivered.
His automated quality infrastructure will be mature — not just the PR quality pipeline but a second pipeline for documentation quality and a third for security scan integration. The team's post-merge defect rate will be 25% below his pre-AI baseline.
The junior developer AI literacy program will have run two full cohorts. He'll have data on whether the no-AI practice built into the program is actually improving debugging skills (he believes it is but hasn't measured it systematically yet).
He'll have begun contributing to an open-source AI tooling project in a meaningful way — not just using it, but contributing.
One-year metrics: - Capability battery: publicly documented and in use beyond his team - Post-merge defect rate: 25% below pre-AI baseline - Junior developer program: two full cohorts completed, outcome data collected - Conference talk: delivered - Open-source contribution: at least one merged pull request in a relevant project
What Raj Will Do Differently Next Week
Start: Drafting the capability battery documentation for public sharing. This week: the introduction and the first test case description.
Stop: Running his battery evaluations as an internal-only exercise. He'll share his next evaluation results — and his methodology — on LinkedIn, to get feedback from developers he respects.
Improve: His security pre-scan prompt. He's been using the same prompt template for six months. He'll run the current version against his five most complex recent security failures and see whether it would have caught them. If not, improve.
His Closing Reflection
"When AI coding tools first got good, I was worried about what it would mean for developers. Not worried about my job — worried about the craft. Worried that the struggle of debugging, the satisfaction of figuring out why something was wrong, would get automated away.
Two years in, I think I was half right and half wrong. The struggle doesn't go away — it moves to a different level. You're not debugging a loop; you're debugging why AI's approach to a class of problems is subtly wrong. You're not figuring out a data structure; you're figuring out why AI's suggested architecture doesn't account for your system's specific constraints. The problems are harder and more interesting than the problems AI can solve.
What I didn't expect was that AI would make me more aware of what I actually know. Catching AI's mistakes requires understanding what it got wrong. And that understanding, built over two years of using AI critically, is sharper than anything I learned from the manual work. I know my domain better because I've had to explain it to AI — and explain why AI's answers are right or wrong.
That's not the outcome I expected. But it's the one I got. And I'll take it."
Raj's conference talk — "AI Quality Standards for Development Teams: What Two Years of Running the Battery Has Taught Me" — is accepted. He presents in eight months.