Xaira Therapeutics says a robot arm it built in-house from low-cost, off-the-shelf parts kept a laboratory instrument working almost non-stop for seven days, analysing 18 plates of protein samples. The AI drug-discovery company says that is 80% more than a modelled weekday 9-to-5 loading schedule would have managed.
The figures come from a technical post Xaira published on 23 October introducing X-Automate, its laboratory automation system. All of the numbers are the company’s own and have not been independently checked. The system pairs a vision-guided arm, X-Bot, with X-Scientist, the agentic AI that Xaira described in a separate write-up two days earlier.
In brief
- What happened: Xaira says X-Automate loaded and ran 18 96-well plates on an HPLC instrument over seven days, keeping it in use 97.2% of the time.
- Why it matters: it targets the hand-offs between machines that still need a person to open a door, load a plate or start a run.
- What’s unknown: how the system holds up over months, and on instruments other than this one.
What X-Automate is
Xaira says X-Automate has three parts: automation workcells that scale up lab work, X-Bot, and X-Scientist. X-Bot is a six-degree-of-freedom robotic arm. The company says it paired an inexpensive commercial arm with open-source tools, wrote the control and integration software itself, machined metal gripper fingers in its own workshop and 3D-printed other parts.
X-Scientist uses a vision-language model, an AI that reads images as well as text, to interpret the scene, choose operations and coordinate the workflow. In its 21 October post, Xaira presented X-Scientist as an AI “co-scientist” for drug discovery and said its next phase was to link scientific reasoning with automated experiments. TSN covered that announcement when it was released.
The problem it targets
Xaira says its industrial workcells can already express and purify proteins at a rate of thousands of samples a week. The slowdown comes afterwards. Analytical instruments have manual doors, drawers and software controls, so a plate that is ready after a few hours may wait until the morning, or through a weekend, for a person to load it.
The first target was an Agilent HPLC (high-performance liquid chromatography) machine, which Xaira calls the first bottleneck in its protein pipeline. According to the post, the instrument has no motorised sample door, little room for error in placing a plate, and a drawer that can shift at the slightest touch. Access to its automation interface also needs an expensive licence. X-Bot’s answer is a camera on the arm that finds each plate and judges its angle, with continuous visual feedback so the arm can correct its approach if the plate is rotated or the drawer moves.
The seven-day run
In what Xaira calls its first field deployment, the system linked the protein-production workcell to the HPLC. Plates came out of refrigerated storage, went through centrifugation and seal removal, and were carried to the instrument by a mobile robot called Wall-E. X-Bot then opened the drawer, swapped the plates and closed it again.
Each plate took 552 minutes from workcell to finished analysis, Xaira says: 537 minutes on the HPLC and 15 minutes for automated preparation, transport and loading. With the analytical method unchanged, the instrument was in use 97.2% of the time, against 53.6% in Xaira’s model of a human 9-to-5 working week.
The run was not trouble-free. Four interruptions in the upstream plate storage needed manual recovery. Xaira says staff restored operation without significant delays further down the line and that no errors were seen in X-Bot or X-Scientist. Separately, the company says it ran more than 100 pick-and-place trials with no handling failures, and calls those results encouraging but says more testing is needed to establish long-term reliability.
Safety stops that bypass the AI
Because X-Bot shares space with scientists, Xaira has added safeguards on top of the arm’s built-in collision detection. An overhead camera watches the area around the arm. An intrusion pauses the work, which can resume once the area has been clear for a set time, and losing the camera feed also triggers a stop. The safety system is hardwired to the arm’s control box. Motion corrections and protective stops run on local computers without waiting for the AI model, which Xaira says matters for speed.
The company says these measures are designed in line with guidance from OSHA, the US workplace safety regulator. It also says each installation still needs its own risk assessment, including whether automatic restarts are appropriate.
A narrow test with a clear point
The strongest part of the claim is also its most modest. Xaira did not change how the HPLC analyses samples. It simply stopped the machine from sitting idle overnight and at weekends, which is where the gap between 97.2% and 53.6% comes from. For any lab whose expensive instruments spend half their week waiting for staff, that is a practical problem, and a low-cost arm that can work with machines built for people is a cheaper idea than replacing the instruments.
The evidence is thin in places, though. The 80% figure compares a real week against a modelled schedule, not against measured human performance. The run covered one instrument, one assay and 18 plates, and the four storage failures show that one reliable robot is not enough. Xaira itself says every connected system has to be reliable before a lab can run itself. All of the figures come from the company’s own post, with no preprint, code release or outside audit to check them against.
From one arm to a fleet
Xaira describes its broader goal as a self-driving lab with a fleet of about 100 instrument-level arms, linked by mobile robots and coordinated by X-Scientist. It is also developing Eve, a home-built mobile platform carrying an X-Bot. For now, that fleet and Eve are a stated vision rather than a deployed system.
Other groups are chasing the same goal. Google Research has put AI agents on wet-lab robots, Warp Sciences says it has raised $215m for AI and robot chemistry labs, and one AI-guided lab says it has found new hydrogen catalysts.
Xaira’s next candidates are plate-based flow cytometry, nanoDSF stability tests and ELISA binding assays. Each one, the company says, could become the next bottleneck as protein production grows.
TSN tracks AI used for science and health on its Tech for Good hub.
Sources
- Xaira Therapeutics, “X-Automate brings physical and agentic AI to the lab”, 23 October 2026
- Xaira Therapeutics, “X-Scientist: A Recursive Self-Improving Agentic AI Platform for Drug Discovery”, 21 October 2026
Related stories
Follow @tsnmediahub on X for breaking AI news as it happens.
Follow on X
