id,row_type,year,name,system,cells,neurons_low,neurons_high,neurons_basis,neurons_where,electrodes,task,task_class,claimed,shown,controls,replicates,n_units,peer_reviewed,preprint_only,data_open,independent_replication,lab,lab_type,country,funding,conflicts,text_read,c_peer,c_control,c_reps,c_data,c_stat,evidence_grade,fly_brains_equal,nearest_animal_below,nearest_animal_above,source_ids
shahaf2001,experiment,2001,Shahaf and Marom: learning in cortical networks,cultured neurons,Cortical neurons on a multielectrode array (species not stated in the abstract),,,not in the abstract,,multielectrode array (number not in the abstract),Low-rate focal stimulation (0.3-1 Hz) repeated until a chosen response appears 50 +/- 10 ms after the stimulus; stimulus then paused 5 min,stimulus-response conditioning,"Selective learning of an arbitrary task in a network of real cortical neurons, first outside the body",Learning curves: the number of stimuli needed to reach the target response in each cycle (per the abstract),not stated in the abstract,not stated in the abstract,,Journal of Neuroscience 2001;21(22):8782-8788,no,not assessed (full text not read),no study found that repeats the same protocol in another lab,"Technion, Department of Physiology and Biophysics",academic,Israel,,,abstract,,,,,,not graded,,,,shahaf2001
demarse2001,experiment,2001,DeMarse et al.: the neurally controlled animat,cultured neurons,"Rat cortical cells (embryonic day 18), neurons and glia, on a 60-electrode array",,,"words only: ""hundreds or thousands"" of cells (figure 1 legend)",Figure 1 legend,"60 (8 x 8 grid, 200 um spacing)",Neural activity patterns moved a simulated animal around a virtual room; collisions and moves fed back as stimulation,closed-loop control,A system that responds and gives feedback to a living network in real time,One session of about 50 minutes (after 8 minutes without feedback): 51 distinct activity patterns found; the authors say they do not yet know how the feedback changed the network; no learning shown,none described,"one culture, one session described",1,Autonomous Robots 2001;11(3):305-310,no,no data statement,not applicable (no performance claim),California Institute of Technology (Potter lab),academic,USA,,,full,Y,N,N,N,N,D,,,,demarse2001
ruaro2005,experiment,2005,Ruaro et al.: image processing with neuronal cultures,cultured neurons,Neuronal cultures on commercial multielectrode arrays (cell type not in the abstract),,,not in the abstract,,commercial multielectrode arrays (number not in the abstract),Images mapped to patterns of stimulation; responses to a pattern strengthened by tetanic training,pattern recognition,Low-pass and band-pass filtering of images; learning allowed pattern recognition and feature extraction,as claimed in the abstract; numbers not in the abstract,not stated in the abstract,not stated in the abstract,,IEEE Transactions on Biomedical Engineering 2005;52(3):371-383,no,not assessed (full text not read),none found,"Scuola Internazionale Superiore di Studi Avanzati (SISSA), as listed on the paper",academic,Italy,,,abstract,,,,,,not graded,,,,ruaro2005
bakkum2008,experiment,2008,"Bakkum et al.: embodied cortical network, goal-directed task",cultured neurons,In vitro neocortical network on a multi-electrode array (species not in the abstract),,,not in the abstract,,multi-electrode array (number not in the abstract),An adaptive algorithm searched for training stimuli that moved the network to set activity states in a movement task,closed-loop control,The network learned to modulate its dynamics and reach set activity states within tens of minutes,After 2 hours of training plasticity stayed above baseline for 80 minutes (p < 0.01); replaying the effective stimuli without feedback gave no significant plasticity (p = 0.82),replay of the same stimuli not contingent on behaviour (from the abstract),not stated in the abstract,,Journal of Neural Engineering 2008;5(3):310-323,no,not assessed (full text not read),none found,Georgia Institute of Technology (Potter lab),academic,USA,,,abstract,,,,,,not graded,,,,bakkum2008
