Research: benchmark analysis and expansion candidates

[Tyronita]

Research branch for growing Text2WetLab beyond its 11 tasks. Nothing here changes the benchmark itself; everything is under research/expansion/.

• Candidate (paper, code, results) triples from four searches, merged in merged_candidates.jsonl and summarised in SUMMARY.md
• Simulator verification of the 25 ready liquid-handling candidates (VERIFIED.md)
• A deterministic fault detector and 3D replay tools (tools/, fault_report.md)
• Per-task breakdown and complexity ratings (ANALYSIS.md, also posted below as a comment)

🤖 Generated with Claude Code

[Tyronita]

Text2WetLab: benchmark analysis and expansion plan

How tasks are classified

Every task gets three discrete labels instead of a numeric score.

• Structure is what the robot has to coordinate, read from the protocol:
  S1 single transfer (one source to some wells) · S2 plate set-up (several reagents added to the same wells) ·
  S3 wash cycles (liquid added then removed again) · S4 module-coordinated (pipetting interleaved with magnet,
  temperature or thermocycler steps and timed waits) · S5 multi-reaction workflow (one reaction's product is the next reaction's input).
• Specification is what the model has to work out: E explicit (every step and volume given) ·
  P paper-assisted (steps given, paper fills gaps) · D paper-derived (the protocol must be extracted from the paper).
• Observed is how models did in Harbor (pass@1, 2026-10-04): solved by all · split · unsolved · not run.

A fourth, expert-labelled tag records why failures happen: bench-ready, needs training (bead handling, magnet timing),
or needs domain judgement (rules a paper does not state, such as not pipette-mixing competent cells).

Non-liquid-handling candidates have no pipetting structure, so they get a family instead: PL protocol program
(XDL, LabOP, Autoprotocol) · IM microscope acquisition · BR closed-loop culture control · OR workcell
orchestration and scheduling · SD self-driving lab campaign · OT other.

1. The 11 current tasks

Task	Structure	Spec	Observed	Opus / Sonnet / Fable	Tag	Description
split-200ul-two-wells	S1	E	Solved by all	1 / 1 / 1	Bench-ready	Calibration task: split one reservoir volume equally into two wells. Checks labware loading, pipette choice and a single transfer.
a1-a12-100ul	S1	E	Solved by all	1 / 1 / 1	Bench-ready	Calibration task: dispense the same volume into every well of one plate row. Checks well addressing and repeated dispensing.
ecoli-heat-shock-transformation	S2	E	Split	0.8 / 1 / 1	Domain judgement	Introduce a plasmid into competent E. coli in one tube and add recovery medium; the heat shock and incubation are operator steps. The pipetting is trivial; the catch is handling fragile cells gently.
ecoli-heat-shock-transformation-hard	S4	D	Not run	–	Domain judgement	The same transformation for eight plasmids on the thermocycler module, with volumes and the temperature program taken from the APEX paper.
colony-pcr-screening	S2	P	Solved by all	1 / 1 / 1	Bench-ready	Set up 96 colony PCR reactions to find which colonies carry the right plasmid: master mix, one colony and its primers per well, a fresh tip for every colony.
colony-pcr-screening-hard	S2	D	Not run	–	Training	The same 96-well colony PCR with quantities and order worked out from the Slowpoke paper and adapted to the deck's reagents.
ampure-bead-cleanup	S3	E	Solved by all	1 / 1 / 1	Training	Clean 96 PCR products with magnetic beads: bind, remove supernatant, two ethanol washes, dry, elute into a fresh plate. The largest task by robot actions.
golden-gate-assembly	S5	P	Solved by all	1 / 1 / 1	Training	AssemblyTron's DNA assembly: PCR seven fragments, remove template, join fragments into four plasmids by Golden Gate, then transform. Each stage's product feeds the next.
golden-gate-assembly-hard	S5	D	Not run	–	Training	The same AssemblyTron workflow with reaction quantities and fragment ratios derived from the paper.
opentrons-rna-extraction	S4	P	Split	0.6 / 0.4 / 0.8	Domain judgement	Extract viral RNA from 48 samples with magnetic beads, washes and a chilled elution plate, using two pipettes and two modules. All models recovered more eluate than the method specifies; Sonnet also added samples before the beads.
opentrons-rna-extraction-hard	S4	D	Not run	–	Domain judgement	The same extraction with volumes, order and timings taken from the paper.

Current tasks on the grid

	E Explicit	P Paper-assisted	D Paper-derived
S1 Single transfer	split ✅, a1-a12 ✅	–	–
S2 Plate set-up	heat shock ⚠️	colony PCR ✅	colony PCR hard ◻
S3 Wash cycles	AMPure ✅	–	–
S4 Module-coordinated	–	RNA extraction ⚠️	RNA hard ◻, heat shock hard ◻
S5 Multi-reaction workflow	–	Golden Gate ✅	Golden Gate hard ◻

✅ solved by all · ⚠️ split · ◻ not run. Every paper-derived task is unrun, and nothing at S1–S3 is paper-derived.
Failures come from domain judgement, not size: all models pass Golden Gate (S5) and AMPure (over 5,000 liquid moves).

2. The new candidates

All 210 new candidates are listed with a plain-English description in CANDIDATES.md
(also posted below as follow-up comments). Structure and family are assigned by the same keyword rules for every candidate,
so they are estimates until each task is built.

New liquid-handling candidates by structure and readiness

Structure	Ready	Partial	No usable code	Total
S1 Single transfer	1	3	2	6
S2 Plate set-up	13	32	18	63
S3 Wash cycles	4	9	1	14
S4 Module-coordinated	1	2	4	7
S5 Multi-reaction workflow	6	12	0	18

New non-liquid-handling candidates by family and readiness

Family	Ready	Partial	No usable code	Total
PL Protocol program (XDL, LabOP, Autoprotocol)	5	5	0	10
IM Microscope acquisition	7	8	0	15
BR Closed-loop culture control	1	5	2	8
OR Workcell orchestration / scheduling	3	6	2	11
SD Self-driving lab campaign	3	14	4	21
OT Other automation	0	12	25	37

Which specification levels the new candidates can support

Specification	Candidates
P + D	138
D (needs a written oracle)	51
E only (no paper)	21

Every candidate with both a paper and code (138) can become a paper-assisted and a paper-derived task, which fills the
empty D column above. The largest gains are in S2 plate set-ups (13 ready) and S5 multi-reaction workflows (6 ready);
S4 module-coordinated work stays thin (1 ready), so it needs hand-built tasks.

3. Where the models go wrong

A deterministic diff of each model's run log against the oracle's (tools/fault_detect.py, fault_report.md) agrees with the judge on every task:

Task	Opus 5.5	Sonnet 5.5	Fable 5.1
Heat shock	mixes the competent cells	clean	mixes the competent cells
RNA extraction	recovers 100 µL eluate (oracle 80)	samples before beads; 100 µL eluate; 5% extra waste removed	100 µL eluate

• Judge inconsistency: Opus and Fable do the same thing on heat shock but score 0.8 and 1.0. Make this a deterministic check.
• The RNA oracle is the authors' script, not the task spec: it waits longer and uses the magnet more than the task asks, and scores 0.6 itself.
• The simulator does not track liquid, so over-aspiration and harm to fragile cells pass simulation.

4. Size of the expanded benchmark

Scenario	Tasks
Today	11
+ 12 corpus papers that already have IR but no task	~61
+ new ready candidates (2 tasks each)	~149
+ all new paper + code candidates	~340–410
Theoretical ceiling (incl. partial, ACRA, AEGIS)	~500–600

Simulator check (VERIFIED.md): of 128 protocol files in the 25 ready liquid-handling repos, 30 pass opentrons_simulate unchanged (10 repos).
The 98 failures are setup issues: labware files on the author's disk (43), newer API (24), extra packages (15), missing inputs (5), other (11).

5. Recommendations

1. Run the four -hard variants in Harbor; they fill the D column and are the most demanding tasks.
2. Turn domain-rule judge items (gentle cell handling, elution volume) into deterministic checks; add a spec oracle for RNA extraction.
3. Build next from the 12 corpus papers with IR and the 10 repos that already simulate, prioritising D variants at S1–S3.
4. Re-run verification with labware vendored and a newer Opentrons release.
5. Add new grader families only where hardware-free execution exists: PL (Cronin ACRA, 100+ executed literature→XDL pairs), IM, OR.

[Tyronita]

All new expansion candidates (plain-English descriptions)

210 new candidates found by the four searches, after removing papers already in the corpus.
Structure (S1–S5, liquid handling) or family (non-liquid-handling) is assigned by fixed keyword rules over each
candidate's recorded robot work, so it is an estimate until the task is built. Specification: P = paper-assisted,
D = paper-derived, E = explicit steps. Simulator status is from VERIFIED.md (ready liquid-handling candidates only).

Structure levels: S1 single transfer · S2 plate set-up · S3 wash cycles · S4 module-coordinated · S5 multi-reaction workflow.
Families: PL protocol program · IM microscope acquisition · BR closed-loop culture control · OR workcell orchestration · SD self-driving lab campaign · OT other.

