https://rlsok.com/log/We are building an AI-designed bird in public.enhttps://rlsok.com/log/#2026-10-26rlsok-log-2026-10-26-enMon, 26 Oct 2026 12:00:00 +0000We found the E-Flap flight logs published by the GRVC team in Seville and used them to check our two-wing aerodynamics model: flap frequency predicted within 8 %. R2 was re-run with the new model and completed all 24 simulated flights. The SmartBird reference point is still 1.6 points outside the pre-registered limit, so the model is formally marked as not passed; the attitude model could not be validated predictively because the logs are closed-loop flights. Next: a tethered test. The real bird flaps on a stand while we measure thrust, current and pitch response.https://rlsok.com/log/#2026-10-23rlsok-log-2026-10-23-enFri, 23 Oct 2026 12:00:00 +0000We first checked the UVLM high-fidelity solver itself against classical theory, then found that the previous check had fed it a wing-tip twist three times too large, fixed it and re-ran. On Eye-Bird ornithopter slow-flight data, four aerodynamics models all under-predicted lift by about half. Find real flight data that includes the control inputs.https://rlsok.com/log/#2026-10-22rlsok-log-2026-10-22-enThu, 22 Oct 2026 12:00:00 +0000The shape now follows gull and kite proportions: five separated primary feathers at the wing tip and a fan tail of seven feathers. It completed every simulated flight. There is almost no public measured data for two-wing flappers like this to check the aerodynamics against. Find public measurements and cross-check with a high-fidelity simulation.https://rlsok.com/log/#2026-10-21rlsok-log-2026-10-21-enWed, 21 Oct 2026 12:00:00 +0000The website and the Windows installer went live for the first time; the AI started designing the first bionic bird, R1, which then completed 16 simulated flights. The bird's attitude model failed validation three times in a row. Make the shape look more like a real bird and keep looking for data to validate the attitude.