Hacker News Reader

Clearly making it happen will require a revolution in manufacturing, which may or may not already be underway, but making it happen for software required decades of continuous revolution in semiconductors and telecommunications.

Some kinds of analysis machinery, like GC, ICP, and DSC or DTA, are probably inherently fairly large; other kinds, like FT-IR, other kinds of spectrometry, TLC, HPLC, other kinds of liquid chromatography, XRF, XRD, and NMR, can be miniaturized and mass-produced. There hasn't been much pressure to do this because bio and chem labs don't care if their spectrophotometer costs US$0.12 or US$12000 or whether it weighs 100 mg or 100 kg; they need one to get their work done, they don't need it to be portable, and they aren't going to lose it because it stays in the lab. But that doesn't mean it can't be done. Even Victorian-era-style reagent testing can be made quantitative in some cases!

> Some kinds of analysis machinery, like GC, ICP, and DSC or DTA, are probably inherently fairly large; other kinds, like FT-IR, other kinds of spectrometry, TLC, HPLC, other kinds of liquid chromatography, XRF, XRD, and NMR, can be miniaturized and mass-produced

Many of the types of analysis listed here are elemental analysis only, which are useless for trying to identify pharmaceutical analytes or determine their concentration.

Out of all of these, microfluidic liquid chromatography is the least science fiction. There's plenty of literature about it but nobody really "has it working", and the reality is that it's not likely to ever have the same capability as benchtop HPLC.

> Many of the types of analysis listed here are elemental analysis only, which are useless for trying to identify pharmaceutical analytes or determine their concentration.

That's mostly true, but if a pill has significant amounts of lead, arsenic, and mercury in it, you know something went wrong, and you shouldn't take it. Even XRF might be enough to allow you to safely use lead-based or arsenic-based catalysts in your synthesis.

> Out of all of these, microfluidic liquid chromatography is the least science fiction. There's plenty of literature about it but nobody really "has it working", and the reality is that it's not likely to ever have the same capability as benchtop HPLC.

Thanks! Can you think of any other plausibly miniaturizable general-purpose analysis techniques? Those are just the ones I came up with off the top of my head. I think microfluidic liquid chromatography doesn't actually have to run faster than the bear, just faster than color-changing DanceSafe test kits.

As for science fiction, https://news.ycombinator.com/item?id=29816434 talks a bit about how today's science fiction is tomorrow's old news.

Miniaturized NMR? can you explain how that would work?

Benchtop NMR spectrometers already exist (for decades now), and some are already cryogen-free, permitting room-temperature measurements, eliminating the dewar and cryogens which account for a lot of the mass and volume of traditional NMR spectrometers. We now have room-temperature superconductors, which might work to eliminate the bulky, heavy permanent magnets in current benchtop devices, though the pressures required may turn out to be impractical. Beyond that I can handwave at improved electronics and SQUIDs, but I don't really know.

Do you think there are some fundamental obstacles to miniaturizing NMR, and if so, what?

benchtop NMR doesn't solve this problem, it's not powerful enough.

If you had improvements to NMR they would actually go first to other things than doing chemical analysis of anarchist drug batches. IE there are other industries that will buy all your machines if they existed.

The real question is why would you EVER use NMR for just about anything? It's really high cost and the total value of the data is lower than just about any other technique. It really only makes sense in research situations.

What are the companies, or at least the industries, that would buy all the machines?

Ultimately what everyday people will end up using is whatever is cheap and works well enough. Right now NMR isn't cheap, and neither is FT-IR or XRD, but these things change over time. Benchtop NMR is already good enough for distinguishing between significant classes of contaminants that could be in your purported insulin.

I'm typing this on a 50-gigaflops computer, which is faster than the Cray Y-MP Los Alamos had back in the 01990s, and people routinely buy teraflops video cards now, any one of which is faster than ASCI Option Red, if you remember that. I just drank a mass-produced soft drink out of a can made of aluminum, the metal Napoleon III preferred to gold to exhibit his wealth. Last year Chinese companies brought three covid vaccines to market within six months of the disease's discovery and started mass vaccinations, though most observers had predicted a minimum of 18 months. SpaceX is routinely landing reusable rockets on their tails now, and the world's energy infrastructure is rapidly shifting from fossil fuels to solar.

Things change. Today's science fiction is tomorrow's old news.

