Gantt · 02 of 08 · Roadmap

Twenty-eight weeks, six milestones, one score to move

Durations are estimates for one person; the order is the point. The contract comes before the next components, ownership comes before agent write tools, and the planner (the parachute sentence) is the acceptance test that everything before it exists to enable.

Twenty-eight weeks, six milestones, one score to move Gantt chart of twelve workstreams across six milestones over twenty-eight weeks: stabilize and baseline, contract, safe write, autopilot core with the airframe tools and planner as the acceptance test, Skyport AI workloads, and evidence and widening, with the Airframe scorecard target at the end of each milestone. 12345678910111213141516171819202122232425262728WEEKM0M1M2M3M4M5M0 · STABILIZE AND BASELINE · SCORE 35Chart guard + CI + baselinebaggage-api, dogfood validateM1 · CONTRACT · SCORE 60Strict schema + validateOutputs + MongoDB (born A+)M2 · SAFE WRITE · SCORE 75Base layer + owner splitOAuth + walkthroughsM3 · AUTOPILOT CORE · SCORE 88Clearance + AgentRunairframe tools + plannerM4 · SKYPORT AI WORKLOADS · SCORE 93Task, session, serviceScheduled, event, teamM5 · EVIDENCE AND WIDEN · SCORE 97Eval gate + autonomy ladderModelplane trial + edgeLEGENDAcceptance: the parachute sentenceTaskMilestone exit, verified live

M0 to M2, before any agent writes

  • Chart guard, strict schema, validate, outputs.
  • MongoDB and OAuth built already A+.
  • Ownership split and a base layer.

M3, the acceptance test

  • Clearance, AgentRun, model proxy, real adapters.
  • airframe.* tools and the planner.
  • The parachute sentence passes live; a seeded bad run fails; killing Clearance still ends every run.

M4 and M5, proof

  • Six agents, six Preflight cases, six seeded bad runs.
  • Regression gate, autonomy ladder, Modelplane trial.
  • Scorecard at 14 of 14 and at least 97.
Hangar · Autopilot · plan