The Hard Thing #5: The Myth of Platform Biotech

“No matter who you are, most of the smartest people work for someone else.” - Bill Joy (co-founder of Sun Microsystems)

Biology is the most powerful technology on the planet (come at me, bros).

And there’s no doubt that AI is increasing our ability to understand and leverage the biological “toolkit” at a headspinning rate.

And yet… Techbio (or whatever you want to call it) has failed to have the transformative impact that has been “just around the corner” for the past 20 or so years.

In 2011, Marc Andreessen published a prescient Op-Ed in the WSJ titled Why Software is Eating the World.¹ A few years later, his fund, a16z, launched a dedicated platform bio effort and argued that the confluence of exponentially decreasing costs (across both compute as well as the tools to read and write DNA) would create a similar dynamic in biology. Biotech would be transformed by broad platforms in the same way that platform business in tech soared to capture about a third of the S&P 500’s total value and usher in an unprecedented wave of innovation.² YAAASSSS!!! As a one-time molecular biologist turned tech investor, this was catnip - I bought deeply into this vision. But as we enter 2025, we are faced with the stark reality of a consistent observation:

Platform businesses in biology generally haven’t worked.

The corollary is even more distressing: the generally accepted business models in biology do not sustain long-term innovation. When we pursue cool science but aren’t realistic about how we’ll make money with that science, the science inevitably dies. The only companies that have built anything close to platforms have actively fought “the system,” have largely eschewed institutional funding, and are headquartered in places like Lebanon, New Hampshire and Danvers, Massachusetts - decidedly not the San Francisco Bay or Kendall Square.

To be clear, I’m not arguing that there’s anything wrong with making drugs or selling services - those are the lifeblood of this industry! But let’s stop pretending that advancing a drug through the clinic means getting to do more science - it means diverting resources to non-scientific endeavors!

To be clear, I’m not arguing that there’s anything wrong with making drugs or selling services - this is the lifeblood of the biotech industry. But I remain deeply concerned about our broad inability to fund and built the infrastructure layer that would improve productivity across the industry. The only practical option for most companies (who generally do need to raise institutional capital in order to fund their efforts) is to vertically integrate and make drugs. But when we think about it for one second, we realize that it is peak insanity to make every niche scientific innovator vertically integrate into everything from target selection to clinical trial design, rather than focusing on their differentiated niche and capturing value with it as broadly as possible.

If we want to take advantage of compounding benefits to technology and learning, we have to solve this dynamic.

I know, I know… Some of you reading this think you are the one true platform that will change this dynamic and I am cheering you on, I swear! But before we get to that, let’s get some grounding…

What is a platform?

We are all deeply familiar with platform business models in technology. Virtually everything we use today is created by platform businesses.

Most tech platforms facilitate some sort of interaction between two or more interdependent groups (e.g. producers and consumers). Marketplaces and social networks are obvious examples of platforms (eBay connects buyers and sellers, Uber connects riders and drivers, LinkedIn connects workers to each other and potential jobs, etc.) but some of the most valuable platforms created over the last couple decades look more like the technical infrastructure and standards on top of which the rest of the world has been built (e.g. AWS and subsequent cloud providers have enabled the entire SaaS industry and the integrations between these applications) and this is typically what “techbio” platforms are aspiring to be - the infrastructure, the tooling. I particularly like the following quote attributed to Bill Gates:

“A platform is when the economic value of everybody that uses it exceeds the value of the company that creates it.” - Bill Gates (according to Chamath Palihapitiya… not someone I’d typically source, but I’ve heard Gates say something similar in other settings, so letting it slide this time)

This manifests in a feature that has contributed more than anything to the inexorable dominance of platform business models over the past 20 years: platforms experience compounding benefits (“flywheels”). Users benefit and use more, platform benefits from growth and gets better, users benefit and use more, etc.

These compounding benefits typically manifest in a few ways, and I consider these “table stakes” for any platform business:

• Network Effects: The platform becomes more valuable the more people use it. (For a truly fantastic primer on network effects, I recommend The Network Effects Bible by NfX, given they named their whole fund after these, they probably know what they’re talking about…). For example:

  • Direct network effects: more users = better product, immediately.

    • Between groups (most common): more buyers = better for sellers (eBay), more riders = better for drivers (Uber), etc.