isomura2015,experiment,2015,Isomura et al.: blind source separation by cultured neurons,cultured neurons,"Rat embryonic cortical cells (dissociated), 64-electrode array",500000,500000,cells seeded per array (not a neuron count),"Methods, culture preparation: 5 x 10^5 cells seeded per array",64,Two hidden sources mixed into stimulation patterns; 100 training trials (about 14 hours) per culture,stimulus-response learning,Neurons learned to represent particular sources while filtering out others; first in vitro demonstration of blind source separation,Evoked responses became selective for one source in the 23 trained cultures (p < 10^-15 at the electrode level); 37 further cultures served as control groups,37 control-group cultures,23 trained cultures,23,PLoS Computational Biology 2015;11(12):e1004643,no,spike trains on the lab website and in the S1 dataset,repeated by the same lab in 2023; no other lab found,University of Tokyo (Jimbo lab),academic,Japan,Japan Society for the Promotion of Science grants,none declared,full,Y,Y,Y,Y,Y,A,3.59,"Fruit fly (139,255)","Algerian sand gecko (1,771,000)",isomura2015
cai2023,experiment,2023,Cai et al.: Brainoware,organoid,"Human cortical organoids from pluripotent stem cells, on a multielectrode array",,,not stated in the preprint,,"multielectrode array (4 x 4 stimulation patterns; full layout in the supplement, not read)",Predict the next value of a chaotic Henon map from pulse trains; speech recognition (published version),reservoir computing,"Living AI hardware: learning from training data, solving non-linear equations, speech recognition","Preprint v1: regression score on the Henon map rose from 0.356 (+/- 0.071) to 0.812 (+/- 0.043) after four daily training epochs; blocking plasticity (K252a) gave 0.31 to 0.385; chip without organoid scored 0; n = 5 per group. The speech result is in the published paper, which we could not read",plasticity blocker K252a; chip and medium without an organoid; untrained organoids,n = 5 organoid experiments per group,5,Nature Electronics 2023;6(12):1032-1039,no (preprint v1 read; published text not read),no data statement in the preprint,none found,Indiana University Bloomington (Guo lab) with Cincinnati Children's Hospital and others,academic,USA,"US National Institutes of Health awards DP2AI160242, R01DK133864, U01DA056242",none declared,preprint,Y,Y,Y,N,Y,B,,,,brainoware-preprint;cai2023
sumi2023,experiment,2023,Sumi et al.: neurons as reservoirs,cultured neurons,Rat cortical neurons on micropatterned modular substrates; optogenetic stimulation and calcium imaging,60,60,neurons read out per network (15 per module); culture size not extracted,"Results, first reservoir experiment","none (calcium imaging, not electrodes)","Tell which of three regions (top, middle, bottom) of a culture was lit, with a linear decoder on 60 neurons",reservoir computing,Biological neurons act as generalization filters; modular wiring improves classification,Classification accuracy 74.8 +/- 20.2% against 32.4 +/- 13.0% for shuffled labels (3 classes; mean +/- SD; n = 21 networks),label-shuffled null model,21 networks,21,PNAS 2023;120(25):e2217008120,no,Zenodo,none found,Tohoku University,academic,Japan,,none declared,full,Y,Y,Y,Y,Y,A,0.0,,C. elegans (302),sumi2023
isomura2023,experiment,2023,Isomura et al.: causal inference in cultured networks,cultured neurons,Rat embryonic cortical cells on a 64-electrode array,,,not stated in the text we extracted,,64 (32 used for stimulation),Learn two hidden sources from mixed stimulation; drugs shifted network excitability,stimulus-response learning,Quantitative predictions of a theory of neural self-organisation (Friston) confirmed in vitro,"30 control experiments: selective responses emerged; the NMDA blocker APV (9 experiments), diazepam (7) and bicuculline (6) disrupted inference; synaptic changes matched the model",vehicle controls and three drug conditions,30 independent control experiments,30,Nature Communications 2023;14:4547,no,GitHub (data and code),same lab as 2015; no other lab found,RIKEN Center for Brain Science and University of Tokyo,academic,Japan,JSPS KAKENHI; Japan Science and Technology Agency,none declared,full,Y,Y,Y,Y,Y,A,,,,isomura2023
