100 Lessons Learned from Shutting down my Anti-Aging company

Making the most of the intangible assets

I created this post to record some key lessons from shutting down the company I founded, Project AgeTuneUp.

Project AgeTuneUp logo. That’s right, we got around to designing a new logo before we got around to picking a better name

This was a company aimed at reducing the cost of NAD+ precursors by several orders of magnitude by scaling up production using Algae. Born out of a research project I worked on during undergrad, we were able to bootstrap this project and turn it into an actual company (admittedly with the help of some windfall from massive Ethereum gains). We assembled a lab space, got a team together, built a prototype, and raised money from angel investors. However, at the advice of some much more experienced Life Sciences VCs, we realized that our strategy was untennable. We ended up returning the other investors all of their money, all $500,000.

Peter Thiel says that startup failures are overdetermined, and that it’s only really success that one can learn from. On the scale of Thielisms from “contrarian” to “common knowledge”, this is pretty close to the “No shit, Sherlock” side of that spectrum. After all, if you’re going to ask someone for advice on creating a biotech company, who would you rather talk to?

Still, when any company goes down, there’s usually some attempt to sell assets and make sure they’re usable by someone else. Now that most of the equipment and consumeables have been sold, I decided to fully document the more intangible things I learned from this misadventure. Here’s 100 lessons learned, divided into a few key categories: Starting up, Assembling and Managing a Team, Growth, Personal Optimization, Running a Lab as a whole, Startup Finances, Strategy, Legal, Interfacing with Investors, VCs, and Advisors, and Final Thoughts

Starting Up

1. Laboratory liquidation sales can be goldmines for your early resources.

When it comes to startup costs, biotech startups can be almost pohibitively expensive. Fortunately, you can get a lot of the equipment you need second-hand. These are invaluable sources of materials or consumables for biotech projects. While there are a lot of projects aimed at making DIY lab equipment, like Open QPCR, most of these are redundant if you can get your hands on old equipment.

2. BUT, you should not rely on them forever.

There is always the danger that the items you get from liquidation sales might not be in ideal shape. Measuring equipment might be broken. The seals on containers of consumables might be broken. You can probably use these to put together a proof of concept, but eventually you’ll see that there’s a tradeoff in using these. The items might cost less monetarily, but they might cost more temporally.

3. Finding a co-founder is critical, verging on Mandatory.

Even though going solo is preferable to having a bad co-founder, comparing those two is like asking whether you or Javier Sotomayor could get closer to the moon by jumping: Absolute chances of success are not great for either, and there is clearly a better strategy. For finding a co-founder, optimize for complimentary skills, continuous learning and self-improvement, flexibility, integrity, energy, and emotional intelligence

Assembling and Managing a Team

4. PhD in biology or biochemistry is not specific enough.

Your team will probably need to be much more specialized. One of the issues we had was that, while some of our team members did have PhDs from schools like UC Berkeley in biochemistry, their specific areas of study were more heavily on eukaryotic biology, and not necessarily in humans. We were trying to make NMN cheaper by using metabolic engineering in bacteria (and later algae). If you’re working on a compound or device that’s going to be affecting a specific organ system, you need to find people that have actually done research in that field.

5. After a certain point, being a genius in a subject becomes less relevant if other components are missing.

IQ does not get you as far as it does in spaces like Mathematics or Physics. Yes, there is a minimum level of intellect needed to run a biotech company, as with any endeavor humans can do. You’ll find that success is much more dependent on other qualities like teamwork, communication, “EQ”, domain knowledge, and willingness to continue learning whatever you need to about law, operations, and new discoveries.

6. Make sure there’s enough buffer for everyone.

Yes, you need to carefully plan your schedule down to the hour, but given the inherrent riskiness of this, you need to have some kinds of contingency plans. Make sure your schedules, and those of everyone on your team, can be quickly changed in the event certain experiments or plans have to change

7. Stay away from incessant and persistent negativity.

It is exhausting and draining, especially when there is very little use for it. Positive, high-performing employees can have positive effects on the productivity of team-members closest to them, but an overwhelmingly negative team-mate can have detrimental effects on everyon they come into contact with.

