Vik Olliver's open-source RepRapMicron project reports roughly one-micrometre positioning repeatability from PIKA V0.02, a machine built around 3D-printed parts and familiar CNC hardware. The figure needs context: the project's stated accuracy is 10µm or better across a working volume roughly 8mm on each side. These are project-reported figures. MrKeyoor has not built or measured the machine.
The project appeared on Hacker News's October 18 front page. Its XY-stage assembly instructions were updated on October 17, 2026. The design itself is older: the project wiki dates a completed PIKA V0.02 to around May 2026. For prospective builders, the separate specifications help define what to check on a machine of their own.
Repeatability doesn't settle part accuracy
Returning consistently to a point and reaching the intended position are different requirements. A mechanism can repeat a small positioning error. Reading the project's two specifications together, the roughly 1µm repeatability figure alone cannot establish 1µm accuracy in a finished part.
The repository also reports a functional resin-deposition system for droplets smaller than 4µm. That figure concerns deposited material. If you're evaluating a fabrication job, finished-part accuracy still needs its own measurement.
Ordinary printing makes the mechanism
PIKA's construction guide describes parts printed on a Prusa Mk4 with a 0.4mm nozzle, taking about 25 hours and using less than 500g of filament. It specifies plain PLA or eSun PLA+. Those are the project's build details. They don't establish the cost of a complete machine.
The XY-stage design prints the X-axis flexures and frame as one piece, with the upper stage levelled by three adjustable screws. Assembly still affects the mechanism: the instructions tell builders to tighten the X linkage evenly to avoid lateral strain. Substantial stage-height adjustments require loosening and retightening the Y linkage screws to avoid stressing its flexures. Fine motion depends on what happens after the parts leave the printer, too.
Software contributors have a route in through the repository. It includes OpenSCAD models and Python code that converts slicer G-code into individual probe-deposition points, with provision for recharging the probe after a set number of points. The README describes an alpha-stage project intended for developers and experimenters, and warns that downloadable Printables files aren't automatically synchronised with the source or instructions.
Before attempting tiny parts, follow the build guide's calibration procedure: it recommends a microscope calibration slide for a more accurate determination of axis movement and says output measurements require a bench microscope. The useful next evidence would be results from other builds that report positioning repeatability and finished-part dimensions separately. That would help establish how well these documented figures survive different printers, assembly work and calibration.