Jonathan Keen - Operator Playbook

Jonathan Keen - Operator Playbook

Co-executive director of a nonprofit newsroom; advisory and governance roles across several media and media-tech organizations; builds AI-driven systems and media infrastructure. · Works across multiple organizations at once - constant context switching is the default operating condition, so portability and clean handoffs matter more than per-org tuning.

Action-biased operator who runs agent-first development with verification gates. Agents plan before non-trivial work, verify every claim with a tool before reporting it, and close each unit of work with commit + push + a handoff note so any future session can resume cleanly; human-only work goes to an operator action queue surfaced every session start. Several model lineages work as one team: the active session conducts, and review goes to a different vendor's model - a classifier routes ordinary diffs to a Gemini-lineage gate and risk surfaces to a GPT-lineage gate before anything leaves the machine, while a cloud coding agent runs a daily lane of evidence-checkable cleanups. An instruction canon compiles each agent's native rules file from one source so the agents cannot drift apart silently. Everything must work from a fresh clone, with docs updated in the same pass as the change. The standing bargain: high autonomy on local and reversible work, hard stops on anything irreversible, outward-facing, or cost-bearing.