    • Within a group: more of my friends on Facebook = better experience for me

  • Indirect network effects: more users = better product, eventually. Ride sharing is a perfect example… when there are more riders, I have a worse experience immediately (surge pricing! long waits!) but a better experience in the long term because growth in my side of the network spurs growth in my complements (drivers), yielding a more durable, efficient product for me to use. This is also the type of network effect you see with infrastructure layers (e.g. AWS) - it looks a bit more like scale economies than network effects.

  • Learning effects: the more users a platform gets, the better the platform itself gets (e.g. if I work with the platform, the platform learns from me and gets better for me and for everyone else who uses it).

  

  Jeff Bezos famously drew the Amazon “flywheel” on a napkin. Here he shows that more sellers → better selection → more buyers → sales potential for sellers → more sellers (and on and on it goes).

• Ecosystem/Third-Party Contributions: Platforms generally encourage (or depend on) third-party developers, producers, or users to create complementary products or services on top of the platform. E.g. Microsoft and Apple don’t make every application we use on our devices, they created a platform on which the many developers who do not work for them could create. In other words, there is a level of “openness” inherent in platforms and the platforms get better because of that openness.

  “No matter who you are, most of the smartest people work for someone else.” - Bill Joy (co-founder of Sun Microsystems)

• Distribution / Scalability: platforms tend to be able to scale quickly and get better with scale. This can be driven in part by network effects (rapid growth + more efficient transactions) but, just as importantly, platforms figure out how to very efficiently distribute their product to users and avoid doing the “nonscalable” work (e.g. facilitating the transaction without manufacturing the product being purchased). While established platforms will often expand their interests into more traditional lines of business (e.g. Amazon has a private label line of merchandise now, with the ability to benefit from deep market intelligence and an already scaled distribution infrastructure), they don’t start there.

…and one last thing… when platforms work, they make a f**k ton of money (technical term). If a platform isn’t making a ton of money (let’s define it as gross margin for now…ignoring reinvestment), there is something wrong with one or more of the features above. Where there are real network effects, platforms tend to be “winner take all” and if the market is large enough, they win A LOT.³ While there are valuable, high margin businesses in biotech (namely big pharma), these are clearly not platforms (as evidenced by: a) the relative lack of consolidation, b) the lack of network effects, c) the failure of the “Bill Gates test” - pharma captures essentially all of the economic value created (I’m not getting into the economic value of “health” - it’s too broadly distributed). ⁴

Why haven’t platforms worked in bio?

1. Distribution is too expensive (i.e. it’s not scalable and doesn’t lend itself to third-party collaborations)

   If I write a new piece of computer code, I can ship it to anyone in the world for free and I know it will run on their machine because I can simulate other environments to reliably test new products before I release them.

   By contrast, if I write a new piece of genetic code, invent a new assay, or create a new method for manipulating a biological system, we’re looking at a months-long tech transfer process involving half a dozen PhDs from our teams collaborating to debug the new system in your environment. This creates an inherent scarcity mindset in both the sellers (“well, if I can only afford to do one or two of these collaborations a year, I might as well vertically integrate as much as possible and keep more of the economics for myself!”) AND the buyers (“ugh, is it really worth it to spend the next 6 months validating this tech? there are so many other things I could be looking at…”).

   Collaborating in bio is hard, slow, and expensive. The level of interdependence between systems means there’s much more bespoke vertical integration than true collaboration. We leverage standardized suppliers and service providers for the “easy stuff” but still take full ownership of far more of the value chain than we want to.

   I’d be really excited to see the agentic AI wave working on this problem rather than just discovering more molecules that we won’t be able to test, debug, and distribute anyway.

2. Minimal network effects
   
   The lack of network effects in biology today is, I hope, only a cultural and technical problem, rather than a “physics” problem (i.e. something we could change if we wanted to and not an immutable fact of life), though likely hard to change unless/until we create a viable distribution model that creates economic opportunity in collaborating.

   Biology presents an incredibly rich substrate for network effects, learning effects in particular given the nature of evolution (there’s a lot more shared biology across disciplines than differences).

The bottom line: If we want platforms to work in biology, we need to be able to efficiently distribute biological innovation.

