Failed projects


A therapeutic for levodopa-induced dyskinesia in Parkinson’s

The hypotheses:

  1. Dyskinesia in Parkinson’s is poorly treated. It’s a serious problem, and the only real solutions are amantadine and deep brain stimulation. Amantadine is not an easy drug: it can cause nightmares and hallucinations. Deep brain stimulation requires invasive surgery.
  2. Dyskinesia in Parkinson’s is caused by dopaminergic dysfunction, which can be treated by NMDA antagonists. This should be a safe, effective treatment which will produce dramatic results that are easy to see (unlike, say, disease modifying treatments for Parkinson’s, which are masked by levodopa until levodopa stops working).
  3. Other dyskinesias will also benefit from this therapeutic.


  1. Not enough to make it worthwhile.

What I learned:

Not every interesting scientific idea is a business.

A therapeutic for glutamate excitoxicity

The hypotheses:

  1. Glutamate excitoxicity isn’t responsible for all neurodegeneration. However, there is some neurodegeneration that is caused by glutamate excitoxicity.
  2. In neurodegeneration that is caused by glutamate excitoxicity, NMDA antagonists will prove effective.
  3. Neurodegeneration, on the whole, is poorly treated and frankly terrifying. For ALS, traumatic brain injuries with neurodegeneration, Alzheimer’s, etc. , there is really nothing that can be done currently.

Evidence: not enough to be worthwhile

What I learned: It’s really, really expensive and difficult to get to the point of getting real evidence for treatments of neurodegeneration. Any neurodegenerative treatment, in order to get funding, needs to follow what people already expect will help neurodegeneration. Out-of-the-box ideas are very tough.

A therapeutic for chronic heart failure

The hypotheses:

  1. Despite being common, chronic heart failure is not well treated. If there are no obvious causes (and there usually aren’t), care is more palliative than curative.
  2. The unfolded protein response is a major cause of damage in chronic heart failure, and, when untreated, can lead to decline in heart function.
  3. Some combination of TUDCA, 4-PBA, and a sigma-one receptor agonist can treat this major cause of damage.


  1. ER Stress definitely occurs in heart failure (it makes sense, there’s a lot of damage happening all around).
  2. Medications to alleviate ER stress (TUDCA, 4-PBA) help in mouse models of heart failure, which are done mechanically.
  3. Sigma one receptor agonists can help in animal models of heart failure, seemingly by acting as molecular chaperones in the ER.

What I learned: someone beat me to it. They tested sertraline, an antidepressant with significant affinity for the sigma one receptor, in patients with depression and heart failure. It helped the depression but not the heart failure.

More broadly, there was a big problem with the gap between the animal model and the human condition. Humans experience heart failure chronically. There’s no equivalent of mechanically restricting the aortic valve. So I think there’s definitely something missing about the model.

Using RNAi for agriculture

Hypothesis: RNAi can be used to delete any RNA strands that it matches with. It could be used as an agricultural technique for pesky pests. For example, oysters are often devastated by herpes. Infecting the oysters with a bacteria that produces RNAi to target the RNA of herpes could be effective prevention.

Evidence: there are a lot of pretty amazing papers showing the use of RNAi against various types of pests. The original one I saw showed RNAi being effectively used in honeybees against the agents responsible for colony collapse disorder.

What I learned: there are a lot of RNAi companies that have tried and failed. I think it’s a combination of a mess of patents, difficulty with scaling up RNA production to be cost effective, and the caution people have for “genetic engineering”.

Exploring ketamine for blood pressure

Hypothesis: never quite got there. Something to do with raising blood pressure.

Evidence: Ketamine raises blood pressure, which is surprising for a dissociative. This is probably because it inhibits the norepinephrine transporter, and norepinephrine is used to raise blood pressure. Ketamine is also an anesthetic. Surgeons use it to anesthetize while raising blood pressure, as anesthesia/surgery can lower blood pressure.

What I learned: blood pressure and arrhythmia are really specific. We have pretty good tools for most stuff with blood pressure, and any new tools probably need to be very specific. Also, ketamine on its own isn’t a good tool to market, because you can’t patent protect it.

The roles of anything other than autoimmunity in type I diabetes

Hypothesis: there might be other sides of diabetes besides autoimmunity that could be targeted.