Ready (44)

Candidate	Level	Spec	Simulator	Licence	Description
evolver-2018	BR	P + D	not run	MIT	Precise, automated control of conditions for high-throughput growth of yeast and bacteria with eVOLVER. Robot work: Closed-loop culture control. Reported result: Turbidostat/chemostat growth curves and selection experiments across 16 vials reported with fitness measurements.
eda-event-driven	IM	P + D	not run	BSD-3-Clause	Event-driven acquisition for content-enriched microscopy. Robot work: Microscope acquisition. Reported result: Adaptive frame-rate imaging of mitochondrial and bacterial division events with reported photobleaching reduction.
etsted	IM	P + D	not run	GPL-3.0	Event-triggered STED imaging. Robot work: Microscope acquisition. Reported result: Event-triggered STED of vesicle/synaptic events with reported detection rates.
mantis-shrimpy	IM	P + D	not run	BSD-3-Clause	Mantis: High-throughput 4D imaging and analysis of the molecular and physical architecture of cells. Robot work: Microscope acquisition. Reported result: Multi-well 4D imaging of cell lines acquired and analysed with reported throughput.
navigate-2024	IM	P + D	not run	BSD-style (LICENSE.md)	Navigate: an open-source platform for smart light-sheet microscopy. Robot work: Microscope acquisition. Reported result: Smart acquisition (autofocus, tiling, decision-based imaging) demonstrated on multiple microscopes with images.
openflexure-autofocus	IM	P + D	not run	GPL-3.0	Fast, high-precision autofocus on a motorised microscope: Automating blood sample imaging on the OpenFlexure Microscope. Robot work: Microscope acquisition. Reported result: Automated malaria blood-smear tiling scans with autofocus success rates reported.
pycromanager-2021	IM	P + D	not run	BSD-3-Clause	Pycro-Manager: open-source software for customized and reproducible microscope control. Robot work: Microscope acquisition. Reported result: Applications in paper (adaptive illumination, light-sheet, high-throughput imaging) with acquired data.
pymmcore-plus-mda	IM	P + D	not run	BSD-3-Clause	pymmcore-plus + useq-schema (multi-dimensional acquisition engine). Robot work: Microscope acquisition. Reported result: Software; validated through downstream papers (e.g., Mantis PNAS Nexus 2024).
alabos-alab	OR	P + D	not run	MIT	AlabOS: a Python-based reconfigurable workflow management framework for autonomous laboratories (A-Lab). Robot work: Workcell orchestration / scheduling. Reported result: AlabOS ran A-Lab for months (>3500 samples) of solid-state synthesis; A-Lab Nature 2023 reported 41 synthesized targets (later disputed).
openworkstation	OR	P + D	not run	CC BY-SA (repo)	OpenWorkstation: A modular open-source technology for automated in vitro workflows. Robot work: Workcell orchestration / scheduling. Reported result: Automated microphysiometry/cell-culture workflow demonstrated with measurements.
slab-scheduling	OR	P + D	not run	MIT	Optimal Scheduling for Laboratory Automation of Life Science Experiments with Time Constraints. Robot work: Workcell orchestration / scheduling. Reported result: Schedules for cell-culture protocols computed and compared to baseline; executed on Maholo.
autoprotocol-dnada	PL	P + D	not run	Apache-2.0	Automated Platform for the Plasmid Construction Process (DNAda -> Autoprotocol). Robot work: Protocol program (XDL, LabOP, Autoprotocol). Reported result: Hundreds of plasmid constructions executed with reported success rates.
chemputer-acra	PL	P + D	not run	(LICENSE present,…	Verification and execution of the scientific literature via chemputation augmented by large language models. Robot work: Protocol program (XDL, LabOP, Autoprotocol). Reported result: LLM-generated XDL from literature procedures executed on the Chemputer with yields reported; >100 reference XDLs released.
chemputer-blueprints	PL	P + D	not run	see GitLab (unverified)	Reaction blueprints and logical control flow for parallelized chiral synthesis in the Chemputer. Robot work: Protocol program (XDL, LabOP, Autoprotocol). Reported result: Parallel chiral syntheses executed with reported yields/ee using XDL blueprints.
clairify-xdl	PL	P + D	not run	MIT	Large language models for chemistry robotics (CLAIRify: NL -> XDL). Robot work: Protocol program (XDL, LabOP, Autoprotocol). Reported result: Generated XDL executed on a Franka robot for solubility/recrystallisation; expert-evaluated XDL validity on 108 protocols.
labop	PL	P + D	not run	MIT	LabOP (Laboratory Open Protocol language, formerly PAML) - iGEM interlab growth-curve/calibration protocols. Robot work: Protocol program (XDL, LabOP, Autoprotocol). Reported result: Protocols used in the iGEM Interlab study (results published separately); executable via converters.
roche-pipetting-performance-ot2	S1	P + D	fails (setup)	GPL-3.0	Deep integration of low-cost liquid handling robots in an industrial pharmaceutical development environment. Robot work: liquid-handler QC: photometric pipetting accuracy/precision (dye dilution). Reported result: SLAS Technology 2024 (Roche pRED) reports OT-2 integration into AutoLab and pipetting accuracy/precision benchmarks (CV, systematic error) per volume.
blots-ot2	S2	P + D	fails (setup)	none	Bacterial Lysis Observation Over Time Assay (BLOTS) for High-Throughput Isolation and Screening. Robot work: 384-well assay plate assembly / lysis kinetics screening. Reported result: Assay plates assembled by the robot yielded time-resolved lysis curves used to recover and sequence isolates.
ester-library-synthesis	S2	P + D	fails (setup)	MIT	Repurposing amine and carboxylic acid building blocks with an automatable esterification reaction. Robot work: 96-well library synthesis, quench, solvent/salt screen. Reported result: 96-well esterification library synthesised on OT-2 with yields/conversions measured by LC-MS.
flat-agar-celegans-ot2	S2	P + D	fails (setup)	MIT	Flat agar and optimized media enable high-throughput C. elegans imaging in 96-well plates. Robot work: C. elegans bleach synchronisation, OP50 agar spotting, worm plating for imaging. Reported result: Pub reports flat-agar 96-well imaging quality (focus/worm detection) enabled by robot-prepared plates.
formulations-prep-shampoo	S2	P + D	fails (setup)	none	Accelerating Formulation Design via Machine Learning: Generating a High-throughput Shampoo Formulations Dataset. Robot work: viscous liquid formulation preparation with gravimetric verification. Reported result: Gravimetric analysis of robot-prepared formulations and downstream property measurements form the published dataset.
gfp-codon-burden-ot2	S2	P + D	7/15 files pass	MIT	Codon usage modulates the relationship between the burden and yield of protein overexpression. Robot work: bacterial growth/induction burden assay (IPTG titration, GFP + OD plate-reader). Reported result: J Biol Eng 2025 (preprint 10.1101/2024.11.28.625058) reports growth-rate burden vs GFP yield across codon-deoptimised strains; raw plate-reader data and processed data are in the repo.
greenaway-halogenation	S2	P + D	fails (setup)	BSD-3-Clause	An automated platform for tailored late-stage halogenation of pharmaceuticals. Robot work: reaction array setup, quench, NMR/LC-MS sample prep. Reported result: Screening arrays dispensed by OT-2 gave halogenation conversions/selectivities measured by NMR and LC-MS across substrate panels.
greenaway-mocs	S2	P + D	fails (setup)	none	Development of an Automated Workflow for Screening the Assembly and Host-Guest Behavior of Metal-Organic Cages. Robot work: solution-phase screening array + host-guest assay. Reported result: Automated screen identified assembled cages confirmed by MS/NMR and host-guest binding measured across the array.
greenaway-pocs	S2	P + D	fails (setup)	none	Streamlining the automated discovery of porous organic cages. Robot work: cage synthesis plate + analytical transfer protocols. Reported result: Automated workflow produced porous organic cages characterised by HRMS/NMR, with hit rates reported vs manual discovery.
lucks-cellfree-biosensor	S2	P + D	fails (setup)	none	Semi-automated production of cell-free biosensors. Robot work: cell-free reaction assembly (384-well biosensor build). Reported result: A full 384-well plate of fluoride-sensing cell-free reactions assembled robotically matched expected colorimetric/fluorescent detection performance.
lysostaphin-turbidity	S2	P + D	fails (setup)	AGPL-3.0	Turbid but accurate: automating lysostaphin quantification including uncertainty quantification. Robot work: antibacterial activity turbidity assay (serial dilution + OD lysis kinetics). Reported result: Turbidity assay quantified lysostaphin over a 0.63-10 mg/L dynamic range with Bayesian uncertainty quantification using autoclaved S. carnosus indicator cells.
msd-biochem-assays-ot2	S2	P + D	passes	Apache-2.0	Automation of biochemical assays using an open-sourced, inexpensive robotic liquid handler. Robot work: Bradford and PicoGreen standard curves / quantitation assays. Reported result: SLAS Technology 2024 paper compares OT-2 Bradford/PicoGreen standard curves and sample quantitation to manual (linearity, CV, accuracy).
plant-phenotyping-ot2	S2	P + D	fails (setup)	MIT	Automated Discovery of Multicellular Behavior for Optimized Plant Growth and Climate Resilience. Robot work: plate filling and treatment dosing for plant phenotyping. Reported result: Automated phenotyping screen measured growth responses used to select stress-resilient conditions.
sapp-dmx-protein-design	S2	P + D	fails (setup)	MIT	Accelerating protein design by scaling experimental characterization. Robot work: SEC fraction pooling and normalisation for protein characterisation. Reported result: High-throughput expression/characterisation of designed proteins; thousands of designs assayed with the automated pipeline.
cayman-his-elisa	S3	P + D	fails (setup)	MIT	An Opentrons OT-2 workflow for the Cayman His-tag detection ELISA kit. Robot work: ELISA sample/standard loading, plate washing, substrate addition. Reported result: Standard curves and His-tag quantification generated from the automated ELISA runs in the Arcadia protocol record.
massspecpreppy	S3	P + D	3/10 files pass	GPL-3.0	MassSpecPreppy - an end-to-end solution for automated protein concentration determination and flexible sample digestion for proteomics. Robot work: BCA assay, reduction/alkylation, SP3 digestion, Evotip loading. Reported result: BCA recovery 101.3% (SD 7.8) for the automated workflow vs 96.3% manual; 94.5% of protein groups passed thresholds for OT-2 digestion.
roboctab	S3	P + D	fails (setup)	none	High-Throughput DNA Extraction Using Robotic Automation (RoboCTAB) for Large-Scale Genotyping. Robot work: CTAB plant DNA extraction. Reported result: DNA yield/purity and genotyping call rates from robot-extracted samples compared with manual CTAB extraction.
tidytron	S3	P + D	3/15 files pass	Apache-2.0	TidyTron: Reducing lab waste using validated wash-and-reuse protocols for common plasticware in Opentrons OT-2 robots. Robot work: plasticware wash-and-reuse / tip and plate cleaning. Reported result: Carryover after wash validated by qPCR/transformation and fluorescence assays showing no detectable cross-contamination vs fresh plasticware.
hyrad-adna-capture	S4	P + D	passes	none	Performance and automation of ancient DNA capture with RNA hyRAD probes. Robot work: hybridization capture + magnetic bead washes (magnetic module, thermocycler). Reported result: Capture efficiency and on-target read fractions for automated vs manual hybridization reported for ancient DNA libraries.
ant-cloning	S5	P + D	7/10 files pass	none	Automated cloning by natural transformation (ANT) OT-2 workflow. Robot work: PCR, digestion, overlap-extension PCR, natural transformation, plating. Reported result: Cloning efficiency and correct-assembly rates for automated natural transformation reported in the paper.
beckham-enzyme-pipeline	S5	P + D	2/15 files pass	none (no LICENSE file)	Enabling high-throughput enzyme discovery and engineering with a low-cost, robot-assisted pipeline. Robot work: E. coli transformation, inoculation, expression, lysis, enzyme assay. Reported result: Robot-assisted pipeline screened enzyme variant libraries; activity/expression measured and compared with manual handling.
culturomics-isolation	S5	P + D	2/11 files pass	none	Modular automated high-throughput isolation and phylogenetic identification of bacteria from complex communities. Robot work: microvolume DNA extraction, two-step amplicon PCR, bead purification, pooling. Reported result: Isolates recovered and phylogenetically identified from automated plates; sequencing success rates reported vs manual pipeline.
damplab-moclo-transformation	S5	P + D	no protocol file	BSD-3-Clause	A linear programming-based strategy to save pipette tips in automated DNA assembly (with DAMP Lab MoClo/transformation workflow). Robot work: MoClo Golden Gate assembly + E. coli transformation + plating. Reported result: Tip-saving assembly strategy evaluated on OT-2 runs; assembly/transformation success measured against standard tip usage.
exfab-qiaseq-flex	S5	P + D	passes	GPL-3.0	Automated QIAseq FX DNA Library Preparation on the Opentrons Flex. Robot work: NGS library preparation (fragmentation, adapter ligation, bead cleanup, PCR). Reported result: Protocol entry reports libraries prepared for Illumina short-read sequencing of bee gut samples (sequencing-ready yields).
jubiotech-ot2-autoclone	S5	P + D	2/7 files pass	AGPL-3.0	Accelerated strain construction and characterization of C. glutamicum protein secretion by laboratory automation. Robot work: Golden Gate assembly, heat-shock transformation, plasmid prep, plating. Reported result: Automated strain construction produced verified C. glutamicum secretion strains characterised in microbioreactor cultivations.
chemputer-peptide	SD	P + D	not run	Zenodo (check)	Universal peptide synthesis via solid-phase methods fused with chemputation. Robot work: Self-driving lab campaign. Reported result: Synthesised multiple peptides (incl. semaglutide-type) with purity/yield reported.
panda-bear	SD	P + D	not run	GPL-3.0	PANDA: a self-driving lab for studying electrodeposited polymer films. Robot work: Self-driving lab campaign. Reported result: BO campaign optimised PEDOT:PSS electrochromic switching; data reported.
sdl-demo-clslab	SD	P + D	not run	MIT	What is a minimal working example for a self-driving laboratory? (SDL-Demo / CLSLab:Light). Robot work: Self-driving lab campaign. Reported result: Optimisation of LED mixing to match target spectra; simulator vs hardware comparison reported.

