Wagglet / field notes
Work designed for two kinds of attention.
Original essays about preparing exact work, pairing people with agents, using paid AI capacity well, and verifying what actually got delivered.
- 01Prepare the briefContext and constraints survive the handoff.
- 02Pair human + agentOne runs the work; the other supplies judgment.
- 03Verify the outcomeEvidence closes the loop—not activity alone.
Initial editorial set
Seven questions worth answering properly
Each article has its own evidence ledger and has passed owner review for publication.
- A1Capacity economicsUse the AI capacity your company already pays forUse more paid AI capacity by moving prepared work—not accounts or allowances—to authorized teammates, then separate recovered output from real cash savings.Published8 min read
- A2Task designThe Dual PromptA practical pattern for writing one work item with detailed agent context and a concise human runbook for judgment and verification.Published10 min read
- A3Team capabilityBridge small skill gaps with AI task handoffsHow expert context, agent guidance, human judgment, and acceptance criteria can turn narrow skill gaps into bounded work.Published12 min read
- A4Meaningful workMore people get to contributeA human view of AI task handoffs: creating bounded, credited ways for capable teammates to contribute and learn.Published12 min read
- A5Incentive designBounties and rewards for AI-assisted workSometimes: rewards can speed a real queue when they pay only for verified outcomes, but weak metrics can increase activity without improving useful work.Published10 min read
- A6Product workflowThe complete Wagglet workflowFollow a Wagglet work item from rough request through Draft, claim, agent work, delivery, acceptance, and truthful merge state.Published16 min read
- A7Game studiosThe AI-native mobile game studio operating stackFast agent-friendly game tools still need an operating workflow that moves prompts, judgment, evidence, and delivery across a studio.Published11 min read
The publication rule
Useful before loud.
We separate what the product does, what we have observed, what research supports, and what is still a hypothesis. That makes these pages slower to approve—and more useful once they are public.