kagan2022,experiment,2022,Kagan et al.: DishBrain (Pong),cultured neurons,"Mouse embryonic cortical cells (E15.5) and human iPSC-derived cortical cells on a 26,400-electrode array",800000,1000000,"cells plated per array (not a neuron count): about 800,000 mouse cells; about 10^6 human cells","Methods: ""Approximately 800,000 cells were plated down onto each ... HD-MEA"" (mouse); ""approximately 10^6 cells plated onto each MEA"" (human)","26,400 (1,024 read at once); Pong input on 8 electrodes","Simplified Pong: ball position as stimulation, two motor regions move the paddle; a hit gives predictable stimulation, a miss unpredictable",closed-loop game,In vitro neurons learn and exhibit sentience when embodied in a game-world,"Rally length rose between the first 5 and the last 15 minutes in mouse (9 cultures) and human (11 cultures) cultures, not in media-only controls (6), rest sessions (20) or an in-silico control (3 seeds); human cultures started worse than controls",media-only arrays; rest sessions; in-silico random control,9 mouse and 11 human cultures (399 sessions in the main test),20,Neuron 2022;110(23):3952-3969,no,"Open Science Framework (spikes, game metrics, code)",none found (the follow-ups share authors and the company),"Cortical Labs with Monash University, University College London and others",company + academic,Australia,Wellcome core funding; Australian Research Council; Cortical Labs,lead author is an employee and shareholder of Cortical Labs and holds related patents,full,Y,Y,Y,Y,Y,A,6.46,"Fruit fly (139,255)","Algerian sand gecko (1,771,000)",kagan2022
habibollahi2023,experiment,2023,Habibollahi et al.: criticality in DishBrain,cultured neurons,Mouse and human cortical cells on high-density arrays,1000000,1000000,cells plated per array (not a neuron count),Methods: approximately 10^6 cells plated on each array,high-density array (as DishBrain),Pong gameplay versus rest; feedback types compared,closed-loop game,Critical dynamics arise when networks receive structured task input; criticality alone does not give learning,"14 cultures, 308 sessions (192 gameplay, 116 rest): markers of criticality during gameplay but not rest; the no-feedback condition did not learn","rest sessions; stimulus, silent and no-feedback conditions",14 cultures,14,Nature Communications 2023;14:5287,no,Open Science Framework (data and code),none found,Cortical Labs with the University of Melbourne,company + academic,Australia,Melbourne Research Scholarship; Australian Research Council DP220101166; Royal Melbourne Hospital Neuroscience Foundation,two authors are employees of Cortical Labs; one is a shareholder with related patents,full,Y,Y,Y,Y,Y,A,7.18,"Fruit fly (139,255)","Algerian sand gecko (1,771,000)",habibollahi2023
khajehnejad2025,experiment,2025,Khajehnejad et al.: DishBrain against deep reinforcement learning,cultured neurons,Mouse (E15) and human iPSC-derived cortical cultures on high-density arrays,,,plating count not extracted,,high-density array (as DishBrain),Simplified Pong; 20 minutes of real-time-equivalent play,closed-loop game,Even simple biological cultures outperformed deep reinforcement learning when samples are limited to a real-world time course,"In the last 15 minutes mouse and human cultures reached more hits per rally than DQN, A2C and PPO (Tukey post-hoc tests); within-group improvement significant only for the two culture types and A2C",three deep RL algorithms as comparison; rest sessions,cultures from the DishBrain data set (count not extracted),,Cyborg and Bionic Systems 2025;6:0336,no (arXiv version 2024),Open Science Framework,none found (same group),"Cortical Labs with Monash University, University College London and others",company + academic,Australia,,senior author is the Cortical Labs chief scientist,full,Y,Y,Y,Y,Y,A,,,,khajehnejad2025;arxiv-2405.16946