Partial (108)

Candidate	Level	Spec	Simulator	Licence	Description
chibio-2020	BR	P + D	not run	none (no LICENSE file)	In situ characterisation and manipulation of biological systems with Chi.Bio. Robot work: Closed-loop culture control. Reported result: Turbidostat growth, optogenetic feedback control and fluorescence measurements demonstrated with reported traces.
evolver-ddr-elife	BR	P + D	not run	MIT	Environmental fluctuations reshape an unexpected diversity-disturbance relationship in a microbial community (eVOLVER). Robot work: Closed-loop culture control. Reported result: Fluctuating dilution regimes run on eVOLVER with community composition measured by sequencing.
photobioreactor-fluorescence	BR	P + D	not run	unverified	Integration of Fluorescence Spectroscopy into a Photobioreactor for the Monitoring of Cyanobacteria. Robot work: Closed-loop culture control. Reported result: Cyanobacterial growth monitored by fluorescence with reported correlation to OD.
pioreactor	BR	P + D	not run	MIT	Pioreactor open-source bioreactor software (used in published studies). Robot work: Closed-loop culture control. Reported result: No first-party results paper; third-party papers (e.g., 10.1186/s12934-025-02670-8) use it for culture experiments.
reacsight	BR	P + D	not run	check Zenodo	Enhancing bioreactor arrays for automated measurements and reactive control with ReacSight. Robot work: Closed-loop culture control. Reported result: Automated cytometry-based optogenetic control of yeast populations and co-cultures with reported dynamics.
brightfield-biofilm	IM	P + D	not run	none	Label-free microscopy enables high-throughput identification of genes controlling biofilm development. Robot work: Microscope acquisition. Reported result: Genome-wide biofilm screen with hit genes reported.
culturomics-colony-picking	IM	P + D	not run	unverified	High-throughput microbial culturomics using automation and machine learning (CAMII colony picker). Robot work: Microscope acquisition. Reported result: Isolated >26,000 gut isolates with morphology-guided picking efficiency reported.
daxi-lightsheet	IM	P + D	not run	CC BY (repo)	DaXi: high-resolution, large imaging volume and multi-view single-objective light-sheet microscopy. Robot work: Microscope acquisition. Reported result: Zebrafish/drosophila volumetric imaging reported.
fish-fluidics	IM	P + D	not run	unverified	An open source 16-channel fluidics system for automating sequential FISH-based imaging. Robot work: Microscope acquisition. Reported result: seqFISH rounds automated with imaging results.
imswitch	IM	P + D	not run	GPL-3.0	ImSwitch: generalizing microscope control in Python. Robot work: Microscope acquisition. Reported result: Software paper; underpins etSTED and UC2 experiments.
openframe-openaf	IM	P + D	not run	BSD	openFrame: modular open microscopy platform with single-shot optical autofocus. Robot work: Microscope acquisition. Reported result: Autofocus accuracy/stability characterised and used in HCA imaging.
phage-cocktails-robotics	IM	P + D	not run	check Zenodo	High-throughput methods leveraging robotics and computer vision for the development of therapeutic phage cocktails. Robot work: Microscope acquisition. Reported result: Phage cocktails designed and validated with plaque/plate-reader data.
pysted	IM	P + D	not run	CC (Attribution, check NC)	Development of AI-assisted microscopy frameworks through realistic simulation with pySTED. Robot work: Microscope acquisition. Reported result: RL policies trained in simulation were deployed on a real STED microscope with image quality reported.
ac-training-lab	OR	P + D	not run	MIT	Acceleration Consortium Training Lab (remote-access SDL modules incl. Hello-World, sample transfer). Robot work: Workcell orchestration / scheduling. Reported result: Educational SDL modules; no single results paper verified.
chemos-2020	OR	P + D	not run	Apache-2.0	ChemOS: An orchestration software to democratize autonomous discovery. Robot work: Workcell orchestration / scheduling. Reported result: Demonstrates closed-loop optimisation on a cocktail-mixing robot and HPLC calibration, with reported optimisation traces.
madsci-ad-sdl	OR	P + D	not run	MIT	MADSci: A modular Python-based framework to enable autonomous science. Robot work: Workcell orchestration / scheduling. Reported result: Framework; experiments reported via ANL RPL deployments but no single peer-reviewed results paper verified.
mobile-robots-dai-2024	OR	P + D	not run	Zenodo (check per record)	Autonomous mobile robots for exploratory synthetic chemistry. Robot work: Workcell orchestration / scheduling. Reported result: Supramolecular/structural diversification chemistry run autonomously with UPLC-MS/NMR-based decisions; hits reported.
star-hardwarex	OR	P + D	not run	check Zenodo	Sequential Treatment Application Robot (STAR) for high-replication marine experimentation. Robot work: Workcell orchestration / scheduling. Reported result: Marine organism treatment experiments with high replication reported.
zhao-autonomous-enzyme	OR	P + D	not run	MIT	A generalized platform for artificial intelligence-powered autonomous enzyme engineering (iBioFAB). Robot work: Workcell orchestration / scheduling. Reported result: Engineered AtHMT and YmPhytase variants with 16-26x improvements in 4 rounds.
chemos2-atlas	OT	P + D	not run	MIT	Atlas: a brain for self-driving laboratories (BO library used in ChemOS 2.0 campaigns). Robot work: Other automation. Reported result: BO planners benchmarked on emulators and deployed on SDL campaigns (e.g., electrochemistry with ChemOS 2.0).
covid-mastermix-omix	OT	D (needs a written oracle)	not run	none	Development and validation of automated methods for COVID-19 PCR Master Mix preparation. Robot work: Other automation. Reported result: Automated PCR master mix preparation matched manual pipetting accuracy in a diagnostic lab (Ct value comparisons, gravimetric checks).
duckbot-jubilee	OT	P + D	not run	MIT	The Duckbot: A system for automated imaging and manipulation of duckweed. Robot work: Other automation. Reported result: Automated growth assays measured duckweed area/growth across treatments.
icfree-cfps	OT	P + D	not run	MIT	An AI-driven workflow for the accelerated optimization of cell-free protein synthesis. Robot work: Other automation. Reported result: Iterative ML rounds improved cell-free protein yields measured on plate reader across generated compositions.
labdroid-cell-culture	OT	P + D	not run	none	Robotic search for optimal cell culture in regenerative medicine (LabDroid Maholo, iPSC-RPE). Robot work: Other automation. Reported result: BO over 7 culture parameters improved iPSC-RPE differentiation scores (pigmentation) in 111 days.
llm-rdf-synthesis	OT	P + D	not run	MIT (per README badge)	An automatic end-to-end chemical synthesis development platform powered by large language models. Robot work: Other automation. Reported result: End-to-end synthesis campaigns executed with reported yields and optimised conditions.
memplex-printing	OT	P + D	not run	MIT	Designer artificial environments for membrane protein synthesis. Robot work: Other automation. Reported result: Thousands of cell-free membrane-protein reactions printed and measured; formulations identified.
myra-cfe-platform	OT	P + D	not run	none	A liquid handling platform for standardized quantification of cell-free enzymatic activity. Robot work: Other automation. Reported result: Standardised enzymatic activity quantification reproducible across operators/platforms reported in the paper.
polymerization-planner	OT	P + D	not run	MIT	Automation-Assisted Photoinduced Atom Transfer Radical Polymerization. Robot work: Other automation. Reported result: Polymerisations run from planner recipes characterised by GPC (conversion, dispersity).
sdl-guide-gormley	OT	P + D	not run	MIT	A user's guide to your first self-driving liquid handling lab. Robot work: Other automation. Reported result: Closed-loop campaign optimised enzyme assay conditions with plate-reader readouts.
spatial-tx-bravo	OT	D (needs a written oracle)	not run	none	Automation of Spatial Transcriptomics library preparation to enable rapid and robust insights into spatial organization of tissues. Robot work: Other automation. Reported result: Automated ST library prep cut prep time by 35%, increased samples per run and showed near-zero technical variability between preparations.
teemi	OT	P + D	not run	MIT	teemi: open-source literate programming for iterative DBTL cycles. Robot work: Other automation. Reported result: Yeast strain library built and screened with titres reported.
aquarium	PL	P + D	not run	MIT	Aquarium laboratory operating system (UW BIOFAB) - used in SD2 robustness studies. Robot work: Protocol program (XDL, LabOP, Autoprotocol). Reported result: Aquarium-run yeast logic circuit experiments with cytometry outcomes reported.
autoprotocol-python	PL	P + D	not run	BSD-3-Clause (LICENSE.rst)	Autoprotocol-python (Strateos/Transcriptic protocol IR). Robot work: Protocol program (XDL, LabOP, Autoprotocol). Reported result: Specification; experimental results come from cloud-lab papers (e.g., SD2 replication papers).
coscientist	PL	P + D	not run	(LICENSE present,…	Autonomous chemical research with large language models (Coscientist). Robot work: Protocol program (XDL, LabOP, Autoprotocol). Reported result: Suzuki/Sonogashira couplings executed and confirmed by GC-MS; plate-reader colour tasks executed.
sample-protein-sdl	PL	P + D	not run	check Zenodo	Self-driving laboratories to autonomously navigate the protein fitness landscape (SAMPLE). Robot work: Protocol program (XDL, LabOP, Autoprotocol). Reported result: Four agents converged on glycoside hydrolases >12 C more thermostable.
sd2-yeast-replication	PL	P + D	not run	none	Highly-automated, high-throughput replication of yeast-based logic circuit design assessments (SD2/Strateos/Aquarium). Robot work: Protocol program (XDL, LabOP, Autoprotocol). Reported result: Logic-circuit assays replicated across labs with cytometry results compared.
aliquot-niaid	S1	E only (no paper)	not run	none	ALIQUOT: Automated Liquid handling Infection Quantification Using OpenTrons (NIH/NIAID QVEU). Robot work: virus infection quantification: serial dilution, fix, immunostain (OT-2 + Flex).