8. Much like 10X programmers, there are 10X research associates.

A highly skilled research associate can do wonders. They can run experiments with minimal human error, they can give you some wiggle room when it comes to scaling up and automating, and they can be a valuable source of insight into how practical your research strategy is. A bad RA or even an average RA may not even be worth the low salary.

9. Background for RAs is critical.

On the subject of RAs, there’s a reason companies like Ginkgo tend to focus most of their hiring from MIT and the UC schools. Many elite schools’ biology curriculums have suprisingly little emphasis on practical laboratory techniques than you would think. It’s also important to focus more on the background because much of this stuff cannot be self-taught like computer science, and there are enormous qualitative differences between someone who can ace a bunch of tests on theory, and someone that’s worked in a wet lab throughout high school and all through college (not just during the summers).

10. Management isn’t just a designated job title.

It’s the requirement that everyone on your team be capable of owning their work and executing properly.

11. Hire fast, but don’t be afraid to fire fast, too.

Yes, if you do fire someone, you need to provide some kind of severance pay. Still, keeping a poorly performing teammate on the project for too long will be much more costly.

Growth

12. Nothing works the way you would naively think it works (for better and for worse)

You are going to very quickly find out that your preconceptions of how VCs work, how research works, how technology transfer works, and even how companies themselves work are all going to be challenged sooner rather than later.

13. Life Sciences are speeding up.

Biomedical sciences are moving much faster than you think, even if you think they’re already moving pretty damn fast to begin with. One week you may have thought of an impressive protocol or modification strategy. Next week some other group may have demonstrated the exact same thing, or even rendered your strategy obselete.

14. When budgeting and planning for growth, focus on automation.

Setting up a budget for opentrons is a good place to start. They’re python-programmable, they can use a lot of the materials you might already be using, and they’re much cheaper than a Tecan Evo (you can probably buy 20 OpenTrons for the yearly salary of an RA II).

15. Make sure you focus on optimizing your tools and workflows

Whether it’s resupplying reagents or processing emails, you want to make sure you dedicate some time to making sure these processes can be sped up. Not doing so will cost you in the long-run.

16. If given the choice between a loyal employee and a cheap employee (competence otherwise being equal), pick loyal.

This is another case of getting what you pay for.

17. Starting out in a new organization can be both a blessing and a curse.

Pro: You’ve got a clean slate to work with compared to established companies or labs. None of the bureacracy that you saw as coutnerproductive. Con: You’ve got to fill it in with something, and fast, otherwise your organization is doomed.

18. Momentum is critical.

If you have very little, your company is in grave and immediate danger. Once lost, Momentum can be very difficult to get back

19. There is such thing as moving too fast and breaking too much.

Later on in this post I mention regulatory compliance. There is a much greater burden of proof that your drug or product does what it says. There is plenty that can go wrong in the quality control process.

Personal Optimization

20. Management is something that can be taught and learned (and you absolutely need to learn it).

Nobody is born a competent manager. This is an area you can be coached and taught in. Regardless of your interest in the subject, you’re going to need to learn to become a manager.

21. Scientists will need to brush up on a lot of new terminology, and fast.

There is going to be a lot of new business and operations terms and concepts that are unfamiliar to scientists. This is often the case with companies in general, but it is especially so for biotech companies

22. For most projects, especially biomedical research, an undergrad degree is nowhere near enough.

Yes, there are plenty of people who started technically challenging companies with undergrad degrees, or even after dropping out. However, Computer science is relatively easy to get started in at a young age (e.g., most people seem to forget that Mark Zuckerburg already had experience creating a company and selling it to Microsoft before he even began at Harvard). As we all saw with the case of Theranos, there are big differences between CS dropouts and Biomedical engineering dropouts.

23. Take emotion out of the equation

This whole process is going to be emotionally taxing. While you do need enough emotional involvement to maintain minimum emotion, going too far beyond that is dangerous.

24. Knowing how to say “I don’t know” is going to be the difference between life and death.

After all, “It’s not what you don’t know that hurts you. It’s what you think you do that just ain’t so”.

