Agentic Coding
Coding agents changed the job. One engineer can now run a small team of tireless juniors — if they know how to orchestrate, specify, contextualize, and review. This board is Hireboard's playbook for doing 6x the work without shipping 6x the risk.
The Agentic Coding Doctrine
Five principles behind every module on this board.
1. You are the orchestrator now
Agents make generation cheap; your prompts and your reviews are the scarce resources. Engineering skill is shifting from typing code to specifying, dispatching, and judging it.
2. Rotate, don't wait
Short runs (1-5 minutes), tiny reviews, and N agents stacked by pace — long POC lanes in the background, tight important work in the foreground. The Rotating Door drives idle time to zero.
3. Chunk size follows the quality bar
One-shot the disposable POC after a diagrammed plan; slice production work vertically — one function through every layer per prompt.
4. Context is infrastructure
A sharp CLAUDE.md and agent-written session notes make every future session smarter. Recurring corrections are missing context, not model failures.
5. Never merge unread code
You own every line you merge. Review the plan before the code, hunt AI-specific smells, verify by running — and let the research keep your confidence calibrated.
Training Modules
Parallel Agents & the Rotating Door
Short runs, tiny reviews, N as a function of run length, shift-stacking lanes by importance — and the isolation that prevents conflicts.
The Chunking Policy
What to one-shot and what to slice: POCs get a plan and one shot with Fable; production gets vertical slices — MVC for one function per prompt.
CLAUDE.md & Session Notes
Context that compounds: standing orders the agent loads every session, and agent-written notes that carry decisions across sessions and people.
Reviewing AI Code
The discipline that makes 6x safe: plan-first review, the AI-specific smell catalog, verify-don't-trust, and depth scaled to consequence.
Research Corner
What the studies actually show — speedups, slowdowns, the perception gap, code-quality trendlines, and how to read benchmarks sanely.
Practice Questions
Comprehensive scenarios from real agentic-coding days — rotation jams, failed one-shots, suspicious diffs, and research claims.
How This Board Connects
Agentic Coding is the daily practice; its siblings supply the judgment. The prompts you write are specifications — the CTO board's planning method (research, diagrams, buy-in) scales down to a single Lucidchart diagram before a one-shot. The code you review is judged with Technical Foundations' design principles. And the systems you build with agents follow Agentic Theory's reliability doctrine.
Work through it in order: parallelism gives you throughput, chunking and context make the throughput good, review makes it safe, and the research keeps you honest about all of it.
Start with the Rotating Door
One technique, six agents, and your afternoons stop being spent watching a single terminal think.
Begin