watmuff2025,experiment,2025,Watmuff et al.: drugs change DishBrain gameplay,cultured neurons,Human iPSC-derived NGN2 neurons plus primary human astrocytes on MaxOne arrays,200000,200000,"NGN2 cells plated per chip (plus 50,000 astrocytes)","Methods: 200,000 NGN2 cells and 50,000 primary human astrocytes mixed per chip","MaxOne high-density array (26,400 electrodes)","Pong in DishBrain after three anti-seizure drugs (phenytoin, perampanel, carbamazepine)",closed-loop game,First demonstration of altered synthetic biological intelligence after drug treatment,"Carbamazepine at 200 uM improved gameplay metrics; sessions: 354 rest and 400 game controls, 42 carbamazepine, 24 phenytoin, 24 perampanel",untreated control sessions; rest versus game,recording sessions on many chips (counts per drug above),,Communications Biology 2025;8:916,no,Open Science Framework (DishBrain code not public),none found,"Cortical Labs with Monash, Cambridge and bit.bio",company + academic,Australia,,authors affiliated with Cortical Labs; competing interests declared,full,Y,Y,Y,Y,Y,A,1.44,"Fruit fly (139,255)","Algerian sand gecko (1,771,000)",smirnova2025-drug
alamel2025,experiment,2025,Alam El Din et al.: organoid building blocks of learning,organoid,"Human iPSC-derived neural organoids, up to 14 weeks, on high-density arrays",,,not stated (diameter and cell counts per organoid not extracted),,high-density array (2-5 organoids per well),Theta-burst stimulation to induce potentiation; receptor blockers; no behavioural task,plasticity assay,Organoids show the building blocks necessary for basic learning and memory,"Synapse and receptor markers, activity-gene expression, criticality, and potentiation and depression after theta-burst stimulation; wells: 11-12 per age group from 2 experiments",pharmacological conditions,"2 independent experiments, 5-6 wells each",,Communications Biology 2025;8:1237,no,RNA-seq at GEO; code on GitHub; other data on request,none found,Johns Hopkins University (CAAT) with Cortical Labs and others,company + academic,USA,,patents and consulting disclosed (organoid production; Cortical Labs),full,Y,Y,Y,Y,Y,A,,,,smirnova2025-learning
repetitive2025,experiment,2025,Shao et al.: repeated training of cultured networks,cultured neurons,Primary cortical cells (the paper says mouse E18 in the results and Wistar rat in the methods) on a 59-electrode array,80000,135000,"cells seeded per array, IQ Metrics calculation: 1,000-1,500 cells per microlitre x 80-90 microlitres",Methods: cell concentration and seeding volume,59 (8 x 8 layout),Tell stimulation patterns apart from evoked responses (support vector machine); repeat the training,pattern recognition,Repeated training increased recognition accuracy and changed network structure,"Accuracy rose with repeated rounds (n = 5-10 cultures per analysis, paired t-tests)","no untrained, blocked or empty-dish control found in the text",5-10 cultures,10,PLoS Computational Biology 2025;21(4):e1013043,no,GitHub (data files and code),none found,Tianjin University,academic,China,National Key R&D Program of China; National Natural Science Foundation,none declared,full,Y,N,Y,Y,Y,C,0.77,C. elegans (302),"Fruit fly (139,255)",repetitive2025