25. There’s a very curious “grass is always greener” dynamic between CS experts and biologists.

Both seem to think the other has some kind of advantage, and some hidden ways ov avoiding the blind alleys altogether.

26. Being a good multi-tasker is not quite accurate.

It’s more a question of intense prioritization, scheduling, filtering, and process optimization.

27. You must ruthlessly prioritize everything in your life, not just in your startup’s operations.

Do a 5/25 analysis of everything in your life at the very least.

28. When you’re working on a startup, this is one of the worst possible times to be in a new relationship.

Even when using strategies like “setting a designated date night once per week”, it is tough. This is probably something best reserved for when and if your venture is making a profit or has customers.

29. Make sure your feedback loops are tight, even if they’re already pretty damn tight.

Set up news alerts for events related to the problem you’re working on. Make sure experiments don’t fail silently.

30. Make sure you never stop receiving constructive criticism.

Receiving no feedback or criticism at all is much worse than receiving negative feedback.

31. “First Principles Thinking” is critical.

It’s obviously tricker to do verifiably in biological sciences than in computer science or physics, but it is still doable. In our case, one of the things we did correctly was recognizing that our model organism’s NMN output was a function of the amount of energy it was able to take in. Hence, we were able to switch to a promising spieces of Algae as a production vehicle.

32. If you’re in a leadership role, you absolutely need to seem like you have your life together.

If you’ve spending a lot of your net worth on this project, or if you’re going through some kind of personal trouble, or even if you’re not getting enough sleep, you need to do whatever it takes to make sure your teammates, advisors, and investors do not catch on. When it comes to signalling, bragging on Twitter about working 100-hour weeks to achieve a goal is usually seen as a positive. Actually looking like you just finished with a 100-hour week has the opposite effect.

33. Don’t get derailed by all-nighters.

Keep a habit tracker to make sure you’re getting proper sleep and at least minimal exercise (e.g., a 100 push-ups/100 crunches/100 squats/10km per day regimen should be both doable with a tight schedule, cheap, and at least partially mitigate stress).

34. Entrepreneurship will reveal some painful truths to you

Starting a company will at some point involve the painful process of making your bare motivations, and how strong they are, fully visible to you.

35. There is an enormous difference between learning in school, and learning to make sure you run a company right.

Unless you see to getting the feedback yourself, you’re not going to get grades on being able to parrot trivia, your’re not going to get any credit for something unless the project actually works.

36. Read EVERYTHING you can on the history of biotech companies.

These are at least 10X more valueable than any other business resources. There are entire stores’ worth of books on running business where the stakes are relatively low.

Running a Lab as a whole

37. There are different, context-specific approaches to interviewing.

Many of these are situation- and context-specific. Some approaches resemble trying to simulate what it’s like to have a candidate as a co-worker. Others seem to resemble the approaches of the Spanish Inquisition more. Ultimately, it comes down to how badly you’re in need of someone with the requiements, but you should not be afraid to take the “inquisition” approach. The consequences of a false positve can be fatal, after all.

38. Make sure there’s enough literal Buffer for everyone

(unlike in lesson #6, this time I mean actual buffer solution). It’s like having clean socks or clean underwhere. You usually don’t put much thought into it, until you suddenly have the horrifying realization too late that you should have stockpiled that much sooner.

39. Most simulations of biological systems aren’t worth jack.

Even if experimental results line up with the simulations, those simulations are still worth much less than you think. Many of the underlying assumptions that went into the simulation, especially for coarse-grained simulations of molecular or protein dynamics, may be less justifiable than you think.

40. When making biologics, or doing anything with sythetic biology, the important equations will change drastically on a larger production scale.

Issues such as crystalization, nutrient flow throughout large containers, and oxygen/light gradients become much more important. It’s like transitioning from Newtonian physics to Relativistic physics when the objects in question are massive enough or moving very fast. Make sure you take this into account right at the drawing board, especially since making the production scale larger and faster is exactly what you’re trying to do.