You can't really escape the physics here, you need a very strong and very homogeneous magnetic field for NMR and very sensitive electronics to detect the signal. That's not something we can do for cheap right now.

And even beyond that I don't think the area has enough volumen and is competitive enough to produce significantly lower prices. The high-field NMR area is almost a monopoly right now, the benchtops and lower field instruments are somewhat competitive. But even those are in price areas far beyond someone doing synthesis at home.

As I asked you above, how sensitive do the electronics need to be? How weak is the signal?

Some things change, but to make NMR cost effective would be cheating mother nature. The entire technique is based on producing a strong and homogenous magnetic field that can hold a lot of sample, only to probe the sample with weak RF and listen to its faint echoes.

Industrial diamonds got "cheap" but large ones never did.

You could be right; certainly you know a great deal more about the subject than I do.

What are the companies, or at least the industries, that would buy all the machines if you could make them much cheaper and/or better?

any purveyor of fine chemicals (polymers, etc) and petroleum processing company (hydrocarbon analysis).

Thank you very much!

Benchtops are at 60-90 MHz field strengths. That is not really enough to look at more complex molecules, the bigger routine NMR spectrometers are at 400-600 MHz (and there are even larger ones, but those are not used for small molecules that much). And even then those benchtops cost something close to 100k USD, that's quite far from affordable.

The "room temperature" superconductors are not used at room temperature in these cases, they're still cooled down. And so far the only spectrometer I know of where they are used is the still extremely new 1.2 GHz Bruker. And that one is almost certainly somewhere between 10 and 20 million USD. The new superconductors are low temperature superconductors, not room temperature. And even then they still work better at lower temperatures. At best you can remove the liquid helium from the system and use liquid nitrogen only, which is an advantage but still really far from room temperature.

Thank you for explaining!

Yes, I don't know if the current room-temperature superconductor material (which really is room temperature, 15°C) will ever be useful for this; it was only discovered in 02020, so it is very unlikely that anyone is using it in a product today, even if they find a way to apply the necessary pressure (267 GPa, thus requiring ultrahard anvils). You're probably thinking of something like YBCO, which is "high-temperature" in the sense you're describing, requiring only LN₂, not "room-temperature".

Costs change over time. There was a time when solar panels cost 100k USD, too. A lot of the costs you're describing are NRE; others are costs that can be reduced.

The new spectrometers are using YBCO, and they are many years beyond schedule. That whole thing turned out to be a lot harder than many people seemed to expect.

The magnets are not the only cost in NMR spectrometers, I think you're seriously understimating the amount of electronics in them. You need to detect very weak signals at several hundreds of MHz, that's not trivial.

How weak are they? Detecting very weak signals (-110 dBm) at hundreds of MHz and even GHz is routinely done by Wi-Fi cards and cellphone radios, and GPS receivers detect signals that are orders of magnitude weaker than that (routinely -150 dBm), but only at tens of MHz. My eyes routinely detect submillilux signals when I look at the stars at night, with an integration time of well under a second; if I'm doing the calculations correctly, that's about -70 or -80 dBm in the 100–1000 THz band. PMTs (including microchannel plates) and SPADs routinely detect optical signals much weaker than that.

To a significant extent you can detect arbitrarily weak signals with coding gain and longer averaging times, although if your benchtop machine already takes ten minutes to give you a result, you probably can't afford to wait more than about 36 dB longer, give you another 18 dB of SNR).

So, I'm not worried about the electronics or the signal processing; there's no such thing as an "amount of electronics". Precision analog equipment is not easy to design, calibrate, and build, but you only need a very small "amount" of it, and it can be mass-produced.

Take resistors. When I was a kid back in the 01980s normal resistors were ±20% carbon composition, which would drift by more than 20% over time or if overvolted, with fiendish temperature coefficients. Now you can't buy a ±20% resistor; most resistors are ±1%, ±0.1% resistors are commonplace, and ±0.01% resistors are easily available for a dollar or two. Precision resistors are now made with an extremum of resistance around room temperature, so the temperature coefficient there is literally zero.

No, what I'm worried about is the physics. I'm not surprised YBCO spectrometers turned out to be a pain in the ass; YBCO is a huge pain in the ass in every possible way. What do you think the physical obstacles are?

The signals are weak, but more importantly, the signals are being detected in a volume wiht a huge magnetic field. You can't even put a digital circuit anywhere in the amplifier. When I worked on the NMR all the monitors in the room were shifted toward the magnet, it would wipe your credit cards, and to adjust the amplifier involved sitting under a multi-ton device twiddling knobs to minimize the impedence.