logic2025,experiment,2025,Kuchler et al.: neurons as NAND and OR gates,cultured neurons,Rat embryonic (E18) hippocampal neurons in microstructures on a high-density array,70000,70000,cells seeded per chip (up to 10 two-input networks per chip),"Methods, cell seeding: 70,000 cells per array","26,400 (about 280 per network)","Two input nodes, one output node: stimulate one or both inputs and read the output",logic gates,Basic Boolean computations (NAND and OR) in a living network of controlled shape,Networks gave non-linear two-input responses; rate coding and spike-timing coding compared (data from 17 networks); accuracy figures not extracted,comparison of encoding schemes; no empty-dish control extracted,17 networks,17,Frontiers in Computational Neuroscience 2025;19:1559936,no,ETH Research Collection and GitHub,none found,ETH Zurich (Laboratory of Biosensors and Bioelectronics),academic,Switzerland,,,full,Y,N,Y,Y,N,C,0.5,C. elegans (302),"Fruit fly (139,255)",logic2025
jbo2025,experiment,2025,"Deng et al.: optogenetic reservoir, obstacle avoidance",cultured neurons,Mouse embryonic (E15) cortical neurons with optogenetic stimulation and calcium imaging,15,350,neurons read out (15) and imaged in the field of view (350),Abstract and Figure 2: 350 neurons imaged; 15 neurons enough for control,none (optical),A virtual car avoids obstacles; a linear read-out trained with FORCE learning,reservoir computing,A minimal output of 15 neurons gives stable control with a success rate above 95%,More than 95% obstacle-avoidance success after about 150 s of training; optogenetic stimulation improved success,with and without optogenetic stimulation,number of cultures not extracted,,Journal of Biomedical Optics 2025;30(10):105004,no,on reasonable request,none found,Beijing University of Posts and Telecommunications with Peking Union Medical College,academic,China,,none declared,full,Y,Y,N,N,Y,C,0.0,,C. elegans (302),jbo2025
cellrep2026,experiment,2026,Robbins et al.: goal-directed learning in cortical organoids (cartpole),organoid,Mouse embryonic-stem-cell cortical organoids (plated at day 25) on MaxOne arrays,,,not stated (organoid cell counts not given),,"MaxOne high-density array, 26,400 electrodes","Balance a virtual pole on a cart; training pulses chosen by reinforcement learning, at random, or not at all",closed-loop control,First goal-directed learning in brain organoids through feedback-driven plasticity,"16 organoids, 38 experiments, over 125 hours: adaptive training beat random (p = 1.06e-3) and null (p = 2.45e-8) in the preprint; gains did not persist after a 45-minute rest; blocking AMPA and NMDA receptors removed them",no-training and random-training conditions; glutamate-receptor blockade,"16 organoids, 38 experiments",16,Cell Reports 2026;45(2):116984,no (preprint 2024 read for numbers),analysis code on GitHub (braingeneers),none found,University of California Santa Cruz (Braingeneers) with UC Santa Barbara,academic,USA,US National Science Foundation; Qualcomm; University of California,,preprint,Y,Y,Y,Y,Y,A,,,,cellrep2026;cellrep-preprint
braille2026,experiment,2026,Liu et al.: Braille letters read by organoids,organoid,Human forebrain organoids (about 500 um across) on the FinalSpark Neuroplatform,,,not stated,,8 per organoid,"Classify 26 Braille letters from a touch sensor; sensor events encoded as stimulation, responses read by a support vector machine (open loop, no learning inside the organoid)",pattern recognition,Organoids can serve as bio-hybrid computing elements for tactile classification,"Average accuracy 61% with one organoid and 83% with a three-organoid ensemble (chance with 26 letters is 3.8%, IQ Metrics calculation); the ensemble was more robust to added noise","no empty-dish, blocked or shuffled control found in the text",1 organoid; ensemble of 3,3,Frontiers in Artificial Intelligence 2026;9:1800219,no,University of Bristol data repository,none found,University of Bristol (remote use of the FinalSpark platform),academic,United Kingdom,,,full,Y,N,Y,Y,Y,C,,,,braille2026