labscriptai-protein-eng	S1	P + D	not run	none	Autonomous liquid-handling robotics scripting through large language models enables accessible and safe protein engineering workflows. Robot work: LLM protocol generation; cell-free GFP variant characterisation; enzyme screening; fluorescence calibration. Reported result: Fluorescence standard curves CV<5%, R2>0.999; 318 GFP variants characterised (best 5.0x avGFP); enzyme double mutants >3x catalytic efficiency.
olh-turbidostat-plr	S1	P + D	not run	none	A Purpose-Built Open Source Liquid Handler for Industry-Class Automated Experiments. Robot work: turbidostat / continuous culture; pipetting accuracy calibration. Reported result: OLH pipetting MAE 4.1 vs 7.2 uL manual (p<0.005); closed-loop turbidostat held ~200 cultures at OD setpoints.
aml-drug-synergy-pyhamilton	S2	P + D	not run	none	High-throughput approaches to uncover synergistic drug combinations in leukemia. Robot work: drug combination dilution matrix and dosing for synergy screening (leukemia cells). Reported result: Multidimensional drug-combination screens identified synergistic combinations in leukemia (H3K27 modification-related), with viability readouts and synergy scores.
artic-normalisation-esr	S2	E only (no paper)	not run	none	ESR-NZ ARTIC normalisation/end-repair/barcoding (OT-2). Robot work: SARS-CoV-2 ARTIC DNA normalisation + nanopore end-repair/barcoding.
auto-sp-proteomics	S2	D (needs a written oracle)	not run	n/a (no repo)	AUTO-SP: Automated Sample Preparation for Analyzing Proteins and Protein Modifications. Robot work: proteomics BCA, digestion and PTM enrichment. Reported result: AUTO-SP automated BCA, digestion and PTM enrichment on PDX breast tumour tissue, with protein/PTM IDs comparable to manual.
detectr-covid-ot2	S2	E only (no paper)	not run	none	Mammoth DETECTR COVID-19 assay on OT-2. Robot work: CRISPR-Cas12 diagnostic (RNP prep + DETECTR reaction).
dtu-synbio-automation	S2	P + D	not run	MIT	DTU Synbio lab automation protocol collection. Robot work: teaching/library protocols (agar plating, liquid-class optimisation, parameterised transfers). Reported result: No published results; teaching and internal-use repository.
evotron-evolver-sampling	S2	P + D	not run	none	Dynamic cybergenetic control of bacterial co-culture composition via optogenetic feedback. Robot work: automated sampling from eVOLVER bioreactors into plates. Reported result: Closed-loop optogenetic control held co-culture ratios at setpoints, measured by cytometry/OD from robot-sampled timepoints.
exposomics-384-flex	S2	P + D	not run	MIT	Opentrons Flex 384-well protein-precipitation sample preparation for nontargeted exposomics. Robot work: protein precipitation / dilution in 384-well format for LC-HRMS. Reported result: README reports up to 288 biological samples plus QC per plate feeding multiple LC-HRMS assays, but no peer-reviewed results found.
flex-genomics-di-omics	S2	P + D	not run	none	Opentrons Flex whole-genome / methylation / TIP-seq workflows (di-omics). Robot work: NGS library prep on Flex with off-deck handoffs. Reported result: README tracks validation status per protocol but no publication.
hydrogel-crosslink-ot2	S2	E only (no paper)	not run	none	Semi-autonomous synthesis and modelling of chemically cross-linked hydrogels (OT-2). Robot work: materials: hydrogel formulation synthesis.
imperial-colony-picker	S2	P + D	not run	none	Automated colony picker add-on for the OT-2 (Imperial SSB MRes; cited by Slowpoke). Robot work: colony PCR reagent prep and colony resuspension. Reported result: Vision-guided colony picking demonstrated on OT-2; results reported only in the project write-up and via the Slowpoke citation.
ipsc-robot-pyhamilton	S2	P + D	not run	none	Automated hiPSC culture and sample preparation for 3D live cell microscopy. Robot work: iPSC culture: media exchange, seeding, imaging scan. Reported result: Code implements Allen Institute iPSC culture protocols (bioRxiv 2020.12.18.423371); robot-specific results not in repo.
krisp-sars2-pooling	S2	P + D	not run	MIT	Implementation of an efficient SARS-CoV-2 specimen pooling strategy for high throughput diagnostic testing. Robot work: SARS-CoV-2 hypercube sample pooling. Reported result: Infected samples reliably detected in pools of 27 and 81 with minimal sensitivity loss for individual Ct up to 32.
metclo-ot2	S2	P + D	not run	none	MetCloOT2: Automated MetClo assembly on Opentrons OT-2 (MSc project, based on Lin & O Callaghan 2018/2020). Robot work: DNA assembly (MetClo methylation-based Golden Gate). Reported result: Underlying MetClo method papers (NAR 2018; Methods Mol Biol 2020) report manual assembly results; OT-2 implementation is an MSc project without published validation.
myrplate-flex-casting	S2	P + D	not run	MIT	myrPlate engineered connective tissue casting on Opentrons Flex. Robot work: collagen/cell suspension casting into 48-well tissue plates. Reported result: README describes casting engineered connective tissues; no publication located.
nioz-ot2-database	S2	E only (no paper)	not run	none	MolLabNIOZ/OT2 protocol database (NIOZ molecular ecology). Robot work: amplicon PCR setup, Qubit quantification, dilution, pooling.
openblend-media	S2	P + D	not run	MIT	Modular development enables rapid design of media for alternative hosts. Robot work: combinatorial media blending into microplates. Reported result: Blended media designs were tested in growth experiments for alternative microbial hosts, with growth rates reported per formulation.
ot2-cherrypick	S2	P + D	not run	GPL-3.0	OT2-CherryPick: A Zero-Install Web Platform for Orchestrating Complex Liquid Handling on the Opentrons OT-2. Robot work: CSV-driven cherry-picking / liquid transfer generator. Reported result: ChemRxiv preprint describing a GUI/CSV cherry-picking platform; validation is mainly simulation/operational, check for gravimetric data.
ot2-yeast-spot-assay	S2	P + D	not run	none	Automation of yeast spot assays using an affordable liquid handling robot. Robot work: yeast serial dilution spot assay on agar (3D-printed Petri mount, z-adjust). Reported result: Automated spot assays of budding yeast reproduced manual growth phenotypes, imaged with a flatbed scanner (spot intensity quantified).
pac832-galr1-ot2	S2	E only (no paper)	not run	none	PAC-832: A selective small-molecule antagonist of galanin receptor 1 (GalR1) - supplementary data and OT-2 protocols. Robot work: cell-based GPCR pharmacology assays (cAMP, acetylcholine release) dose-response plating. Reported result: Repo includes raw and processed cAMP/ACh assay data for PAC-832 vs galanin; paper DOI not yet resolved (likely preprint).
pasta-spatial	S2	P + D	not run	none	PASTA: versatile tyramide-oligonucleotide amplification for multimodal spatial biology. Robot work: spatial proteomics staining (tyramide-oligo amplification cycles on slides). Reported result: High-plex protein and RNA codetection on FFPE tissue via cyclical tyramide-oligonucleotide amplification and imaging (Nature Methods figures).
peg-solubility-assay	S2	P + D	not run	none	An open-source automated PEG precipitation assay to measure the relative solubility of proteins with low material requirement. Robot work: protein solubility assay (PEG titration, plate prep and transfer). Reported result: Relative solubility of proteins/antibodies quantified from PEG-precipitation midpoints using ~200 ug material; accuracy compared against established methods.
pozzo-ot2-doe	S2	P + D	not run	MIT	Pozzo group OT-2 design-of-experiments protocols. Robot work: formulation DoE sample preparation. Reported result: Results appear in separate group publications, not this repo.
prism-protocol-gen	S2	P + D	not run	MIT	PRISM: Protocol Refinement through Intelligent Simulation Modeling. Robot work: LLM protocol generation + simulation validation; PCR master mix/sample prep. Reported result: Digital Discovery 2026: LLM-generated, simulation-refined protocols executed on OT-2 (PCR prep).
pvlab-solubility-dropcast-ot2	S2	E only (no paper)	not run	MIT	PV-Lab Robotic_Liquid_Handling (MIT AMLS OT-2 solubility testing / drop casting). Robot work: materials: solubility testing, thin-film drop casting.
retro-10x-ot2	S2	E only (no paper)	not run	none	OT-2 10x scRNA-seq library prep (Retro Biosciences). Robot work: 10x Genomics single-cell library prep.
robotic-titration-cv-ot2	S2	P + D	not run	MIT	High-throughput robotic colourimetric titrations using computer vision. Robot work: chemistry: colourimetric acid-base titration with webcam endpoint detection. Reported result: Digital Discovery paper reports titration accuracy vs manual/reference concentrations using OT-2 + webcam colour analysis.
rtwizard-antigen-ot2	S2	E only (no paper)	not run	none	rtWIZARD: OT-2 running 42 rapid antigen tests in parallel. Robot work: lateral-flow rapid antigen test loading (ASF, PPA).
sdl-enzyme-flex	S2	D (needs a written oracle)	not run	MIT	Optimized Machine Learning for Autonomous Enzymatic Reaction Intensification in a Self-Driving Lab. Robot work: self-driving enzymatic reaction optimisation (pH, temperature, cosubstrate). Reported result: ML-driven Opentrons Flex platform optimised enzymatic reaction conditions after >10,000 simulated campaigns selected the best algorithm; experimental campaigns improved activity.
trypanosomatid-drug-combo-ot2	S2	E only (no paper)	not run	none	OT-2 protocols for drug combination screening (Trypanosomatics Lab, UNSAM). Robot work: drug dilution/combination matrix, cell lysis, 384-well transfer (luciferase readout). Reported result: README states protocols are described in the lab paper (not cited by DOI).
vadr-vacuum-miniprep-flex	S2	P + D	not run	none	An Open-Source 3D-Printed Vacuum Manifold for Automated DNA Isolation Improves Yields in the Opentrons Flex and OT-2. Robot work: plasmid miniprep with 3D-printed vacuum manifold. Reported result: bioRxiv 2026 reports improved DNA isolation yields with the open-source vacuum manifold on Opentrons Flex (link between repo and preprint inferred from content/date; verify).
viscous-liquid-bo-ot2	S2	P + D	not run	MIT	Optimization of liquid handling parameters for viscous liquid transfers with pipetting robots, a sticky situation. Robot work: liquid-class optimisation (viscous liquids, gravimetric, Bayesian optimisation). Reported result: Digital Discovery 2024 reports gravimetric transfer error and speed after multi-objective BO for viscous liquids on OT-2 and Sartorius rLine.
worm-food-preference	S2	P + D	not run	GPL-3.0	C. elegans food preference protocol (bacteria and worm spotting on OT-2). Robot work: spotting bacteria and worms onto agar plates. Reported result: Protocol reproduces a published behavioural assay; no new results in the repo.
