A tweet is about 1kb of memory*. All of the pictured things can store one tweet:
Two Babbage Analytical Engines
37 punch cards
A Russian woven electronic core memory module
This hard drive built inside Minecraft
*280 characters of up to 4 eight byte octets each in UTF-8.
For reference, the Apollo Guidance Computer, which got humans to the moon, did so with an amount of memory equivalent to 8 tweets.
Scientists have also successfully built logic gates by using swarms of soldier crabs. It takes about 80 🦀 to operate a logic gate, and there are 8 logic gates in a byte, so 640,000 crabs can be used to store a single tweet. Which seems kind of horrifying. 🦀🦀🦀🦀🦀🦀🦀🦀
So a typical crab swarm like the one below has 100M crabs. If the 🦀 were used to operate logic gates, that would store 156 tweets. Horrible, horrible tweets.
There also exists a tournament-legal deck of Magic the Gathering cards that turns any Magic game into a Turing Complete computer, capable of reading and writing data. In this case, you can encode a tweet with around 512 creature tokens (of 16 types). arxiv.org/pdf/1904.09828…
Maybe the most efficient super weird encoding method so far: slime mold circuits! It would only take a 2.5 inch by 2.5 inch slime mold to hold a single tweet, and, as the paper concludes: they are "self-growing & self-repairing & can be incorporated into hybrid wetware" (shudder)
• • •
Missing some Tweet in this thread? You can try to
force a refresh
A third of all genetics papers published in Nature over a decade (and 20% across all journals) had errors due to the fact that many gene have names like SEPT2 (the official name of Septin 2), which were automatically coded as dates by Microsoft Excel. genomebiology.biomedcentral.com/articles/10.11…
Also an Excel error cost led to an estimated 1,500 deaths in the UK last year. The UK COVID contact tracing effort saved to an old Excel file format (.XLS) rather than the new one (.XLSX) & smaller row limits resulted in names being dropped. They were never contacted by tracers.
It was easier to change genetics than to change Excel (or to get people to stop using Excel to track data it wasn’t built for... or to get people to understand how to format cells). 27 genes have so far been renamed to deal with the date problem.
I think we underestimate how much COVID disruption will help big organizations in the long run. So many bad routines were disrupted, and new improvisation rewarded. Reminds me of this paper: when star NBA players are injured the rest of the team has improve & create new patterns.
I’ve been talking to people at big companies that thought they would not have been able to make the switch to online, and it is clear that the sudden disruption to routines led to revelations. It is like a version of @work_matters & Rao’s subtraction game. hbr.org/2014/02/scalin…
Plus, the fact that every other competitor was disrupted in at the same time meant a breather from competitive pressure during the most confusing period. And the backstop of government aid helped, too.
So the real question is: what did your company learn from COVID practices?
I am somewhat thrilled that NFTs immediately moved beyond parody. Here’s people bidding to acquire a signed tweet about people bidding to get a signed Pokémon card.
And now you can buy a signed copy of my tweet about signed copies of another tweet about signed Pokémon cards. Please don’t buy it: v.cent.co/tweet/13700528…@Cent
Probably still beats substack as a long-term business model though.
The most sci-fi aspect of this century: if space development continues at the present rate, it is likely that the last time in history that all of of humanity was on Earth at the same time was November 2, 2000.
Currently, it looks like the moment that the ISS is retired (likely 2028) the private Axiom modules being attached to it will become their own commercial space station - the one in the image. And that assumes that no one else builds another space station first. After that? 🌙 🔴
The population of space. (The next barrier to cross: surpassing the maximum space population record...13) nefariousplots.com/figures/4
You may have heard that learning a second language makes you smarter, but more recent careful work, including meta-analyses of 157 studies, has found no advantage in any cognitive dimension for learning additional languages.
Similar negative news on chess. Previous work found that teaching kids chess improved a whole lot of outcomes, from math to logic to academics.... except that these studies were generally small. A large randomized trial with 4,000 students finds no advantages to learning chess.
But learning to code does actually improve other academic outcomes. 👇
Venture capital is the final big holdout of instinct-based investing in a finance world that has gone quant. This is from a giant survey of VCs & it shows only around 1 in 10 do any sort of quantitative analysis of past decisions & nearly half often make gut investment decisions
This is because analysis doesn’t help that much. But neither does any other method: picking an unexpected first winner appears to require a lot of luck (VCs who do it once don’t do it again), but then that luck can be turned into status, and status generates profitable deal flow!
The issue is that if you are just going on instinct, investment decisions can suffer from many biases; including homophily - the tendency of people to like people like themselves. That creates real consequences for both the types of startups that get funded & VC performance 👇