pnas2026,experiment,2026,Sono et al.: online learning in modular neuron networks,cultured neurons,Rat primary cortical neurons in 128 microfluidic wells on a high-density array,1869,1869,IQ Metrics calculation: 128 wells x 14.6 neurons per well (mean of 30 wells),"Results, first section on the engineered networks","26,400",Closed-loop reservoir: train a linear decoder so the culture outputs periodic and chaotic signals on its own,reservoir computing,Modular networks can be trained online to generate temporal signals; modularity is essential,"Error rose after learning ended in 99% of trials (126 recordings from 22 chips); homogeneous (24), lattice (27) and hierarchical (24) networks compared",homogeneous (non-modular) networks; learning versus post-learning phase,"22 chips, 126 recordings",22,PNAS 2026;123(11):e2521560123,no,Zenodo,none found,Tohoku University,academic,Japan,,none declared,full,Y,Y,Y,Y,Y,A,0.01,C. elegans (302),"Fruit fly (139,255)",pnas2026
paired2025,experiment,2025,"Rountree et al.: paired organoids, potentiation and a maze game",organoid,Pairs of human iPSC brain organoids (about 500 um) joined by an axon channel with ten electrodes,,,not stated,,10 planar electrodes in the channel,"Open-loop potentiation protocols; closed-loop ""maze game"" with reinforcement-learning rules",closed-loop game,"Robust, drug-tunable short- and long-term potentiation; adaptive responses in a maze game that depend on BDNF",about 384 organoid pairs were set up; trained versus untrained networks; NMDA-receptor blocker prevented potentiation (preprint),untrained networks; blocker; BDNF,not extracted,,preprint,yes,no data statement found,none found,Tulane University,academic,USA,,,preprint,N,Y,N,N,N,D,,,,pp-paired2025
digits2026,experiment,2026,Loeffler et al.: handwritten digits on the CL1,mixed,"Human iPSC-derived cortical and hippocampal neurons as monolayers, organoids and microfluidic modules, on CL1 chips",500000,500000,total cells seeded per chip (monolayers; not a neuron count),Methods 5.2: 0.5 x 10^6 total cells per array chip,64 per chip (CL1 system),Spatio-temporal handwritten-digit (MNIST) classification with a linear read-out; chance 10%,reservoir computing,Modular cortical cultures outperform monolayers and organoids; rigorous artifact control is needed,"Modular cortical cultures scored highest; an empty saline dish scored above chance when stimulation artifacts were not blanked, so some apparent accuracy came from the stimulus itself",empty saline dish; shuffled output; synthetic spike trains; chance,"for example 22 modular cortical, 8 hippocampal monolayer chips",,preprint,yes,no data statement found,none found,Cortical Labs with Tohoku University and the University of Melbourne,company + academic,Australia,,,preprint,N,Y,Y,N,Y,C,3.59,"Fruit fly (139,255)","Algerian sand gecko (1,771,000)",pp-digits2026
robot2026,experiment,2026,"Cai et al.: Brainobot, an organoid-controlled humanoid robot",organoid,Human cortical organoids (3 months old) on a multielectrode array,,,not stated,,multielectrode array (6 stimulation electrodes in the grasping task),A camera image becomes pulses; organoid responses choose robot actions for object grasping and laser chasing,closed-loop control,"Cross-task adaptivity, high computing efficiency and low energy consumption",Accuracy rose from 72% to 92% over three laser-chasing epochs; after a switch to grasping 86% then 92%; plasticity blocker K252a held it at 62-64%; the toxin TTX removed it; the authors state the supporting equipment is bulky and energy-intensive,TTX; K252a; fully connected networks as comparison,n = 5 organoids from three independent experiments,5,preprint,yes,"""main text or supplementary""",none found,Indiana University Bloomington with Harvard Medical School and others,academic,USA,,,preprint,N,Y,Y,N,Y,C,,,,pp-robot2026
infoproc2026,experiment,2026,"Cai et al.: Brainopheno, organoid pattern classification",organoid,"Human cortical organoids on a multielectrode array (control, APP-mutant and Alzheimer-monocyte conditions)",,,not stated,,multielectrode array,Classify evoked responses to different spatial stimulation patterns,pattern recognition,"A quantifiable information-processing-like phenotype that appears with maturation, is disrupted by drugs and is reduced in Alzheimer models",classification above chance that tracks maturation and pharmacology (preprint; accuracy numbers not extracted),pharmacological perturbation,not extracted,,preprint,yes,RNA-seq at GEO; electrophysiology on request,none found,"Indiana University Bloomington with UCLA, Weizmann Institute and others",academic,USA,,,preprint,N,Y,N,N,N,D,,,,pp-infoproc2026