41. There is very large variation in CRO quality.

Be very thorough in your homework when choosing a CRO to offload certain experiments to. If they’re cheap, you may be getting exactly what you’re paying for in the worst sense. In our case, we ended up using a mass spec facility that was used by Ginkgo Bioworks in their earlier years. We found out from some experienced life sciences VCs that there was a good reason why they stopped using that facility: The machines were poorly maintained and unrelaible compared to in-house mass spec.

42. Synthetic biology is a lot messier than electrical engineering or computer programming.

In theory, synthetic biology is supposed to operate as though cells are circuitboards. In practice, they’re more like tacos. You will find out very quickly how useless modelling organisms as discrete, deterministic logic gates is (and that’s before you take into account the fact that your model organisms can mutate pretty easily).

43. Waste disposal is an enormous cost, especially for a lab.

We very quckly found ourselves spending thousands of dollars on waste disposal from terracycle. Make sure you budget for worst-case scenarios here like spills and accidents.

44. Most investors and experienced academics will (wisely) not put much stake in fancy ML algorithms.

Aside from the obvious fact that neural networks do not actually resemble biological neurons, there’s still he fact that much of the data may be unreliable. No matter how many different layers of latent variables a model can come up with, it’s practicaly useless if it cannot properly account for alleoteric or epistemic uncertainty.

45. Deterministic simulations leave A LOT to be desired.

Stochastic simulations are under-used (though none of these will be truly useful until quantum computers are cheap).

46. Liquid nitrogen and -80°C freezers are the heart of your lab.

Make sure you keep backup generators for these. These are to a biotech company what hard drives and back-ups are to a software company.

47. Make sure you set up phone alerts for any incubators or fridges or freezers losing power.

If a freezer or incubator fails, it doesn’t matter if your alert wakes you up in the middle of the night. You need to be able to respond quickly.

49. Focus on maximizing the usefulness of meetings.

Make sure whomever calls a meeting creates a meeting agenda ahead of time.

Startup Finances

49. Budget conservatively.

Startups cost way more than you think. Create budgets for worst-case scenarios, while still making sure to cut unnecessary costs where possible

50. Money is gasoline.

Legally, the investors come first. Practically, the employees come first. If you cannot make payroll, you’re finished.

51. If you’re going to fail, fail fast.

Given the choice between a fast and capital-efficient “no” and a long “maybe”, choose the former. Delaying failure will be tempting, but the longer you do it, the nastier it will be when the bubble finally pops.

52. It is important to make distinctions between financial risk, scientific risk, and legal risk.

This was another mistake we made. We put so much effort into making sure our cheaper NMN system was scientifically sound. We relied on everything from finite element simulations to actual experiments on microbes to validate the idea. Unfortunately, we did not put as much effort into reducing the financial or legal risk (more on that later).

53. Your top priority should be getting your product to the clinic.

If you’re coming from a research background, you may be tempted to use published research as a signalling method. If this is the case, you absolutely need to abandon this mindset.

54. Tales of founders self-funding projects are filled with survivorship bias.

Chances are if the company is at a stage where you’re considering funding it yourself, it’s probably in more trouble than you realize. If you’re deciding to fund it because other investors won’t buy into it, that should also raise some red flags.

Strategy

55. Platforms are nice, but killer applications are the ideal

Foundational technology platforms (like genomics, antibodies, and CRISPR) have led to some major biotech successes. However, the founding technology itself is usually of little value compared to the drugs ultimately enabled by the discovery. Selection of the platform’s killer apps — the clinical indications to which the platform will be applied — will make or break your company. In our case we took a platform for optimizing production of metabolic compounds that was extremely promising. However, our bet on either NMN, the decision to apply it to Tau Tangles to address Alzheimer’s, or possibly both, contributed to the downfall.

56. Biotech investing is pretty dependent on blockbuster drugs or monopolies.

Raising money, recruiting people and building a movement around your mission almost demands irrational optimism, especially in failure-riddled drug development. The endeavor will either be #1 in the industry, or it will slide into irrelevance. You need to strike a balance between this kind of optimism and acknowledgement of the true risks.

