A progressive path through simulation, control, data, policy training, and closed-loop evaluation for learners with basic Python experience
The 14 lessons currently comprise 8
cpu-verifiedand 6gpu-verifiedlessons, with 0plannedand 0published. Every lesson has reached its declared verification status; none is markedpublishedbefore the versioned artifacts and release acceptance in M5. L01–L07 arecpu-verifiedafter their clean-kernel CPU verification; their additional AMD paths also passed their respective contracts on the reference R9700. L08 isgpu-verifiedafter its bilingual CPU numerical paths, CPU+EGL path, and scripted-expert path on a reference R9700 all passed. L09 is alsogpu-verified: its bilingual CPU diagnostics, CPU+EGL path, and dual-camera recording and readback on a reference R9700 all passed. The renumbered L12 training content isgpu-verifiedafter ACT and SmolVLA GPU smoke, checkpoint audit, and reload checks. L10 iscpu-verifiedafter its bilingual CPU-only data-readback and inspection paths passed. L11 isgpu-verifiedafter its bilingual diagnostics, CPU+EGL path, and bounded-DR recording, readback, and provenance path on a reference R9700 passed. L13 isgpu-verifiedafter its bilingual diagnostics, single-checkpoint closed-loop evaluation on a reference R9700 with AMD+EGL, matched Reference/shifted rollouts, and H.264 video and report evidence audit passed. This does not establish a stable success rate, OOD robustness, or sim-to-real transfer. L14 is an optional advanced practicum outside the completion path for the 13 core lessons. Its required path isgpu-verifiedin the pinned companion runtime on four W7900D GPUs: base0/10, 100-step LoRA, DCP-to-HF merge, merged0/10, and a 20-video audit. These are independent small aggregate observations, not evidence of improvement; resolved-seed matching and the higher-cost competition extensions remain unverified.
