this post was submitted on 16 Sep 2024
35 points (100.0% liked)
TechTakes
42 readers
20 users here now
Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.
This is not debate club. Unless it’s amusing debate.
For actually-good tech, you want our NotAwfulTech community
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
right, so let's start with equipment. instead of using lab-scale techniques with lab-scale equipment, they instead decided to make an industrial-type reactor but tiny. so for a thing that stirs and heats your reaction mix they came up with arduino-powered jars that cost all in all $300-500. you know what does this job? a hotplate-magnetic stirrer combo with external thermometer that you can get on amazon for $70ish, with some associated glassware that still will be cheaper than $300 total, won't leach plasticizers and that you can actually clean up thoroughly. if you're looking for case of coder mindset where every problem can be solved with iot and 3dprinting, it's there. that thing only really allows purification by extraction and crystallization, but this is straight up not enough for some cases. maybe they're limiting their syntheses to cases where it is enough. sometimes you need to run distillation, if it's something other than solvent this means vacuum pump ($2k ballpark) with dry ice trap (that has to be fused quartz and not normal glass) and dry ice. sometimes you need to run chromatographic column, that alone is easy if you know what you're doing but it requires a lot of solvent that would be ideally recycled because there's a lot of solvent involved (100x mass of compound to be purified is on lower end) (honeywell doesn't want you to know this but recycled solvent is free you can do it in-house. i recycled 1500L). then you need equipment to evaporate that solvent (rotovap) and recycle it. like derek lowe said, it's a lot of like trying to make your own aluminum foil
and then there's entire waste management of it all, and there will be a lot of waste that you just can't flush down the drain
another problem is that they expect anyone to be tied to this rube goldberg model reactor and don't provide instructions in human readable form, but only as instructions for that contraption, and even that only through chatbot (??) if you can pull something let me know. (they have something better than machine generated free association right? as in, something they have tested. i don't expect orgsyn.org level work where they repeat all procedures they publish, but at least something grounded in reality)
like you said there's zero quality control and options for purification are extremely limited, this alone will get people killed. some of their testing was outsourced, and there are some ancient methods of determining what you've got based on things like melting point, but this is critically sensitive to fucking up due to not knowing what you're doing. but this won't tell you what it is, it only allows you to compare your result to someone's else result to check if they match. i don't know how likely in this case would be that fucking up synthesis will get patient killed, but in their previous one where they tried to cook naloxone there was a way to fuck up synthesis that would give compound with opposite activity to what was intended (oxymorphone).
regarding sofosbuvir. it's fluorinated, which means they either
e: by now i suspect they just ran with outputs from 2 without actually trying to make in their jarware. there's also stereogenic centre on phosphorus which can, but doesn't have to, make things significantly harder, but idk how actually important this is
and this is all forgetting about that sofosbuvir is used in combination therapy, where all the other drugs come from
the other drugs are also patented and just from looking at them it seems to me that these require a bit more specialized starting materials and a few steps that involve palladium couplings
synthesis-prediction software is an alleged thing that will sometimes work, i personally never did, but if you do use it, it's critically important to know what you just have cooked because these outputs are predictions, not procedures. in this case point is moot because they just can replicate what is in patents, as patents describe something that already works, even if some details are missing
there are degrees to it. their abortion cards from what i understand just repurpose veterinarian misoprostol, which has similar if not the same requirements for purity and such and because it's also used in humans, i won't be surprised if it (as in, API) rolls off the same production line. but here, they don't cook anything so there's less room to fuck up. this is also not golden standard on how pharmaceutical abortion can be done now, but it's available and it's basically unregulated which i suspect was the entire point
single sane take from hn that is grayed out for some reason. something like quarter of discussion involves nootropics
e: looked up their synthesis planner. for the easy example they mentioned (ivacaftor) 1/3 of the last step reactions are wrong, and these that aren't require mildly to wildly spicy reagents (allergens, toxic or corrosive, that last bit is expected). first few candidates split entire molecule in two parts across amide bond (correctly) into amine and acid. acid is listed at $1/g amine is listed at $28/g, which are prices cited in 404media article actual link provided lists amine at $28 per 10mg. predicted synthesis of amine does not include the actual pathway that i found on reaxys (and one that would work in garage). ivacaftor is listed as commercially available at $86/g, but link provided only lists 10mg at $12. i take they just punched these compounds into their wrong prediction machine and ran with numbers it provided without checking
that's not even synthesis planner like i've seen before that will provide you with suggested reaction conditions, any extra reagents and all that jazz, that still require careful supervision and good analysis of reaction mixture afterwards. no, here you have two bits that you can possibly stitch up to get something, fuck you, you won't get anything better. figure out all the other bits, draw the rest of the fucking owl on your own
Deep dive into complex chemistry may be the most important thing in this thread. The bullshitters need more pushback like this, even though the effort involved means it can't happen nearly as often as the bullshit. Thank you.
if you want a deep dive, show me a procedure this thing is supposed to be using because i found nothing serious
also just noticed that their warrant canary does not include proof of date like news snippet and is not signed