agentex-genetic-codes	S3	P + D	not run	none	Automated prototyping of genetic codes. Robot work: in vitro translation reaction assembly, RNA purification. Reported result: Automated prototyping produced and characterised reassigned genetic codes with mass-spec/translation readouts.
bean-kasp-anthracnose	S3	P + D	not run	none	Automating high-throughput screening for anthracnose resistance in common bean using allele specific PCR. Robot work: plant DNA extraction + KASP/PACE allele-specific PCR genotyping. Reported result: Automated DNA extraction and KASP assays genotyped common bean lines at four anthracnose resistance markers with clear allele clustering, matching expected genotypes.
cfrna-pipeline	S3	P + D	not run	none	An inexpensive semi-automated sample processing pipeline for cell-free RNA (extraction). Robot work: cell-free RNA extraction and cleanup. Reported result: RNA yields and sequencing library quality from the semi-automated pipeline compared with manual extraction.
damplab-taqpath-ot2	S3	P + D	not run	none	Semi-Automated and Miniaturized SARS-CoV-2 Detection using TaqPath COVID-19 Multiplex RT-PCR (BU DAMP Lab). Robot work: MagMAX RNA extraction + TaqPath RT-qPCR setup.
har1eyk-starlet-protein-extraction	S3	E only (no paper)	not run	none	hamilton-starlet-scripts (PyLabRobot STARlet protein extraction / Orange G volume tests). Robot work: protein extraction; volume QC (Orange G).
opentrons-covid19-dev	S3	P + D	not run	Apache-2.0	Opentrons/covid19 (OT-2 COVID-19 RT-qPCR station development repo, Opentrons + Open Medicine Institute). Robot work: SARS-CoV-2 RNA extraction + RT-qPCR setup stations A/B/C. Reported result: Experiment folders contain LOD/contamination study notes; formal results only on protocols.io, no peer-reviewed paper.
ot2-proteomics-thesis	S3	E only (no paper)	not run	none	Automated Versus Manual Proteomics Sample Preparation: A Comparative Evaluation of Liquid Handling Workflows (MSc thesis). Robot work: proteomics: BCA assay + SP3 digestion. Reported result: Thesis raw data in repo comparing OT-2 vs human BCA/SP3 performance; not peer reviewed.
tape-organoid-barcoding	S3	P + D	not run	none	Cell-type-specific signaling networks in heterocellular organoids (TOBis barcoding / robotic organoid extraction). Robot work: mass-cytometry barcode production/aliquoting, organoid extraction. Reported result: TOBis barcoded organoids profiled by mass cytometry; signalling networks resolved per cell type.
ufdutton-16s-flex	S3	E only (no paper)	not run	none	UF Dutton Lab Opentrons 16S microbiome workflow (OT-2 + Flex). Robot work: microbiome DNA extraction + 16S/shotgun library prep.
ht-hcr-sea-urchin	S4	P + D	not run	none	Automated, high-throughput in situ hybridization of sea urchin (Lytechinus pictus) embryos. Robot work: whole-mount in situ HCR on embryos in 96-well plates. Reported result: Processed 192 probe sets in 32 h; produced localisation data for 101 genes across three developmental stages.
tmt-bead-platform-ot2	S4	D (needs a written oracle)	not run	n/a (no repo)	A Semiautomated Paramagnetic Bead-Based Platform for Isobaric Tag Sample Preparation. Robot work: TMTpro isobaric-tag proteomics sample prep. Reported result: Robot-prepared TMTpro16-plex triplicates quantified similar numbers of proteins/peptides with good reproducibility vs manual.
ai4pd-flex-protein-purification	S5	E only (no paper)	not run	MIT	Opentrons Flex protein purification and molecular biology scripts (AI4PD Lab). Robot work: protein expression, Ni-NTA purification, TEV digestion, Golden Gate, miniprep.
amplicon-seq-bisanz	S5	E only (no paper)	not run	none	AmpliconSeq (16S amplicon library OT-2 scripts, Bisanz lab). Robot work: 16S amplicon library prep (extraction wash, PCR, indexing, pooling, AMPure).
angelovangel-ot2-ngs	S5	E only (no paper)	not run	none	angelovangel/opentrons (KAUST Bioscience core NGS/Sanger OT-2 + Flex protocols). Robot work: Sanger PCR setup, ONT rapid library, DNA normalisation, PacBio Kinnex PCR.
biomationscripter	S5	P + D	not run	MIT	BiomationScripter: Python library for scripting bio-automation protocols. Robot work: protocol-generation library (templates for PCR, dilution, transformation). Reported result: No wet-lab results tied to a publication found.
ecoli-protein-expr-purif-ot2	S5	E only (no paper)	not run	none	OT-2 assisted recombinant protein expression and purification in E. coli (Norton-Baker). Robot work: protein expression: transformation, inoculation, lysis, purification transfers. Reported result: README says protocols are described in a paper (citation not resolved in README head); likely reports expression/purification yields.
evogen-ot-library	S5	E only (no paper)	not run	none	EvoGenOT (UCPH EvoGenomics Opentrons protocol library). Robot work: DNA/RNA co-extraction (D-Rex), BEST library build, bead purification.
locost-covid-nanopore-ot2	S5	E only (no paper)	not run	none	Opentrons_LoCost_Covid19 (ARTIC LoCost nanopore library prep on OT-2). Robot work: SARS-CoV-2 nanopore library prep (dilution, end-repair, barcode ligation, pooling).
ot2-moclo-jove	S5	E only (no paper)	not run	MIT	Modular Cloning with the OT2 (OT2_MoClo_JoVE). Robot work: MoClo Golden Gate, heat-shock transformation, plating, colony picking, bead miniprep. Reported result: Repo targets a JoVE article (not resolved in Crossref); JoVE articles typically include representative results.
parseq-pyhamilton	S5	P + D	not run	none	parSEQ: Probe and Rescue Sequencing for Advanced Variant Retrieval from DNA Pool. Robot work: variant library sequencing pipeline: barcoding PCR, cherry picking, Illumina/nanopore library prep. Reported result: bioRxiv 2023 parSEQ preprint reports retrieval of sequence-verified variants from pooled DNA libraries using the Hamilton pipeline.
polar-sarscov2-ot2	S5	P + D	not run	MIT	POLAR: Pathogen-Oriented Low-cost Assembly & Re-sequencing SARS-CoV-2 (OT-2 automation). Robot work: viral RNA magbead extraction + SARS-CoV-2 amplicon library prep.
pudu-synbio-ot2	S5	E only (no paper)	not run	MIT	PUDU (Protocol Unified Design Unit) - OT-2 synthetic biology workflows. Robot work: Golden Gate / Loop assembly, transformation, plating, iGEM GFP/OD calibration.
roborna-nirvana	S5	P + D	not run	none	An open-source, automated, and cost-effective platform for COVID-19 diagnosis and rapid portable genomic surveillance using nanopore sequencing. Robot work: SARS-CoV-2 magnetic-bead RNA extraction + RT-qPCR + nanopore variant sequencing. Reported result: Automated extraction showed high reproducibility with no cross-contamination; limit of detection, sensitivity and specificity comparable to commercial assays; NIRVANA identified circulating variants.
a-lab-2023	SD	P + D	not run	none (ARROWS); see repos	An autonomous laboratory for the accelerated synthesis of novel materials (A-Lab). Robot work: Self-driving lab campaign. Reported result: Claimed 41/58 novel compounds synthesized in 17 days (phase claims later contested).
ada-thin-films	SD	P + D	not run	MIT	Self-driving laboratory for accelerated discovery of thin-film materials (Ada). Robot work: Self-driving lab campaign. Reported result: Optimised hole-transport-layer pseudomobility via closed-loop BO; data released per campaign.
afion-plasmonic	SD	P + D	not run	none	Self-driving lab for the photochemical synthesis of plasmonic nanoparticles (AFION). Robot work: Self-driving lab campaign. Reported result: Targeted Ag nanoparticle morphologies/spectra achieved autonomously with reported spectra.
ahtech-electrochem	SD	P + D	not run	MIT	A high-throughput experimentation platform for data-driven discovery in electrochemistry (AHTech). Robot work: Self-driving lab campaign. Reported result: Thousands of CVs on redox mediators with discovered reactivity trends reported.
bear-mechanics	SD	P + D	not run	(LICENSE present,…	Superlative mechanical energy absorbing efficiency discovered through self-driving lab-human partnership (BEAR). Robot work: Self-driving lab campaign. Reported result: Record 75% energy-absorbing efficiency structures discovered; >25,000 experiments.
chemcrow	SD	P + D	not run	MIT	Augmenting large language models with chemistry tools (ChemCrow). Robot work: Self-driving lab campaign. Reported result: Executed syntheses of DEET and three thioureas plus a chromophore on RoboRXN, with characterisation.
chemputer-modules-acscs	SD	P + D	not run	unverified	Robotic Modules for the Programmable Chemputation of Molecules and Materials. Robot work: Self-driving lab campaign. Reported result: Several syntheses executed on new modules with yields reported.
clio-electrolytes	SD	P + D	not run	none	Autonomous optimization of non-aqueous Li-ion battery electrolytes via robotic experimentation (Clio). Robot work: Self-driving lab campaign. Reported result: Found electrolyte with 13% higher conductivity in 42 experiments; pouch-cell validation.
helao-async	SD	P + D	not run	MIT	HELAO-async: autonomous electrochemistry/high-throughput experimentation orchestration (Caltech/TRI). Robot work: Self-driving lab campaign. Reported result: Used for autonomous electrocatalyst/OER composition screening with reported datasets.
organa	SD	P + D	not run	none	ORGANA: a robotic assistant for automated chemistry experimentation and characterization. Robot work: Self-driving lab campaign. Reported result: Solubility, pH, recrystallisation and electrochemistry (quinone) experiments executed with reported user-study times and measured values.
polybot-pedot	SD	P + D	not run	MIT	Autonomous platform for solution processing of electronic polymers (Polybot). Robot work: Self-driving lab campaign. Reported result: Autonomously optimised PEDOT:PSS film conductivity/defects over closed-loop campaigns; conductivities reported.
raft-polymerization-platform	SD	P + D	not run	check Zenodo	A fully automated platform for photoinitiated RAFT polymerization. Robot work: Self-driving lab campaign. Reported result: Automated photo-RAFT polymerisations with Mn/Đ characterisation reported.
robochem	SD	P + D	not run	MIT	Automated self-optimization, intensification, and scale-up of photocatalysis in flow (RoboChem). Robot work: Self-driving lab campaign. Reported result: 19 photocatalytic transformations self-optimised; yields often exceeded literature.
solubility-hein	SD	P + D	not run	unverified	Automated solubility screening platform using computer vision. Robot work: Self-driving lab campaign. Reported result: Solubility of compounds determined by CV turbidity with comparison to HPLC.