The folks that have gotten closest to this have focused on narrow problems (repeatable / scalable processes) in big markets (e.g. antibodies). It doesn’t (yet) work in areas that are still closer to the cutting edge of science.

The Elusive Promise of Modularity

Modularity is often the key to efficient distribution: I can sell my product to you because our products are designed to work together. While new technologies typically have interdependent architectures as we’re fighting for every ounce of performance gains and don’t even have time to think about cost or efficiency, modularity eventually wins out… again, and again, and again. Every major industry across the world has become widely modularized except biotech.

Modularization allows for much more nimble evolution and faster growth as companies are able to quickly benefit from advances across the component stack. But biotech is stuck. No one expects the fashion designer to be good at harvesting silkworms or managing retail distribution, we want them to focus on what they’re good at: dreaming up fanciful couture. And yet we expect academics who discover an interesting new gene editing system to become good at everything from target selection to clinical trial design if they’re going to capture any value for their innovation.

Part of the issue is misguided valuation. Biotech investors tend to translate “modular” into “commoditized” (low margin) rather than recognizing that in other industries while modularization leads to SOME commoditization across the value chain, value then accrues to the specialized components. Sure, over time, those will get commoditized as well (that’s just physics) and value will accrue to a different component or a newly integrated system, but that is just the natural cycle of innovation and disruption. But today we aren’t ascribing value to those specialized components unless and until they get vertically integrated into a drug.

It’s not as though there aren’t specialists at every stage of the biotech value chain, but that specialized expertise is not widely accessible, it’s not easily distributable, forcing innovators to recreate the wheel thousands of times over, each time risking making potentially avoidable mistakes.⁵ So perhaps more accurately, while we have some modularization, we tend not to have modularization of the specialized components (i.e. the stuff that’s hard to tech transfer) and we force everyone to become a systems integrator (of the often substandard parts that happen to be available for partnership) because it’s too difficult to build a sustainable business as a specialist.

I would be very willing to bet that the vast majority of clinical failures are due NOT to a flaw in the core underlying technology but rather to a mistake somewhere else in the value chain caused by nothing more than a lack of experience.

The Myth of Standards

Modularization typically requires standardized interfaces. I know that your component will work with mine because we have agreed on specifications where we have critical interdependency. While techbio enthusiasts have long talked about biology as the next “engineering discipline,” the reality is that biology is wet, wriggly, squishy, and ever-evolving (even within the timeline of a single experiment). It does not, and likely never will, have truly standardized interfaces. Even when we think we’ve figured something out, you’ll hear a scientist describe something as working “in my hands” because she knows chances are it won’t work in yours (for reasons she can’t really predict and may not understand without months of tinkering).

Some of this is “methods” and I like cooking analogies here… Let’s take my famous french macaron recipe: I have done everything I can to make explicit the secrets to success, but chances are, your first batch of macarons will fall flat (IYKYK).⁶

But some of this is simply the substrate - chemistry and physics are more easily reducible to math: predictable, containable, and controllable. Biology is alive and has a mind of it’s own (figuratively at least, if not literally). The idea that we will someday be able to reduce the wet chaotic mess of interactions in cellular cytoplasm to math is science fiction, and I might go so far as to argue that there is a beauty in accepting that it could/should never occur. There is magic in biological unpredictability.

There have been and continue to be some efforts to create standards in biology, but I fear that these efforts are like filling a leaky bucket. The needs of the field are ever-evolving, technology continuously advances, and customization is still deemed critical.⁷

Many of these efforts have been ongoing for well over a decade and every year they feel like they’re just getting started (no offense intended: it’s certainly not for lack of trying!). The leak in the bucket is just really, really big. As such, I’m most excited about one of the least sexy applications of AI: its ability to act as a translator, obviating the need for ontologies and “standards” in the traditional sense, while still allowing for much greater interoperability and reproducibility of innovation across the value chain.

The companies working here often describe themselves as “AI Scientists” - copilots (or in some cases, pilots) that can engage in both molecular and experimental design, learning from prior art to both generate novel hypotheses and figure out how to test them. Some companies to watch at this interface (A→Z):

• BenchSci: BenchSci’s ASCEND platform seeks to help scientists develop hypotheses and optimize experimental plans by leveraging existing model systems and reagents and highlight known trial design risks.

