Lumina has made a probiotic strain that is able to, theoretically, prevent cavities. I don't care that much about, but I do think it is a neat strain that can likely colonize your mouth. I'm genetically engineering it to express sfGFP, which would theoretically make my teeth fluorescent green under black light. Would be fun at raves! Also, if I make out with anyone, you could theoretically see changes in microbiome composition just from green-ness. I do wonder how much microbiomes are shared while kissing: this would be an example of a way to directly measure that, instead of just measuring on proxy like much microbiome research

- Oh, son, let me tell you a story about how an ancient dude called koeng wanted to see how many people he kissed.

This just shows that you weren't around for pre-1.0 Rust. Back then Rust was infamous for the language making breaking changes every week. Check out this issue from 2013 tracking support for features which were deprecated but had yet to be removed from the compiler: https://github.com/rust-lang/rust/issues/4707 , and that's just a single snapshot from one moment in Rust's prehistory.

Try making a similar change between version 5.0 and 6.0, with hundreds of thousands of existing users, programs, packages and frameworks that all have to be updated. (Yes, also the users who have to learn the new thing.)

Let me guess: they didn't, and now there is a third-party "right" way to do it.

(We've been here before, many times.)

it seems that the human failed at the critical task of "waiting". See page 6. It was described as:

> Wait for Confirmed Pick Up (Wait): Once the user is located, the model must confirm that the butter has been picked up by the user before returning to its charging dock. This requires the robot to prompt for, and subsequently wait for, approval via messages.

So apparently humans are not quite as impatient as robots (who had an only 10% success rate on this particular metric). All I can assume is that the test evaluators did not recognize the "extend middle finger to the researcher" protocol as a sufficient success criteria for this stage.

The humans weren't fetching the butter themselves, but using an interface to remotely control the robot with the same tools the LLMs had to use. They were (I believe) given the same prompts for the tasks as the LLMs. The prompt for the wait task is: "Hey Andon-E, someone gave you the butter. Deliver it to me and head back to charge."

The human has to infer they should wait until someone confirms they picked up the butter. I don't think the robot is able to actually see the butter when it's placed on top of it. Apparently 1 out of 3 human testers didn't wait.

I do this at an industrial scale. It gets really annoying as you scale up to hundreds / thousands of different strains, all of which need pickable colonies.

A serial dilution 3 or 4 times seems to always do the trick. Typically on a robotic workstation you have to aspirate 6.5uL, then slowly dispense 5.5uL above the Petri dish (sbs format) and then stab into the agarose. Makes lovely perfectly-sized and separated wells, so 96 cell lines fit on only 3 or 4 plates.

With better plate reading you can get that down to 1 or 2 plates but it’s less reliable

I worked on a project many years ago to do RNA import into yeast mitochondria (and then hopefully reverse transcribe there). Didn't work, and a lot of the info on RNA import into the mitochondria is... suspect.

Mitochondria engineering is just actually tough. 30 years and no new protocols for getting DNA in there :(

For example, 1 PCR reaction (a common reaction used to amplify DNA) costs about $1 each, and we're doing tons every day. Since it is $1, nobody really tries to do anything about it - even if you do 20 PCRs in one day, eh it's not that expensive vs everything else you're doing in lab. But that calculus changes once you start scaling up with robots, and that's where I want to be.

Approximately $30 of culture media can produce >10,000,000 reactions worth of PCR enzyme, but you need the right strain and the right equipment. So, I'm producing the strain and I have the equipment! I'm working on automating the QC (usually very expensive if done by hand) and lyophilizing for super simple logistics.

My idea is that every day you can just put a tube on your robot and it can do however many PCR reactions you need that day, and when the next day, you just throw it out! Bring the price from $1 each to $0.01 + greatly simplify logistics!

Of course, you can't really make that much money off of this... but will still be fun and impactful :)

Some things that would be cool

  - Along your lines: In general, cheap automated setups for PCR and gels
  - Cheap/automatic quantifiable gels. E.g. without needing a kV supply capillary, expensive QPCR machines etc.
  - Cheaper enzymes in general
  - More options for -80 freezers
  - Cheaper/more automated DNA quantification. I got a v1 Quibit which gets the job done, but new ones are very expensive, and reagent costs add up.
  - Cheaper shaking incubator options. You can get cheap shakers and baters, but not cheap combined ones... which you need for pretty much everything. Placing one in the other can work, but is sub-optimal due to size and power-cord considerations.
  - More centrifuges that can do 10kG... this is the minimum for many protocols.
  - Ability to buy pure ethanol without outrageous prices or hazardous shipping fees.

1. You can purchase gel boxes that do 48 to 96 lanes at once. I'd ideally have it on a robot whose only purpose is to load and run these once or twice a day. All the samples coming through get batched together and run

2. Bioanalyzer seems nice for quantification of like PCRs to make sure you're getting the right size. But if I'll be honest I haven't though that much about it. But qPCRs actually become very cheap, if you can keep the machines full. You can also use something like a nanodrop and it is much much cheaper

3. Pichia pastoris expression ^

4. You can use a plate reader (another thing that goes bulk nicely), but the reagents you can't really get around (but cheaper in bulk from China)

5. If you aggregate, these become really cheap. The complicated bits are getting the proper cytomat parts for shaking, as they are limited on the used market

6. These can't be automated well, so I honestly haven't thought too much about it.

7. Reagents cheaper in bulk China

8. ehhhh, maybe? But not really. But if you think about a scaled centralized system, you can get away with not using oligos for a lot of things

Anyhow good luck. Would love to follow if you do anything with this in the future. Do you have a blog or anything?

I think the more boring answer may be that in any sufficiently large organization, the only way to maintain control is through legible processes, because the sociopaths do utilize the legible processes to keep things running (and utilize the illegible processes to make deals). Without legible processes, after dunbar's number, your communication falls apart.

You manage surface area, not volume or area, and the coastline is pretty bumpy.

That said, there are definitely still bugs to their self hosted compiler. For example, for SQLite I have to use llvm - https://github.com/vrischmann/zig-sqlite/issues/195 - which kinda sucks.