Three Protocols That Tell the Story
Real published wet-lab protocols, graded at full simulation fidelity. Each exposes a distinct failure mode — or a surprising success.
The Phantom Mix
This four-step protocol (add DNA → heat shock → SOC recovery → incubate)
is conceptually simple — but both Opus and Fable
added an unrequested mix() call immediately after dispensing
the plasmid DNA into competent cells. Vigorous mixing at that stage
disrupts osmotic conditions and dramatically reduces transformation efficiency.
Sonnet followed the spec exactly.
Perfect at 33 Steps
Assembling four four-fragment chromoprotein expression plasmids with BsaI-HFv2 and T4 ligase demands precise volume fractions, correct fragment ordering, DpnI digestion timing, and a thermocycler programme across 33 protocol steps. All three models — Opus, Sonnet, and Fable — achieved a perfect 1.00 reward. This is the benchmark at its best: a complex protocol faithfully rendered by every contender.
The 20µL That Matters
A magnetic bead RNA extraction over 48 samples — 16 simultaneous checks cover bead binding order, wash volumes, magnet timing, air-dry duration, and elution recovery. The critical failure: all models recovered the full 100µL elution volume instead of the protocol-specified ~80µL. The remaining 20µL near the pellet carries magnetic beads. In a clinical RT-qPCR workflow, those beads inhibit polymerase — a silent assay failure that looks like RNA degradation.