57. You need to pick a specific condition or disease that your drug or product is going to treat.

Scientifically speaking, many compounds may be useful for treating multiple conditions. Operationally speaking, you’re going to need to explicitly define what disease you want to treat, and how you’re going to get to that space. We chose NMN because it seemed like megadoses could have promising effects on symptoms of aging. The obvious problem is that The FDA doesn’t currently recognize “aging” as a disease. After tearing through every research paper we could find on the NAD+ pathway and it’s role in age-related diseases, we picked ammelioration of tau aggregation in Alzheimer’s patients. Ultimately, we shut the company down before we actually got the chance to verify that this was the case.

58. Your Blue-Ocean opportunity is probably temporary.

Even if you do stumble upon a “Blue Ocean”-opportunity, it can quickly turn “Red” if you’re not fast enough. When the project was started in 2015, it seemed like there was definitely an opportunity. Most of the reputable suppliers of NAD+ precursors either charged a lot for the compound. Meanwhile the suppliers that charged very little were more often than not very ir-reputable. If we could bridge the gap between cost and quality control, we could solve the problem. Come 2017, we realized the situation had changed. While many of the companies that made NAD+ precursors were mainly focusing on Nicotinamide riboside, many of the ones like Chromadex and Elysium were in some truly vicious legal battles. We realized that we had failed to take into account this kind of financial or legal risk of getting into a potential competition (or worse, legal squabble) with them.

59. In SynBio, there is a big gulf between prototype and MVP.

Much of getting reliable results out of synthetic biology, especially when it comes to scaling up production or reducing false-positives or false-negatives, is dependent on rapid optimization. Unfortunately, there are wide-reaching patents on a lot of the technology for directed evolution.

60. One plan of action is not enough.

You need to have back-up plans on top of back-up plans on top of back-up plans. Yes, you need to demonstrate confidence in your idea, but a lot of investors will wisely want to see that you have contingency plans. Basically, you’re probably going to want to draw up at least 5 back-up business plans.

61. Unlike CS, the eccentric founder is far from the ideal in life-sciences startups.

This is especially the case for subfields (like aging) still trying to gain legitimacy in the eyes of institutions and the public.

62. Also, unlike in CS, your project can literally die

You are going to spend a lot of time maintaining your cell lines and keeping things properly preserved. Even then, these don’t last forever (and things like DNA can be much more tough to recover than you think).

63. Biology is inherently noisy. Focus on minimizing variance.

Biotech companies are built on science, brand, and business models. The business model in biotech is usually exit by aquisition or IPO. The brand is typically the stock-photo-filled, generic website. Everything else is usually focused on the science. We made the mistake of trying to tune all 3 at the same time in some cases. This was especially worrisome as, despite our focus on Tau Tangles in Alzheimer’s down the road, we were effectively a platform company that was operating on much longer time-scales than other biotech companies. The fact that we chose a model-organism-based discovery-engine approach like the infamous Exelixis made this even worse. My advice: Try to reduce this maneuvering to one variable, everything else being constants.

64. Huge swaths of scientific literature are either imprecise, or just plain wrong.

A company is where things will really hit the fan. When we were working on metabolic engineering for scaling up NMN production in prokaryotes, we scanned through more than 100 papers and databases for info on the NAD+ pathway. Not only were many of the kinematic constants for the enzyme dynamics too imprecise to be useful, a lot of the information was missing entirely. We were able to make strides with a symbolic regression algorithm that could make esimates of the missing constants without having to resort to n-body problems. We were actually able to use this to pick which genes to modify in the NAD+ pathway and how much to do so. Not only did our model organisms not die, but we even produced results showing that they could actually secrete the excess NMN they produced. Still, many investors pointed out that this was still just a band-aid for a much larger issue with the research we were basing our science off of.

65. Be willing to take advantage of other groups’ work.

Don’t reinvent the wheel. There is so much to be made in re-selling and re-branding drugs for which the patent has expired. I recently saw several friends of mine (Including the brilliant Nathaniel Brooks Horwitz and Nikita Shah of Nivien Therapeutics) participate in the Roviant Biotech pitch competition. Every single participant’s strategy, including that of the winners, was to pick a failed drug from a database like Adis Insight, and apply one with a soon-to-expire patent to some other disease. That’s all. For using this hueristic most effectively. It’s not just pitch competitions. My friends Josh Cohen and Justin Klee of Amylyx Pharmaceuticals also took the same approach, and have made some incredible breakthroughs towards treating Alzheimer’s because of it.

