New guide: what is reservoir computing?
Brainoware and most of the organoid computing work I've written about lean on reservoir computing, so I've given it its own page. You'll find where it came from (liquid state machines in 2002 and echo state networks), why only the readout gets trained, and why I think drifting organoids are its weak point.
New guide: how many neurons are in the human brain?
You'll see 86 billion neurons quoted everywhere, including on this site, so I've written up where it comes from. It's a 2009 count of four brains, and a 2024 paper in Brain argues the data only support somewhere between 61 and 99 billion. I re-did the intervals myself, and you'll also find where in the brain the neurons actually are.
New comparison: FinalSpark vs Cortical Labs
If you've wondered how the two living-neuron computers differ, I've put them side by side. FinalSpark runs a remote lab of brain organoids that research groups use from Python, and Cortical Labs sells the CL1 (about 800,000 neurons on a 64-electrode chip) or rents it through its cloud. You'll find access, price and lifetime in one table, and which I'd pick for what.
New guide: are brain organoids conscious?
If you've wondered whether the organoids in these computers can feel anything, I've written up what the evidence says. The short answer is no, as far as anyone can tell (a 2021 National Academies report called it extremely unlikely for the foreseeable future). You'll also find why EEG-like waves in organoids don't change that, and what I think is still open.
New guide: Brainoware and organoid reservoir computing
If you've seen headlines about brain cells doing speech recognition, they were probably about Brainoware (an Indiana University system from 2023), so I've written it up. They used a human brain organoid as the reservoir in a reservoir computer, and after two days it could tell you which of 8 Japanese men was saying a vowel 78% of the time. I'd still call it a proof of concept, because ordinary AI did better, and you'll find the reasons in the guide.
Read Brainoware: a brain organoid used for reservoir computing
New comparison: neuromorphic computing vs wetware
I've written a comparison of neuromorphic chips and wetware, because I keep seeing the two mixed up. IBM's TrueNorth put 1 million programmable silicon neurons on one chip back in 2014 and ran live video on 63 milliwatts, whereas you'd get about 800,000 living ones in a CL1. I think the chips are well ahead for now, and you'll find the reasons in the guide.
New guide: what brain organoids are
I've added a plain guide to what brain organoids are, since the computing pages assume you know. I'd start with Lancaster's 2013 paper, where organoids grown from stem cells reached up to 4 millimetres across and modelled microcephaly. You'll also find why I wouldn't call them mini-brains.
New guide: multielectrode arrays
I've written up the hardware under all of this, the multielectrode array. One of the first, for cultured cells, was described in 1972, Jerome Pine's 1980 dish had 32 electrodes that could record and stimulate, and the chip used in DishBrain had 26,400. I'd say how many you can read at once matters more than the headline count, and the guide explains why.
New guide: what biocomputing covers
I've written a guide on biocomputing, because you'll see the word used for three quite different things. I'd start with Leonard Adleman, who solved a small graph problem with DNA in 1994, and then a 2017 test that got data back perfectly from DNA at 215 petabytes per gram. The third branch is living neurons (what the rest of this site covers), and I've linked every result to its paper.
New guide: how long brain organoids live
I've written up how long brain organoids actually survive, because the answer depends a lot on where they are. In a lab dish a Stanford and UCLA team kept cortical organoids for up to 694 days, and they reached postnatal-like stages at 250 to 300 days. Wired into a computer it's far shorter, with about 100 days on FinalSpark's Neuroplatform and six months quoted for the CL1. Also, the guide explains why they die, which comes down to having no blood vessels.
New guide: the Cortical Labs CL1
Cortical Labs sells a biological computer called the CL1, and I've written up what is inside it. A culture holds about 800,000 human neurons on a chip with 64 electrodes, and the company says its life support keeps them alive for up to six months. The best result I could find from outside the company is a July 2026 paper where one culture sorted images into two classes with up to 95.6% accuracy. Also, press reports put the price at about $35,000, and you can rent time on one through Cortical Cloud.
Three guides rewritten, same figures
I went back over the wetware explainer, the organoid intelligence guide and the sniper bots case study and rewrote the prose so it reads the way I'd actually say it. Nothing factual moved: the two sniper bots still bought 36 seconds after the pool opened and paid about 21%, and the best organoids on FinalSpark's platform still last up to about 100 days. If you'd read them before, the case study now says a bit more about why I think Bot A's transaction was signed in advance.
Brain vs GPU power and the sniper bots case study, rewritten
I've rewritten the text of two guides and you'll find every figure is the same. Your brain still runs on about 20 watts and one NVIDIA H100 SXM is rated at up to 700. Each on-chain number in the case study still links to its transaction (so you can check it yourself). Also, each page now shows the day it last changed. The sitemap carries the same dates so search engines can tell what's new.
New guide: the FinalSpark Neuroplatform
FinalSpark runs a lab in Vevey in Switzerland. Human brain organoids sit on electrode arrays there and you can stimulate and record them from a Python notebook wherever you are. I went through the 2024 paper by FinalSpark's own staff (it's in Frontiers in Artificial Intelligence). It says they've tested more than 1,000 organoids and recorded over 18 terabytes. The best organoids now last up to 100 days. However, two parts of the paper don't quite agree on some of the numbers, and I may be wrong about which one is right, so I've given both on the page and said where they differ.
New guide: how brain cells in a dish learned to play Pong
DishBrain is the 2022 experiment by Kagan's team, published in Neuron. Neurons grown on a chip moved the paddle in a simple game of Pong. I used the paper's full text for the numbers (a MaxOne chip with 26,400 electrodes and about 800,000 mouse cells on each chip). The team compared the first 5 minutes of play with minutes 6 to 20, and as far as I can tell the cells only improved when the game gave them structured feedback after a miss. You'll also find the 2023 letter in Neuron that challenged the claims of intelligence and sentience. Also, the tables in every guide now fit on your phone.
Guides with sources
I've put four guides on the site and each one lists its sources at the end. They cover what wetware means and organoid intelligence. There's one on the power your brain uses next to a GPU (and where that comparison stops being fair). The last is a case study of the two sniper bots that bought into our pool 36 seconds after it opened on 21 August 2026. Also, the site counts visits without cookies and keeps no IP addresses. The privacy page tells you what's recorded.