Microduck opens pre-orders at $399. Its press kit distinguishes open software from hardware and lists the specifications that remain provisional.
Microduck is an interesting development platform to evaluate, but a pre-order needs a different reading from a finished-product review. The distinction matters particularly for a robot: software can be updated remotely, while an unsuitable sensor or a difficult repair stays on your desk.
What the manufacturer commits to
Pollen Robotics' press kit announces introductory pre-orders at $399 before tax and shipping, with first deliveries targeted before Christmas 2026. The software stack is described as open source; the mechanical and electronic design files are not. The kit also marks camera specifications, LiDAR range and SDK languages as provisional. Delivery is a target, not an observed shipment.
What to examine in the software
Before committing a classroom or research project, open the repository linked by the manufacturer. Check that the version you will use has a licence, installation instructions and an example that connects simulation to the physical robot. An open-source label alone does not establish that every dependency, model weight or training dataset has the same licence.
Separate three tasks in your evaluation plan: controlling the existing behaviours, adding a new behaviour, and reproducing training. They require different amounts of equipment and time. Being able to drive a robot from a game controller does not establish that you can reproduce its training on your own computer.
Record which software version and simulated environment produced a behaviour before transferring it to hardware. Keep the original policy available so a failed experiment can be reversed. That gives a concrete reproducibility criterion without pretending a press demonstration is an independent benchmark.
Budget for the experiment
The price of the robot is only one line in a project budget. Add delivery, taxes, batteries or spares you actually require, and any compute used for training. For a shared lab, also account for time spent charging, resetting and repairing equipment. Those costs can matter more than a small difference in purchase price.
Ask how the robot behaves when a connection drops, a motor stalls or a policy fails. Test progressively in a clear area appropriate to the manufacturer's instructions, with a way to stop motion immediately. A successful simulation is a useful gate before hardware testing, not proof that falls or collisions are harmless.
What would justify a recommendation
Useful independent evidence would include repeatable task completion, recovery after failure, documentation quality and the availability of repair parts. We have not measured those properties on a review unit. The launch image is supplied by Pollen Robotics and credited separately from this analysis.
Microduck's appeal is the proposed path from simulation to a small physical robot. Whether it suits a specific project depends on the delivered software, hardware and support. Treat unresolved specifications as questions to settle before ordering, rather than filling the gaps with assumptions.
September 7 revision: manufacturer sources reviewed, availability clarified and practical analysis rewritten.