66. In general, beware of “Complexity addiction”

This is in many ways a generalization of the previous two points, combined with “Don’t let Perfect be the enemy of Done”. I would characterize “Complexity addiction” this way: An actor comes up with a ridiculously elaborate strategy that is so meticulously planned out that it can’t possibly fail. But it will. So why didn’t they come up with a simpler, more mundane plan? There’s a simple explanation. They’re addicted to coming up with complex schemes, like a gambit roulette. They simply can’t help but make an overdone, overblown plan that requires so many parts to move correctly. Maybe it’s out of boredom or insanity or pride. Whether business planning or even just reaching out to people, don’t make your strategies overly complex.

67. Keep your balance sheet clean

When it comes to fundraising, you can obviously raise too little. Likewise, you can also raise so much, that when you fail to meet certain growth targets by the tiniest amount, the valuation of the company plummets. Similarly, you want to make sure you retain at least some equity for the founders and employees, but squeezing valuation and investment terms to perfection may result in less than perfect relationships with investors.

68. You need to be afilliated with whichever university produced the research your product is based on.

If you’re working on something based on research at a university, you should probably be directly affiliated with the university itself if not the lab. This will make questions about IP and technology transfer much less o a headache. In my case, my work as a research associate at Harvard Medical School was not enough. I should have found a co-founder that was either an undergraduate or graduate student at Harvard.

69. Technology transfer strategies vary widely between universities.

Some universities are quite agreeable when it comes to defining terms for spinning out intellectual property into companies. Others, not so much. We found this out the hard way when talking to advisors of ours at MIT’s Engine. While MIT is a bit more lenient when it comes to technology transfer terms, universities like Harvard have a much tighter grip. Even with the blessing of a professor that worked on the technology in question, that still doesn’t guarantee being able to get IP rights at all, let alone favorable terms.

70. You need to have a plan for when part of your business and/or scientific justification depends on unpublished data.

Some of the objections raised by a few scientists and investors were focused on the differences in available transport pathways for NR and NMN. We had reason to beleive that there more available transport methods of NMN, as discovered by Dr. Shin-ichiro Imai’s group at Washington University School of Medicine in St. Louis, and research out of Dr. R. Grace Zhai’s group at the University of Miami School of Medicine. Unfortunately, at the time none of this was available in any published research (not even in preprint yet), and obviously saying “Just take our word for it” was not an option. Before you find yourself in this space, make sure you reach out to the groups whose unreleased evidence or research supports your conclusions beforehand.

71. You will probably need legal help or legal advice MUCH sooner than you think.

Find some good IP and corporate lawyers. As you can see from the previous points, we very quickly found ourselves in need of expert legal advice.

72. If you’re getting help from any non-profit organization or B-Corp, make sure there are easy communication channels so you don’t accidentally become a headache for them.

You need to make it abundandly clear which organization is doing what. There may be things you want to do, especially those related to startup fundraising, that may conflict with the non-profit’s or B-Corp’s charter.

74. One bad apple can spoil the whole barrel.

After getting the angel investors on board, we talked to some actual life sciences VC firms. This was the part where my background became an enormous liability, as it offered an incredibly easy comparison with Elizabeth Holmes (the part about us only hacing undergrad backgrounds inthe space). The takeway isn’t to be embittered about some scammer ruining chances of raising funds. The investors we talked to were wise to make this comparison. After all, if we failed, our failure could have a similar effect that Holmes had on biotech investing, except with the damage localized to the still-young anti-aging field.

75. However much you’re writing down in notebooks, it’s not enough.

You may have gotten in the habit of keeping notes of your experiments in grad school, but you need to be writing down even more. These are not just your records for scientific purposes, but also for legal and IP reasons.

76. Make sure to keep minutes or transcripts of every meeting.

This will make future arguments and disputes MUCH easier to resolve. Everyone will be much gladder for it.