[Tyronita]

All new expansion candidates (continued)

No (58)

Candidate	Level	Spec	Simulator	Licence	Description
bio-spec-hardwarex	BR	D (needs a written oracle)	not run	CC BY (hardware, per…	BIO-SPEC: An open-source bench-top parallel bioreactor system. Robot work: Closed-loop culture control. Reported result: Validation growth experiments reported in HardwareX.
takahashi-turbidostat	BR	D (needs a written oracle)	not run	unverified	A low cost, customizable turbidostat for use in synthetic circuit characterization. Robot work: Closed-loop culture control. Reported result: Circuit characterisation in turbidostats with growth/fluorescence data.
archemist	OR	D (needs a written oracle)	not run	unverified	ARChemist: Autonomous Robotic Chemistry System Architecture (Fakhruldeen et al.). Robot work: Workcell orchestration / scheduling. Reported result: Solubility screening and crystallisation workflows demonstrated.
mobile-robot-burger-2020	OR	D (needs a written oracle)	not run	n/a	A mobile robotic chemist (Burger et al.). Robot work: Workcell orchestration / scheduling. Reported result: 688 experiments over 8 days found photocatalyst mixtures 6x more active.
adept-framework	OT	D (needs a written oracle)	not run	n/a	The ADePT framework for assessing autonomous laboratory robotics. Robot work: Other automation. Reported result: Assessment framework; useful for rubric design, not tasks.
autopots-proteome	OT	D (needs a written oracle)	not run	n/a (no repo)	Fully Automated Sample Processing and Analysis Workflow for Low-Input Proteome Profiling. Robot work: Other automation. Reported result: autoPOTS quantified proteomes from low-input/single-cell-scale samples using an unmodified low-cost pipetting robot.
autorelacs	OT	D (needs a written oracle)	not run	MIT	AutoRELACS: automated generation and analysis of ultra-parallel ChIP-seq. Robot work: Other automation. Reported result: AutoRELACS generated ChIP-seq data comparable to the manual RELACS protocol across many samples.
baybe	OT	P + D	not run	Apache-2.0	BayBE: Bayesian Back End for design of experiments. Robot work: Other automation. Reported result: Campaign API with benchmarks; industrial case studies reported, few with released wet-lab data.
bioclonebot	OT	D (needs a written oracle)	not run	see OSF (HardwareX requires…	BioCloneBot: A versatile, low-cost, and open-source automated liquid handler. Robot work: Other automation. Reported result: Validated accuracy and repeatability across liquid handling tasks.
catib-screening	OT	D (needs a written oracle)	not run	GPL-3.0	"High-throughput screening of catalytically active inclusion bodies using laboratory automation and Bayesian optimization". Robot work: Other automation. Reported result: Generated 63 BsGDH CatIB variants; automated screening workflow validated and used to rank variant activities in parallel.
diy-ot-linear-motion-qpcr	OT	D (needs a written oracle)	not run	n/a (no repo)	Principles of computer-controlled linear motion applied to an open-source affordable liquid handler for automated micropipetting. Robot work: Other automation. Reported result: Benchmarking showed micropipetting speed/positional accuracy adequate for qPCR setup reproducibility.
diy-stem-lhr	OT	D (needs a written oracle)	not run	n/a (no repo)	DIY liquid handling robots for integrated STEM education and life science research. Robot work: Other automation. Reported result: Qualitative teacher study plus demonstration experiments.
egf-sequeduct	OT	D (needs a written oracle)	not run	GPL-3.0	Biofoundry-Scale DNA Assembly Validation Using Cost-Effective High-Throughput Long-Read Sequencing. Robot work: Other automation. Reported result: Biofoundry-scale validation of assembled plasmids via nanopore long-read sequencing with automated reports.
evobot	OT	D (needs a written oracle)	not run	n/a (no repo)	New Capabilities of EvoBot: A Modular, Open-Source Liquid-Handling Robot. Robot work: Other automation. Reported result: New capabilities demonstrated for long-running artificial-life experiments.
findus	OT	D (needs a written oracle)	not run	GPL-3.0	FINDUS: An Open-Source 3D Printable Liquid-Handling Workstation for Laboratory Automation in Life Sciences. Robot work: Other automation. Reported result: Pipetting errors <0.3%; showcase biology and chemistry applications.
flyspresso-embryo-ihc	OT	D (needs a written oracle)	not run	not checked	An open-source semi-automated robotics pipeline for embryo immunohistochemistry. Robot work: Other automation. Reported result: Automated embryo fixation, colorimetric/antibody staining and adaptive confocal imaging demonstrated.
illumina-dnaprep-flowbot	OT	D (needs a written oracle)	not run	n/a (no repo)	Automating the Illumina DNA library preparation kit for whole genome sequencing applications on the flowbot ONE liquid handler robot. Robot work: Other automation. Reported result: Automated libraries had comparable yields (median 1.5-fold diff), similar assembly quality and 100% cgMLST concordance; hands-on time 25 vs 125 min.
labdroid-phenol-chloroform	OT	D (needs a written oracle)	not run	CC-BY-4.0	Automated phenol-chloroform extraction of high molecular weight genomic DNA for use in long-read single-molecule sequencing. Robot work: Other automation. Reported result: Routine extraction of >100 kb gDNA from algae, bacteria, insects and human cells; validated by pulse-field electrophoresis and nanopore read N50.
lego-lhr-stem	OT	D (needs a written oracle)	not run	n/a (no repo)	Liquid-handling Lego robots and experiments for STEM education and research. Robot work: Other automation. Reported result: Activities tested with school students; build and experiment instructions provided.
maple-fly-robot	OT	D (needs a written oracle)	not run	not checked	MAPLE (modular automated platform for large-scale experiments), a robot for integrated organism-handling and phenotyping. Robot work: Other automation. Reported result: Automated virgin collection and social-behaviour experiments revealing persistent idiosyncrasies.
miniaturized-16s-mosquito	OT	D (needs a written oracle)	not run	protocols.io (CC-BY)	High-Throughput Miniaturized 16S rRNA Amplicon Library Preparation Reduces Costs while Preserving Microbiome Integrity. Robot work: Other automation. Reported result: Miniaturised 5 uL reactions matched standard 25 uL composition; validated on 46 fish x 5 body sites over seven months.
mngs-384-miniaturized	OT	D (needs a written oracle)	not run	n/a (no repo)	Miniaturization and optimization of 384-well compatible RNA sequencing library preparation. Robot work: Other automation. Reported result: ERCC spike-ins showed miniaturised prep fidelity equivalent to full-volume preps at much lower cost/time.
pipette-show	OT	P + D	not run	MIT	Pipette Show: An Open Source Web Application to Support Pipetting into Microplates. Robot work: Other automation. Reported result: Error rates for manual plate filling with vs without the app.
polymer-protein-hybrids	OT	P + D	not run	none	Machine Learning on a Robotic Platform for the Design of Polymer-Protein Hybrids. Robot work: Other automation. Reported result: Enzyme-stabilising copolymers designed and assayed.
proteoautonet	OT	P + D	not run	none	ProteoAutoNet: high-throughput co-eluted protein analysis with robotics and machine learning. Robot work: Other automation. Reported result: Co-elution networks derived from robot-processed samples validated against known complexes.
snails-slow-pipetting	OT	D (needs a written oracle)	not run	none	Consistent ultra-long DNA sequencing with automated slow pipetting. Robot work: Other automation. Reported result: Across six libraries SNAILS preserved more >100 kb reads and increased average read length vs manual slow pipetting.
summit-flow	OT	P + D	not run	MIT	Summit: benchmarking machine learning methods for reaction optimisation. Robot work: Other automation. Reported result: Emulators trained on literature flow-chemistry data; no new wet-lab experiments.
tox-pipetting-robot	OT	D (needs a written oracle)	not run	n/a (no repo)	A versatile and low-cost open source pipetting robot for automation of toxicological and ecotoxicological bioassays. Robot work: Other automation. Reported result: Tested precision of automated pipetting for online toxicity bioassays.
transfection-reagent-agnostic	OT	D (needs a written oracle)	not run	n/a (no repo)	A Novel, Fully Automated, and Reagent-Agnostic Transient Transfection Protocol. Robot work: Other automation. Reported result: HeLa transfection efficiency 40.9-73.3% and A549 42.2-54%, comparable to manual.
endylab-opentrons	S1	P + D	not run	none	Endy Lab Opentrons protocol collection. Robot work: PCR setup, dilution series, calibration. Reported result: No linked publication.
kedar-balance-integration	S1	P + D	not run	none	Opentrons OT-2 mass-balance integration (application note supplement). Robot work: gravimetric calibration with external balance. Reported result: Gravimetric accuracy of dispensed volumes reported in the application note.