Think of NMR as bespoke. Like a large luxury liner built for a rich individual. It's not ever going to be a zodiac.

As I understand it, the magnetic field doesn't have to be voluminous; it just has to be strong and uniform. It can be arbitrarily small as long as it's large compared to an atom, which is why benchtop NMR machines can exist at all, though they won't ever be truly miniature unless either we can also miniaturize the cryostats and refrigerant supply, or we can find a usable room-temperature superconductor. Am I misunderstanding something? As I said, you know a lot more about this than I do.

NMR is an ensemble measurement technique. You can't measure the magnetic moment of a single molecule. Instead, you need a highly pure sample and lots of it. That has to be in solvent (I'm leaving out solid spinning NMR here, that's a different ballgame entirely), enough solvent to keep the sample in solution.

I think you're wasting your time trying to improve NMR. The value of the technique isn't worth it.

They use rare earth permanent magnets and don't offer the resolution of superconducting magnet NMR, but they are much smaller and cheaper (tens of thousands of dollars, new) than superconducting units or the even older resistive electromagnet NMR units. The first one I saw was from picoSpin, which has since been acquired by Thermo Fisher Scientific. I think that there are multiple vendors now. Here's a current picoSpin unit:

https://www.thermofisher.com/order/catalog/product/912A0913

An 80MHz desktop NMR in 2022 is hilarious. This owuld be great to put in a research lab or to teach students, but it's not something that could be used in a high volume, high quality pharma testing situation.

(my phd in nmr is from 20 years ago... even then it was hard to justify the expense of nmr machines in structural biology...)

Probably the 'manufacturing revolution' will be modifying yeast, plants, etc. to produce the chemical compound of interest. Everyone needn't have expensive equipment - testing can be outsourced. Such an effort would need a GPL like agreement to keep people honest, but the technology exists.

Conceivably, but I was a lot more enthusiastic about this possibility 35 years ago before the humans had much experience with it. It turns out DNA is a really shitty programming language for humans. Like, worse than Malbolge.

I'm not arguing with what you're saying, but GC/MS can be bought for like $100 a sample and it's conceivable a system can be invented where private auditors audit the output of an unregulated pharmaceutical manufacturer in such a way that the QC assurances to the consumer are as good or better as in our current system, with much lower regulatory costs.

In the limit what you are describing is a generic drugmaker.

It's certainly possible to audit drug quality by sending it to labs, and people do that for darknet drugs all the time, but there are still problems:

1) There is no way to use ex-post analysis alone to achieve the kind of QA that pharmaceutical companies do; they have visibility into the entire manufacturing process and process control. Put another way, a sample of a drug cannot be used to verify that the process used to manufacture it is safe.

2) There is no assurance that anything you get in the future is made with the same process.

The only way I see this working, honestly, is for a rogue jurisdiction to offer safe harbor to "generic pirates". The rogue jurisdiction would offer legitimate regulatory oversight in exchange for tax revenue, and the drugs would be smuggled out of the jurisdiction for sale. To some extent this is already the case in grey markets where brand name drugs which are sold for less in other countries get arbitraged/smuggled back to high-cost markets.

I think we are seeing the same thing, a system designed to remove corruption/regulatory capture/lobbying as far as possible from the process. I'm seeing a "rogue" auditor that performs the same service, while you are seeing a rogue government. If the outcome is the same, I am fine with either. Your pointed weakness regarding auditing the manufacturing process could be incorporated into either.

> I think we are seeing the same thing, a system designed to remove corruption/regulatory capture/lobbying as far as possible from the process

Not really. It's more like a system to selectively remove intellectual property rights without destroying the financial incentive to develop drugs.

Oh. So you don't want a system that removes corruption, regulatory capture, and lobbying as far as possible? I don't want to destroy the financial incentive either, I think it's great people that bring wanted goods to market profit from it.

IP is an interesting point, although I'm not convinced that generating the IP is more than a small fraction of the cost of a drug. An aggregate I saw from 2011-2018 puts Research at only 17% of revenue, and only some proportion of that is geared towards generating IP. That is to say, with all else removed you could generate an IP only company for 17% the cost of drug sales.

> Oh. So you don't want a system that removes corruption, regulatory capture, and lobbying as far as possible?