77. Regulatory compliance can cost a lot more than you think.

Make sure you include this in the budget, and make sure you’re up to speed on regulations from federal to local level. We learned this the hard way when learning about differences in how biologics are treated between Cambridge, MA and Somerville, MA.

78. Yes, life sciences companies tend to IPO much sooner than Tech companies.

The takeaway is that they need to stand up to much more outside scrutiny much sooner than tech companies.

79. Patent litigation is serious business

Patent trolls are viewed as an annoyance by technology companies. After all, any halfway competent sofftware engineer has probably violated at least dozens of patents accidentally. In the case of biotech, these patents often carry much more weight, and the terms are often phrases so that they can apply to situations and technologies much different than what the original filer had in mind. The legal costs of a patent dispute can also be collosal (or even company-killing).

Interfacing with Investors, VCs, and Advisors

79. Fundraising can take a very long time.

Fundraising can take years even if you’re a famous scientist. Obviously, my teammates and I had far less fame than that. Expect to spend about 12 months trying to raise a round.

80. Your board and other investors matter a lot.

The other investors you have in a round (or even previous rounds) can have an enormous impact on current or prospective investors’s decision process, not necessarily for the better.

81. Startup names are criticaly important, and you should choose the correct one early.

We originally came up with “Project AgeTuneUp” as a placeholder. This was one of the bigger regrets I have about the venture. Names are much more important than you realize not just to investors, but also to regulators. See also the earlier point about keeping the branding and business model constant, while focusing on optimizing the science.

82. Domain expertise in investors, not just founders and advisors, is critical.

You want to bring aboard investors that have experience in biotech investing or running a biotech company. One of the mistakes we made earlier on was getting angel investment from ex-CEOs and ex-CTOs of technology companies. When we started talking to actual biotech VCs, the more experienced biotech investors saw this as a red flag. Even without this, having angel investors with little experience in biotech would have resulted in plenty of other operations problems down the road.

83. Put more weight on advice from people who already have or have done what you want.

Advice is cheap, but advice from people that have been in similar situations as you is golden.

84. For finding teammates, co-founders, advisors, investors, or mentors, mastering contact-marketting is your best bet.

Rather than spraying and praying with social media ads, create a short-list of the people that could help you the most in solving your problems. Keep your communications short, but memorable. Ideally, pick a more unusual medium of communication (i.e., physical mail, meeting them at an event they’re scheduled to be at). Your method of getting their attention should NOT be a stunt.

85. Life Sciences investors treat financial risk and scientific risk VERY differently.

Scientific risk, they can handle. What they cannot handle is financial risk. You need to have your unit economics all figured out. Our problem was that our whole business model relied on using a new, unproven technology to make the unit economics work out. This was an almost foolheardy amount of financial risk

86. Do not compromise on being choosy

Getting a company to an exit can take 10 years or more, and it’s far from a smooth ride. Be relentlessly picky when it comes to choosing where to accept funding from and who you’re going to work with. After all, you’re going to be spending more time with these people than many of your friends and family memebrs. Optimize for objectivity, resilience, and trustworthiness. Stay away from investors or colleagues that only want to work in a sub-field because it’s hot at the moment (they might be likable, but they won’t be long for this world)

**87. Competent investors will do due dilligence correctly. **

You’re better off being completely transparent to investors. If they are experienced, they WILL find out everything eventually.

88. Remember, trust is not permenant. It is perishable.

The half-life of trust is about 6 weeks for in-person interactions. 6 days for online interactions.

89. Optimize for signal-to-noise ratio when communicating with Mentors/Advisors.

If you do have a high-quality advisor or mentor helping you out, chances are they might have their hands full mentoring a lot of other people. Many of the same principles for pitching apply to this: keep your initial communications short. Try to get the gist of your question across in 15 words or fewer.

90. If your background is less than ideal, it may have a bigger impact than any changes in the idea itself.

Founder’s background is critical. This means everything from which school you went to to whether you have any published research in the area you’re working on.