• FutureHouse: A philanthropically-funded research institute that has been researching and publishing a number of great tools (benchmarks, models, and agents) that may help make scientific knowledge more accessible.

• Lila Sciences: The latest mega-bet by Flagship Pioneering, working towards “Scientific Superintelligence” through an integrated set of AI-powered tools and a physical lab, working to accelerate learning and product development in life and materials sciences.

• Medra: A young new robotics company focused on creating a human language interface and flexible robotic system so that human scientists can engage with lab automation in a natural way, without diving into detailed protocol generation / parameter specification.

• Potato: An early stage startup automating literature review and protocol generation, with a vision to improve scientific reproducibility and get to better methods faster. They’re generating a lot of love from scientists and are catching the attention of large publishers.

• Tetsuwan Scientific: A young new company working to scale the “hypothesis testing” step of autonomous science

Most of these companies are still figuring out their business model (software? services? products? gasp! platform???), but it will no doubt be an exciting area to watch.

Modularity is hungry for data

By now, I thankfully no longer feel as though I’m screaming into the void when I talk about the need for [more/better/multimodal] data. But biology is so multifaceted that it’s really easy to call our bluff…

“We need more data!”

“OK, awesome! I’ve got an automated lab sitting idle overnight… What do you want?”

“Ummmm… Well… It’s hard to know whether…”

The reality is that it’s been very difficult to drive the kinds of cross-functional emergent properties in biology that we saw in language and thus, the community has much less confidence in data scaling laws. We basically all accept that they’re there (duh!), but if you asked anybody what they think the next most useful 100,000 datapoints would be, it’s like trying to pick the right droplet out of a bucket and they freeze (scientists really don’t like having to pick favorites).

Most of the data warriors are focused on generating data about the underlying biology. And this is super useful stuff! This is the language of life, it is what we will train models to read and speak.⁸

Where there hasn’t been nearly as much focus, but which we’ll need if we want any of the “AI Scientists” listed above to work, is generating data about the process of executing science, about the art and “taste” of good science, about what it means to have “good hands”.⁹ The hardest thing here is honestly thinking up creative ways to measure human activities and the ensuing lab results: creepy 24/7 cameras, really robust sensors, you name it…

Until we are able to automate / scale up the debugging process of scientific execution, we’ll continue to be bottlenecked not at hypothesis generation but at experimental validation. This affects both discovery and tech transfer (read: distribution).

The hard thing about hard things…

The idea of platform biotech is seductive: modularity, network effects, scalable innovation… the same dynamics that transformed tech. The reality is that we’re not even close. Platforms in bio don’t fail because the idea is flawed, they fail because the underlying systems are broken. Investors equate platforms with low margin commoditized services and distribution of true innovation is too slow and expensive, forcing every innovator to vertically integrate just to attempt survival.

If we want platform biotech to work, we need to fix this fundamental issue. This isn’t about waiting for the latest “foundation model” to reduce biology to code. It’s about finding ways to distribute biological innovation efficiently, making tools and expertise across the value chain truly accessible, speeding up our tech transfer and debugging processes, and rethinking how we align incentives across the board.

The companies that have built platforms thus far have been forced to focus narrowly: solving specific problems in big markets (and without a lot of the baggage that comes with traditional funding models). Platform biotech isn’t impossible, but it’s a long way from inevitable. If we’re serious about realizing its potential, we need to stop pretending it’s just around the corner and get to work on the messy, unsexy infrastructure that will actually make it possible.

Perhaps the most valuable near-term use case for AI won’t be discovery, it will be tech transfer.

By the way - if you haven’t yet checked out the Database of AI for (Life) Science Companies that I put together, please check it out! It’s got over 300 companies, almost 500 products that they make, and 300 announced partnerships! More features and exciting announcements coming soon! If you want to get involved, shoot me a note on LinkedIn.

Conflict of interest disclosure: I hold equity and/or close relationships with several of the companies mentioned in this post but this post is independent of any work I do with those entities. These perspectives are entirely my own and do not reflect the views of any organization I work with (now or in the past).

1. Around that time, it was still relatively popular to short Amazon, the stock was going up like crazy but free cash flows were declining - it definitely wasn’t a given that it would become wildly profitable and valued over $2 trillion - and “vertical SaaS” was becoming the hot new theme. So while software has undoubtedly eaten the world, it wasn’t a given in 2011.