Evidence: eh…there’s suggestive stuff around the animal models, but nothing convincing in human models. Plus, it’s hard to argue with how effective monoclonal antibodies have been in autoimmunity.

What I learned: I should avoid autoimmune diseases without monoclonal antibodies. They’re just too good.

Treating obesity-related asthma/inflammation

Hypothesis: it’d be possible to alleviate obesity-related asthma (and other obesity-related disorders) with natural extracts that would have a beneficial effect both on inflammation and on weight/insulin resistance.

This would be better than the first-line treatment for inflammation, which is corticosteroids, which are notorious for weight loss.

Evidence: there are suggestive papers around curcumin, thymoquinone, and osthole regarding reducing inflammation and encouraging weight loss. Also, using corticosteroids as anti-inflammation in obesity-related diseases is weird, considering how notorious corticosteroids are for causing weight gain.

Unfortunately, I never knew how far to trust the papers on these sort of “natural” remedies, as they always seem to be by people with an agenda for these natural remedies who found exactly the biomarkers they looked for. Then, when I was even trying to clearly link to inflammation at all to obesity-related asthma, I find it’s really only a disease found in women, and I had zero hypotheses for htat.

What I learned: obesity and inflammation are really complicated. It’s probably best not to mess with obesity, and inflammation is really best with a monoclonal antibody. Tnf-alpha inhibitors do pretty great with inflammation, although they have severe side effects.

Education business

$1 live classes

Hypothesis: a good way to make some money would be to offer $1 live classes to mass groups of people.

Evidence: eh… none, really. I just offered it and see how it went.

Why it failed: it’s hard to make money that way. Processing fees are murder on $1 classes (most of them charge some amount of cents per processing fee). Plus, I just don’t like teaching live classes that much. It’s pretty stress inducing.

What I learned: pay attention to processing fees.

SAT/ACT tutoring

Hypothesis: I could hire Harvard undergrads to tutor kids in the SAT/ACT.

Evidence: Seemed pretty obvious.

Why it failed: I didn’t really like it. I don’t like the college admissions game or dealing with undergrads. The entire thing felt bad.

What I learned: I wish it was “don’t try to do what you don’t particularly like”, but that was a repeated mistake I made.

Selling prerecorded LSAT lectures

Hypothesis: I could sell my live LSAT classes as a recording.

Evidence: Seemed pretty obvious.

Why it failed: I was always aware that the LSAT, if they found out, would likely sue me, as I was using their copyrighted material and they’re notoriously litigious. So, it made it hard to promote.

What I learned: Either invite lawsuits or don’t, but don’t half-ass invite a lawsuit.

Other businesses

Group buying for restaurants

Hypothesis: I could lower costs for food businesses by group buying for them, or at least finding out the price from food vendors.

Evidence: I was aware that price discovery for wholesale food was really hard.

Why it failed: food businesses don’t want to talk to people they don’t know about their invoices, in part because a lot of them are cheating on their taxes. Also, didn’t have a convincing sales pitch because I didn’t know anything.

What I learned: don’t get involved in businesses you don’t know about.

Automated invoice entry

Hypothesis: it’d make my group buying business easier if I could automate invoice entry (see the problem here?). Also maybe I could sell the invoice entry app.

Evidence: ehh…

Why it failed: I did almost no research on what options were out there, and just went with my instinct.

What I learned: research beforehand.

Automated receipt extraction from email

Hypothesis: something similar to Slice, where receipts would get extracted from the email. Not sure why I thought this was a good idea, to be honest.

Evidence: ehh…

Why it failed: I did almost no research beforehand.

What I learned: research beforehand (I learned this again).

Collaborative to-do list instead of email

Hypothesis: considering how many emails are either to-dos or checking on to-dos, it’d be better if that was simply built into the email instead. I even created a name/app for it,

Evidence: seemed self-evident, but the big question was always if people were willing to let it be part of their email process

Why it failed: I initially planned it to be something my brother’s company would use, which he agreed to. Then they got crazy busy, and no longer wanted to change any of their task management processes. I lost interest in promoting it because I never was that into it in the first place.

What I learned: if you’re not willing to put in a lot of hours into promoting your business (e.g. if you don’t really like the space), don’t start it.