cellfree-biosensor-ot2	S2	D (needs a written oracle)	not run	n/a (no repo)	Semiautomated Production of Cell-Free Biosensors. Robot work: cell-free biosensor reaction assembly (LacZ colorimetric / fluorescent). Reported result: Compared manual vs semi-automated OT-2 assembly of cell-free reactions; robotic assembly reduced variability in colorimetric and fluorescent outputs.
combinatorial-barcode-dna	S2	D (needs a written oracle)	not run	CC-BY-4.0	Scalable Combinatorial Assembly of Synthetic DNA for Tracking Applications. Robot work: oligo mixing for combinatorial DNA barcode library assembly. Reported result: Built 256- and 512-member modular barcode libraries from <50 oligos; OT-2 used for oligo mixing in Gen_2 library.
covid-fosh-review	S2	D (needs a written oracle)	not run	unknown	High-throughput and automated screening for COVID-19. Robot work: review of open-source automated COVID-19 testing. Reported result: Review of case studies (China, Spain, UK); no new wet-lab data.
curcumin-formulation-sdl	S2	D (needs a written oracle)	not run	none	Efficient discovery of new medicine formulations using a semi-self-driven robotic formulator. Robot work: drug formulation solubility screening (excipient mixing). Reported result: Found 7 curcumin formulations with solubility >10 mg/mL after sampling 256 of 7776 candidates.
duckweed-hte	S2	D (needs a written oracle)	not run	MIT	Accelerated high-throughput imaging and phenotyping system for small organisms. Robot work: modified OT-2 loading duckweed into well plates for plant-microbe growth phenotyping. Reported result: Study of 6,000 experimental units across 2,000 treatments with time-resolved growth data; system scales to 11,520 units.
gpcr-livecell-ot2	S2	D (needs a written oracle)	not run	n/a (no repo)	GPCR Signaling Measurement and Drug Profiling with an Automated Live-Cell Microscopy System. Robot work: automated compound addition during live-cell cAMP imaging. Reported result: Measured GPR68 cAMP dynamics vs extracellular pH and ogerin modulation using OT-2-integrated microscopy.
hydra-hydrogel-plates	S2	D (needs a written oracle)	not run	none	Fabrication of cell culture hydrogels by robotic liquid handling automation for high-throughput drug testing. Robot work: hydrogel coating of 96/384-well plates for cell culture HTS. Reported result: HYDRA produced meniscus-free hydrogel coatings validated by imaging-based anticancer dose-response assays and long-term microscopy.
marchantia-agro-transformation	S2	D (needs a written oracle)	not run	n/a (no repo)	Semi-automated workflow for high-throughput Agrobacterium-mediated plant transformation. Robot work: Agrobacterium-mediated plant (Marchantia) transformation. Reported result: Semi-automated protocol showed transformation efficiency similar to manual manipulation.
mxene-aerogel-al	S2	D (needs a written oracle)	not run	none (data Zenodo…	Machine intelligence accelerated design of conductive MXene aerogels with programmable properties. Robot work: materials mixture preparation (MXene/CNF/gelatin) for active learning. Reported result: OT-2 prepared 264 mixtures; 162 conductive aerogels fabricated/characterised over 8 active-learning cycles to train an ANN for inverse design.
newcastle-neg-opentrons	S2	P + D	not run	none	Newcastle Network Ecology Group OT-2 protocol and labware collection. Robot work: mixed lab protocols and custom labware definitions. Reported result: No linked publication found.
nfkb-luciferase-ot2	S2	D (needs a written oracle)	not run	n/a (no repo)	Luciferase reporter assay for NF-kB activation automated by an open-source liquid handling platform. Robot work: cell seeding, transfection and luciferase reporter assay. Reported result: OT-2 Python protocols for HeLa seeding, NF-kB reporter transfection and luciferase readout gave quantifiable activation signals with good sample-handling quality.
opentrons-physicalai-hf	S2	P + D	not run	unstated	Opentrons/PhysicalAI (Hugging Face dataset of Flex teleoperation episodes). Robot work: Flex plate-handoff manipulation episodes (imitation learning data). Reported result: Inference session summaries record on-robot execution of plate handoffs to a BioTek reader.
ot2eye	S2	D (needs a written oracle)	not run	MIT	OT2Eye: Detection of labware conditions to monitor and extend affordable liquid-handling robots. Robot work: labware/tip detection from OT-2 camera images. Reported result: High accuracy identifying plates and tip racks and their deck positions, and tip presence/absence.
plastic-substitutes-al	S2	D (needs a written oracle)	not run	BSD	Machine intelligence-accelerated discovery of all-natural plastic substitutes. Robot work: nanocomposite film mixture preparation for active learning. Reported result: OT-2 prepared 286 CNF/MMT/gelatin/glycerol nanocomposites to train models discovering all-natural plastic substitutes.
pr-pr	S2	D (needs a written oracle)	not run	n/a (no repo)	PR-PR: cross-platform laboratory automation system. Robot work: cross-platform protocol language (Golden Gate, Kunkel mutagenesis, Gibson). Reported result: Modified Golden Gate assembly run across liquid-handling robot, microfluidic and manual platforms; Kunkel and Gibson also implemented.
radiolabeling-psma-ot2	S2	D (needs a written oracle)	not run	n/a (no repo)	Automated radiolabeling and handling of 177 Lu- and 225 Ac-PSMA-617 using a robotic pipettor. Robot work: radiochemistry (177Lu/225Ac-PSMA-617 labelling). Reported result: Robotic pipettor performed radiometal labelling with reduced operator exposure and synthesis times (radiochemical yields/purity reported).
synthetic-cells-robotics	S2	D (needs a written oracle)	not run	unknown (repo not found)	Scaling Up Synthetic Cell Production Using Robotics and Machine Learning Toward Therapeutic Applications. Robot work: synthetic cell production scale-up. Reported result: Robotic scale-up increased synthetic-cell batch size 30-fold; luciferase SCs monitored in mice.
xchem-opentrons	S2	P + D	not run	Apache-2.0	XChem Opentrons reaction liquid-handling toolkit. Robot work: reaction dispensing arrays for crystallographic screening. Reported result: No single paper ties these scripts to reported results.
swabseq-flex	S3	D (needs a written oracle)	not run	n/a (no repo)	Automated implementation of the SwabSeq COVID-19 diagnostic assay on the opentrons flex liquid-handling robot. Robot work: SwabSeq NGS COVID-19 diagnostic (RT-PCR barcoding, pooling, SPRI cleanup). Reported result: Flex workflow completed in ~2.5 h with complete agreement on negatives, very high agreement on positives and no discordant calls vs manual.
bead-digestion-ann	S4	D (needs a written oracle)	not run	n/a (no repo)	Refinement of paramagnetic bead-based digestion protocol for automatic sample preparation using an artificial neural network. Robot work: proteomics paramagnetic bead digestion. Reported result: ANN-optimised OT-2 bead digestion showed high reproducibility across plasma, Arabidopsis, E. coli and mouse cortex; identified 385 unique rat bile proteins.
kitfree-rna-extraction	S4	D (needs a written oracle)	not run	n/a (no repo)	A high-throughput pipeline for scalable kit-free RNA extraction. Robot work: kit-free magnetic-bead RNA extraction for SARS-CoV-2 RT-qPCR. Reported result: Semi-automated kit-free extraction validated by SARS-CoV-2 RT-qPCR without loss of diagnostic sensitivity/accuracy.
protein-corona-ot2	S4	D (needs a written oracle)	not run	n/a (no repo)	Automation and low-cost proteomics for characterization of the protein corona: experimental methods for big data. Robot work: protein corona formation and magnetic isolation for proteomics. Reported result: Automated corona isolation with magnetic-core nanoparticles; compared low-cost proteomics workflows.
sp3-magbead-protocol-repo	S4	P + D	not run	MIT	Automated magnetic bead-based digestion protocols for Opentrons OT-2 (proteomics CNB-CBM). Robot work: SP3 bead digestion. Reported result: None published.
ai-chemist-ustc	SD	D (needs a written oracle)	not run	n/a	An all-round AI-Chemist with a scientific mind. Robot work: Self-driving lab campaign. Reported result: OER catalysts discovered autonomously.
aurora-materials	SD	P + D	not run	vendor	AURORA - An Automatic Robotic Platform for Materials Discovery. Robot work: Self-driving lab campaign. Reported result: Electrochemical materials screening reported.
cloudlab-chiral-perovskite	SD	D (needs a written oracle)	not run	n/a	Autonomous discovery of optically active chiral inorganic perovskite nanocrystals through an intelligent cloud lab. Robot work: Self-driving lab campaign. Reported result: Chiral perovskite NCs with CD responses discovered.
sdl-nano-acsnano	SD	D (needs a written oracle)	not run	n/a	A Self-Driving Lab for Nano- and Advanced Materials Synthesis. Robot work: Self-driving lab campaign. Reported result: Perspective/platform review.