2. This sentiment is still alive and well. A few of us attending the panel at Recursion’s JPM event very nearly choked on our drinks when a very prominent public market tech investor said something along the lines of: “in tech, we generally like investing in horizontal platforms, that’s where the value is created… why hasn’t anyone tried creating a horizontal platform for biotech yet?” Oh, honey. It’s not like we haven’t tried!!

3. It’s beyond the scope of this post, but Feng Zhu at Harvard Business School has done some great research on why network effects / platforms are often unable to reach that escape velocity, such as: multihoming (I always check BOTH Uber and Lyft before booking and most of my drivers work with both, pushing down their margin), disintermediation risk, clustering, etc.

4. The business models we do have in biotech… could any of these become platforms?

• Big pharma (Lilly, Novo Nordisk, Pfizer, Roche/Genentech, etc): these companies make money by selling drugs to patients. There are no real scale/network effects other than perhaps more robust post-approval data packages, some go to market efficiencies (common to all trad’l distributors), etc. And other than the ripe M&A markets in pharma, it’s famously one of the least “open” and collaborative industries out there. While they do have great gross margins (a function of capitalized R&D and high drug prices relative to manufacturing costs), these are most definitely not winner-take-all businesses with strong network effects. Not platforms… and probably too close to the end product to realistically become platforms

• Equipment and reagent suppliers (Thermo, Danaher, Illumina, Twist, NEB, etc.): I absolutely love all of these companies. BUT, they are vertically integrated traditional product organizations. I would argue that they all pass the Bill Gates test and we do sometimes see some ecosystem-level collaboration in the form of setting standards/interfaces, but even that is pretty minimal. They serve as valuable infrastructure, but there are de minimus network effects and scalability dynamics. Importantly, these folks also generally aren’t at the bleeding edge of innovation, they are distributing products once they have become standard / commoditized (although arguably this could change) and as such they are not typically getting any sort of royalty / “take rate” on the products they enable (compared to Apple’s 30% share of App Store revenue). Not platforms… but honestly probably have some of the best infrastructure to take a real shot at it.

• Contract Research/Development/Manufacturing Organizations (Evotec, Lonza, Charles River Labs, etc.): CROs (and services generally) have historically been a bit of a dirty word in the life sciences - perceived as low-end and low-margin. It’s definitely not where VCs are placing their bets. While it’s true that CROs tend to look more like outsourcers than innovation partners today and do not tend to exhibit network effects or scale economics, I do think this is an area worth watching (as it’s often these “thankless” jobs that become productized and turn into Infrastructure as a Service over time). Not platforms yet, but like with the major suppliers, the infrastructure is there if we decided to take a real shot at this.

• “Next Gen” Biotech/TechBio (Moderna, Recursion, Dyno, BigHat, Mammoth, etc.): These are the companies that are calling themselves “platforms” today and while you do see some collaborations and partnerships, these businesses are pushed by investors (and practical necessity) to become vertically-integrated drug discovery companies. While it’s possible that they have developed technical platforms that will make them more efficient drug developers, they are still limited by the success or failure of their own drug launches (or the very small number of partnerships they strike)… They fail the Bill Gates test in spades. These folks will turn into pharma companies if they find a successful drug. If they don’t, it’ll be a tough road. Who I’m watching: Generate:Biomedicines and Recursion - if they get good clinical hit, they’re both culturally positioned to reinvest in platform. You could imagine some of these companies, if wildly successful build something that looks a bit more like an “operating system” that creates opportunities far beyond what the company itself can prosecute - they may choose to capture the value of a couple key products (e.g. the equivalent of MS Office) but may license their technology out more broadly to the industry and capture a fair take rate on it. But let’s stop calling biotech’s with a snazzy piece of tech in the middle “platforms” - it’s not a platform until you sell it like a platform.

• Software for Bio (e.g. Benchling, TetraScience): This is still a relatively new space and these folks haven’t typically found a way to tap into end-market value (e.g. the “take rates” that are common among marketplaces, the canonical platform businesses of tech) and are limited by a relatively small addressable market (TAM) (see Ankit’s excellent blog on why selling software to R&D is tough). Who I’m watching: TetraScience.