Final Thoughts

91. Ironically, you may regret using a regret-minimization framework in your decision making

We’ve all heard the stores about choosing to start a company because you have some kind of bucket-list item you want to get done. Jeff Bezos is probably the most famous user of this “regret-minimization framework” when he founded Cadabra (fortunately renamed to Amazon). This was part of the reasoning I used when starting out on this.

92. If it will truly benefit the world, you should not be afraid to hand over control to someone else (especially if the other investors recommend it).

This was another piece of feedback I was given. However good my idea might be, it’s likely I’m not the best one to implement it. A lot of Founders complain a lot about being ousted from their companies, but a lot of the time it is truly necessary (especially in the case of Deep-tech companies).

94. This too shall pass.

No matter how good things may be going, no matter how much things are looking up, it can all be taken away in a flash. The compound you’re working on may prove to have unacceptably high toxicity. It may fail at Phase 2. Something might happen to one of your star teammates. You might be excited after closing a deal with an investor, and then you’ll find yourself having to jump out of the way of a car that ran a red light (for us, there was about a week-long interval between those two events, but I still have the road rash from the experience).

95. Failure does not mean resetting to a neutral state.

You can end up either ahead of behind where you started. You may be able to cut your losses early and “fail forward” for lack of a better term. You could also lose a lot of your own money, time, and even friends in the process.

96. Entrepreneurship is a gauntlet.

Elon Musk’s “looking into the abyss and chewing glass” analogy is uncomfortably accurate.

97. Don’t be so quick to sacrifice your own pay/savings.

Even if you yourself can go without pay if it means prolonging the venture’s operation, even if you can keep it afloat with an injection of your own capital, if you’re even considering either of these options the company is probably in more danger than you’re willing to acknowledge.

97. Know your customer.

If you’re working on anything that’s going to be given to patients, take time to actually talk to patients and patient advocates. Don’t over-optimize for pandering to investors. If your north star is solving unmet needs for patients, investors will ultimately line up.

98. The burden of opportunity is real.

There are worse problems to have, but this still can be a death-knell if not addressed right away. In short, this is basically the startup version of “He who chases two rabbits catches neither”. When you tro to focus on too many directions at once, that’s a recipe for failure, overwork, and burnout.

99. The startup route may be less than ideal if you’re trying to make an impact.

In the end, creating a biotech company because you want to make an impact is inherently irrational. High chance of failure. You’re probably better off joining a company like Zymergen or Ginkgo, or working at a VC firm.

100. Make sure you go through all the proper administrative paperwork.

When shutting down, make sure to file an “Articles of Dissolution” or “Certificate of Termination” document with the Secretary of State. Pay any back taxes the business may owe. Cancel permits, cancel licenses, and file abandonment forms for any fictitious business names. I’m fortunate that this is not one of the lessons I’ve learned the hard way (I filed all the shut-down paperwork correctly the first time), but it’s something that can cause trouble for other failed entrepreneurs further down the road even if the company is no longer in operation.


Anyway, I decided to record all this while it was still fresh. I decided to try an make the most productive use of all this. While Project AgeTuneUp still failed, I don’t beleive there was absolutely nothing to be learned from it.

Still, I’m no going to lie. This process hurt. It hurt a lot.

It’s not just the financial loss (yes, I’m aware that a lot of the Ethereum I used to fund this had absurdly high potential future value). What’s worse is that upon reflection, this seems like an enormous moral shortcoming on my part. I feel like I have offended my friends, mentors, and humanity as a whole because my work failed to reach the quality it should have.

As for next steps. At least one positive piece of feedback I gained from this whole endeavor is this: I have an uncommon skill for automation and probabilistic modelling. This was one of the few positive things that came out of the feedback sessions with life science investors, but it became much more apparent in a recent conversation I had with David Botstein, PhD.


If I do eventually want to return to biological aging (and by extension find moral redemption), my best bet is to become as good as I possibly can at machine learning. I post more updates on this process in the weeks, months, and years to come.

Subscribe to know whenever I post new content. I don't spam!


At least this isn't a full screen popup

That would be more annoying. Anyways, if you like what you're reading, consider subscribing to my newsletter! I'll notify you when I publish new posts - no spam.