Well, sure. But that's not what this is actually about. The titular anarchists are not inventing new drugs, they're finding alternative ways to get drugs that have already been invented.

> That is to say, with all else removed you could generate an IP only company for 17% the cost of drug sales.

I am not convinced that accurately reflects the reality of running a pharmaceutical business, even if incentives/competition is somewhat warped in healthcare. There are a lot of penny-stock pharma companies whose only mission is to generate valuable IP. They're notorious for being volatile (since their outcome is basically binary on the results of research), but not for being great investments. If what you are asserting was actually true, those sorts of firms would generate ~5x higher returns, on average.

>Well, sure. But that's not what this is actually about. The titular anarchists are not inventing new drugs, they're finding alternative ways to get drugs that have already been invented.

... yes and they're doing so because in part regulatory capture, lobbying, and corruption have lead to unaffordable drugs. So that is, partially, what this is actually about. But sure there are other reasons for people to manufacture their own drugs outside of formal channels.

>I am not convinced that accurately reflects the reality of running a pharmaceutical business, even if incentives/competition is somewhat warped in healthcare. There are a lot of penny-stock pharma companies whose only mission is to generate valuable IP. They're notorious for being volatile (since their outcome is basically binary on the results of research), but not for being great investments. If what you are asserting was actually true, those sorts of firms would generate ~5x higher returns, on average.

That's actually my point. Generating the IP looks to be low value compared to wading through all the insane regulatory barriers, developing sales channels in-bred with medical board licensed doctors and FDA regulators that examine sales materials, lobbying for regulatory capture and all the other expensive pursuits that arguably may not add much value to the ill person consuming the drug. Another words, I'm not sure removing IP from the equation really changes the cost of drugs that much.

Here's a possible approach.

Buy n = 1024 doses of your insulin or whatever, D(0, i) for i from 0 to n - 1 = 1023, homogenize each dose, and divide each one in half into half-doses called E(0, i) and F(0, i).

Mix pairs F(0, 2*i) and F(0, 2*i + 1) into 512 new doses D(1, i) for i from 0 to 511. Homogenize these new doses and divide each one in half into half-doses called E(1, i) and F(1, i).

Mix pairs F(1, 2*i) and F(1, 2*i + 1) into 256 new doses D(2, i) for i from 0 to 255.

And so on, until in step k = lg n = 10 step you mix the half-doses F(k - 1 = 9, 0) and F(9, 1) into a single dose D(k = 10, 0). Send this D(k, 0) off to the lab to be analyzed.

If the lab is equipped to detect dangerous impurities in your insulin at one-thousandth the danger level, which is reasonable for many contaminants, and the sample comes back clean, then you know that all 1024 doses were safe, though some of them may have the wrong dose. Mix the remaining 1023 doses well so that they all have the same dose and store them safely.

If not, you need to track down the contamination (or massive dilution), so in the next iteration, you send E(9, 0) and E(9, 1) to the lab for analysis. If one of them comes back safe, you know the 511 doses that were mixed into it were okay, and you can mix them well and store them safely, then repeat the process on the contaminated subtree.

Depending on your cost function (latency, shipping and handling costs, etc.) and your priors for correlation among the samples, it might be worthwhile to recurse more deeply on failure: instead of sending E(9, i) for i from 0 to 1 to the lab, you might instead send them E(6, i) for i from 0 to 15. If one out of 30 doses was randomly contaminated, for example, about 14 out of the 16 groups will be bad on average, while if it's one out of 100, then you'll have about 7 bad groups out of 16. At some point you need to give up on the recursion, too, or you'll end up testing almost all 1024 doses when they're all bad.

This of course doesn't solve the problem of future buys, just reduces it by the factor of n.

>it's conceivable a system can be invented where private auditors audit the output of an unregulated pharmaceutical manufacturer in such a way that the QC assurances to the consumer are as good or better as in our current system, with much lower regulatory costs.

No, it's not. This is ideological libertarian nonsense. There's a reason pharma came to be regulated in the first place. All this will lead to are more injuries and death of consumers.

>No, it's not. This is ideological libertarian nonsense.

If I could flag your post, I would. This is purely political nonsense and a god-like attempt to disprove something through fiat.

Heck, even the FDA can't guarantee all generics are equivalent to the branded version (some anticonvulsants come to mind). Might be a while before this ability comes to the masses.

A lot of machinery has been getting smaller. You can do GC/MS with a desktop device now.