cl1,platform,2025,Cortical Labs CL1 and Cortical Cloud,cultured neurons,Human stem-cell-derived neurons on a silicon chip with life support (company page); chips of 64 electrodes in the 2026 papers,500000,500000,"cells seeded per 64-electrode chip in two 2026 Cortical Labs preprints that used a CL1 system (not a neuron count, not a company specification)","cl1.html gives no count; preprints: 5 x 10^5 and 0.5 x 10^6 cells per chip. Press figures of ~800,000 and ~200,000 are not on the company pages",64 per chip (2026 preprints),Platform: deploy code to neurons locally or on the Cortical Cloud,platform,"World's first code-deployable biological computer; neurons alive up to 6 months; ""a fraction of the energy""","The company page gives no price, no neuron count and no energy figure. The CL1 page was archived on 4 March 2025 (Wayback Machine); press prices ($35,000 a unit, $300 a week in the cloud) are not on the company pages",not applicable,not applicable,,platform paper: CL API (arXiv preprint 2026),not applicable,API documentation on GitHub,not applicable,"Cortical Labs Pte Ltd, Melbourne",company,Australia,,,company pages and papers,,,,,,not applicable,3.59,"Fruit fly (139,255)","Algerian sand gecko (1,771,000)",cl-cl1;cl-cloud;cl-home;arxiv-2602.11632;pp-digits2026;biorxiv-2026.01.21.700748
neuroplatform,platform,2024,FinalSpark Neuroplatform,organoid,"Human forebrain organoids about 500 um across on 4 arrays, 4 organoids each",,,no neuron count in the platform paper or on the company pages,Jordan et al. 2024: each array holds 4 organoids with 8 electrodes each; organoid diameter about 500 um,"8 per organoid; 32 per array; 128 on the platform (4 arrays), IQ Metrics calculation",Platform: remote 24/7 stimulation and recording of organoids; demonstrations of stimulation and dopamine uncaging,platform,"Remote access to human neurons for biocomputing; ""up to a million times"" less energy (press; the company's 2023 article says a billion)","Used with over 1,000 organoids in three years and more than 18 TB of recordings (platform paper); plans ""Shared"" and ""Dedicated"" listed, price not shown on the page on 3 November 2026",not applicable,not applicable,,Frontiers in Artificial Intelligence 2024;7:1376042,no,raw data on request (paper); 30 TB offered to researchers (company page),not applicable,"FinalSpark, Vevey",company,Switzerland,none received for the paper (stated),all authors are FinalSpark employees,full + company pages,Y,,,,,not applicable,,,,jordan2024;fs-neuroplatform;fs-vision;fs-home
jhu-oi,programme,2023,Johns Hopkins organoid intelligence programme (roadmap),organoid,Human brain organoids (target),10000000,10000000,"the roadmap's target of about 10 million neural cells per organoid; today's organoids are under 500 um and under 100,000 cells (same paper)","Smirnova et al. 2023, sections on scaling and ethics",not applicable,Roadmap for organoid-based computing and its ethics,programme,"Biocomputing could be faster, more efficient and more powerful than silicon and need only a fraction of the energy","A roadmap and review, not an experiment: sets the 10-million-cell target, lists the 21 MW (supercomputer) against 10-20 W (brain) comparison, and says terms such as intelligence and sentience describe basic functions only",not applicable,not applicable,,Frontiers in Science 2023;1:1017235,no,not applicable,not applicable,Johns Hopkins University with co-authors at other institutions,academic,USA,Johns Hopkins Discovery and SURPASS awards,patent and consulting interests disclosed (organoid production; Cortical Labs),full,,,,,,not applicable,,,,smirnova2023-frontiers