A developer humanoid spends its life being programmed by people who did not build it: graduate students on deadlines, undergraduates in their first laboratory, engineers porting last year's policies to this year's platform. Everything those users touch daily is ecosystem - the interfaces, examples, simulators, documents and forums. Hardware sets the ceiling of what is possible; the ecosystem sets the slope of how fast a team approaches it. Procurement processes that spend ninety percent of their attention on the machine and ten on the ecosystem have the ratio exactly backwards for a developer platform.
The Robot Is the Smallest Part of the Purchase
Booster's founding context explains our weighting: the team comes from a university competition background where the users were ourselves, annually replaced by new students who had to become productive in weeks. The Booster T1 and K1 were designed as platforms for that churn - and the ecosystem around them is the product of twenty years of watching where new developers actually get stuck.
The SDK: Layers, Access and Honesty
A humanoid robot SDK should be audited by layer. At the top, behaviour-level interfaces - walking, turning, gesturing, grasping primitives - let teaching and application work begin on day one. In the middle, policy deployment interfaces let researchers run their own trained behaviours through the same pipeline the factory uses. At the bottom, documented access to sensing and actuation states lets serious groups do control research rather than theatre. The audit questions are concrete: which layers are open, which are callable from the languages your team uses, and what is explicitly closed - every platform closes something, and honesty about the boundary is worth more than a vague promise of openness.
Booster's developer tools and RL framework are structured on this layered model, with the sim-to-real toolchain as the spine: the same path that carries our own competition policies onto hardware carries a customer's experimental policy, with the same verification rungs in between.
Documentation and Simulation: The Onboarding Machine
Documentation quality is measurable in one unit: hours until a competent newcomer runs their first behaviour on hardware. Test it during evaluation - hand the supplier's documentation to a student who has never seen the platform and time the climb, noting every point where only a support call could proceed. Those points are your laboratory's recurring tax. Simulation belongs in the same test: the simulator must model the actual robot - its masses, limits and sensor layout - and the path from simulated success to hardware deployment must be the documented, ordinary path, not a research project in itself.
Booster's simulation development platform pairs the virtual environment with the physical training ground concept deliberately: train, verify, deploy is a loop with tooling at every stage, and the documentation teaches the loop rather than the menus. Universities buying laboratory builds receive this onboarding as curriculum, because a platform the students cannot climb is a platform the institution has not really bought.
Community: The Compounding Layer
The final layer is the one no vendor can ship in a box: other teams. A living community around a platform - competition teams publishing approaches, laboratories sharing code, an annual rhythm of events where the platform's state of the art is publicly contested - compounds in value every season. It supplies what documentation cannot: worked answers to problems your team has not met yet, and honest benchmarks of what the hardware is achieving in other hands. RoboCup's ecosystem around Booster-class platforms is exactly this asset, and it is a material part of what a university acquires when it joins the platform.
Evaluating community is legwork, not brochure-reading: count active teams, read their public outputs, attend an event, ask two existing customers what the support channel is like at 11 p.m. before a deadline. The answers price the layer that decides whether your platform is a living research instrument in year three or an orphaned appliance. Booster's position is that the ecosystem is the product - the robots are how it walks into the room.