[Tyronita]

Which candidate protocols actually run (Docker simulation, 2026-10-13)

Every Opentrons protocol in the experimental candidates was simulated in the t2w-sim image (docker/sim/, Opentrons 7.5 + PyLabRobot, no network, read-only code). 795 protocol files in 69 repos: 203 pass, from 32 repos. Per-repo counts: research/expansion/DOCKER_SIM.md.

Tier 1: paper + code + results, simulating (10 repos, the strongest new tasks)

Candidate	Passing / total	Experiment	Source	Licence
TidyTron	35/37	Washing and reusing tips and plates to cut plastic waste	SLAS Technology 2024	Apache-2.0
GFP codon burden	21/42	Growth and expression burden assay (inducer titration, GFP and OD)	J Biol Eng 2025	MIT
ANT cloning	7/10	Cloning by natural transformation, PCR to plating	ACS Synth Biol 2020	none
Beckham enzyme pipeline	6/21	Enzyme discovery: transformation, expression, lysis, activity assay	Sci Rep 2024	none
MassSpecPreppy	3/10	Proteomics sample prep: protein assay, SP3 digestion, Evotip loading	Proteomics 2023	GPL-3.0
Culturomics	2/11	Bacterial isolation: DNA extraction, two-step PCR, bead clean-up	iMetaOmics 2025	none
OT-2 autoclone	2/7	Golden Gate assembly and strain construction	Appl Microbiol Biotechnol 2022	AGPL-3.0
MSD biochem assays	2/2	Bradford and PicoGreen quantitation	SLAS Technology 2024	Apache-2.0
HyRAD	1/1	Ancient-DNA hybridisation capture (magnet, thermocycler)	Mol Ecol Resour 2022	none
ExFAB QIAseq (Flex)	1/1	NGS library prep on the Opentrons Flex	protocols.io 2025	GPL-3.0

Tier 2: protocols run, weaker paper/results link (17 repos)

Area	Repos (passing / total)
Sequencing library prep	KAUST NGS core (13/22), EvoGenOT (13/25), NIOZ amplicon database (11/202), ARTIC LoCost nanopore (6/6), ESR ARTIC normalisation (4/11), Bisanz 16S (2/5), UF Dutton 16S on Flex (2/17)
COVID-19 diagnostics	Opentrons COVID-19 stations (12/52), DETECTR (2/2), TaqPath RT-qPCR (1/2)
Synthetic biology	PUDU (5/28), DTU teaching protocols (6/10), Imperial colony-picker prep (2/2), E. coli expression and purification (6/21)
Other	MIT PV-Lab solubility and drop-casting (15/20), trypanosome drug combinations (2/3), C. elegans food preference (2/2), OpenBlend media (1/2), PRISM LLM-generated protocols (1/1)

Tier 3: run but not tasks (3 repos)

BiomationScripter, the Newcastle labware collection and LabOP are libraries or templates with no real liquid moves.

Hamilton

The har1eyk STARlet Automated Protein Extraction protocol (PyLabRobot) runs end to end on PyLabRobot's simulated backend: 936 liquid moves, 164 tip pickups and 10 CoRe-gripper moves of the binding plate onto the magnet and back.

Reference videos

Four 3D replays are attached to a draft release (outside the codebase; drafts are visible only to repository collaborators):

1. 1_hamilton_starlet_protein_extraction.mp4: the STARlet run on PyLabRobot's 3D STARlet model
2. 2_tidytron_tip_and_plate_wash.mp4
3. 3_gfp_codon_burden_assay.mp4
4. 4_ant_cloning_overlap_extension_pcr.mp4

Limits: "passes" means the protocol runs in the simulator, which does not track liquid or check the biology. The two Flex candidates and Culturomics simulate but were not rendered, because the replay only reads OT-2 slot numbers.