• Scientific Marketplaces (e.g. ScienceExchange, Scientist.com, Zageno): There have been a handful of proper 2+ sided marketplaces pop up in biotech, helping connect service providers / reagent suppliers with end users. That would certainly exhibit network effects! ✅ They’re not tapping into the end product value (e.g. royalties) but they are often tapping into transaction value (e.g. take rate on the services provided) and in theory, this is a very efficient product to scale. I think at the moment, these folks are held back by the distribution challenges inherent in biotech (e.g. even with a facilitator like ScienceExchange, it’s really hard for two entities to collaborate), but if any of these folks can really figure that out, these might become real platforms.

• GenAI for biological design (e.g. Cradle, EvolutionaryScale, Profluent, Inceptive, etc.): Jury is out on whether some of these can become platforms. As they stand now, these are terrific software products and are certainly scalable and partner-oriented, but it remains to be seen whether any of these will be able to develop real network effects, the critical missing ingredient (as of now: their customers are likely making that reasonably hard). My biggest fear for these companies is that these models are becoming commoditized very quickly… the ones who will win will be those who move beyond a flashy model and figure out the workflows and deployment (all the unsexy stuff). Otherwise they become drug companies and then it’s just a dice roll. Who I’m watching: Cradle.

• A note on Ginkgo… if you’re reading this, you may know that I worked at Ginkgo. As a former tech investor, I loved the platform vision. Ginkgo has a unique level of religion around platform potential in biotech. We avoided vertical integration and fought hard to maintain rights to technical platform improvements to achieve learning economies and network effects over time. But building horizontal platforms in bio is still really tough, mainly just a function of the complexity of mastering and integrating so much diversity. I continue to believe deeply in the mission/vision and am cheering them on as they focus on mastering a few core markets.

• 💌 And now…. a love letter to Adimab. Adimab was founded in 2007 by Tillman Gerngross, Karl Dane Wittrup, and Errik Anderson (who has since founded several other platform-minded companies, most recently Alloy Therapeutics) after Gerngross sold his prior company, GlycoFi to Merck (for $400M, the third highest price paid for a private biotech at the time). Rumors are that Gerngross was so dismayed by what happened to GlycoFi after selling to Merck that he developed a bit of religious zealotry about keeping Adimab independent and focused on platform. They are about the least flashy company out there (I was SHOCKED to see Adimab advertisements at JPM!!) and are still private and headquartered in Lebanon, New Hampshire. But the lack of hype does not indicate a lack of accomplishment… Adimab has partnered with over 130 companies on over 600 royalty bearing programs, of which 78 reached the clinic and 4 are now commercial (with 11 more are currently in Phase II/III). While they don’t publish financials, the rumors are that they are printing money (on the order of $100M) and if you consider that royalties are ~100% margin and add up over time, Adimab has the potential to become an extremely profitable and valuable company. Still nothing compared to tech platforms (and herein lies the conundrum: even in this relatively big market of antibody design, the market is a whole lot smaller than tech), but it’s about the closest thing we have (relatively high revenue to employee ratio).

5. Let’s take a look at the lifecycle of a typical platform-minded techbio company:

1. Spin technology out of academic lab and raise some seed capital

2. Find an angel partner or two who are willing to help validate your technology

3. Get a decent data package, present at a conference, raise a modest Series A

4. Scale up BD efforts, realize that BD is HARD (everyone wants exclusivity, it takes a really long time to get to milestones/royalties, no one is willing to pay enough upfront to sustain the R&D you really want to be doing)

5. Create parallel “pipeline” track to complement BD efforts… this requires you to suddenly:

   1. Be good at picking a therapeutic target

   2. Be good at integrating all the other components that will be necessary to turn your niche technology into a drug

6. Raise Series B on the promise of the therapeutic pipeline you have recently created

7. Do your best to execute on pipeline and existing partnerships (forget about new partnerships, you don’t have the bandwidth)… your pipeline requires you to suddenly:

   1. Be good at therapeutic development (in vitro and animal studies)

   2. Be good at evaluating safety, efficacy, etc.

   3. Be good at clinical trial design

8. If you’re lucky, get acquired by large pharma

This is insane. A company that was created based on the promise of a niche technology now needs to be good at an entire range of therapeutic development activities.

What if, instead, each of the required components and skills across the value chain was easily accessible? What if it was trivially easy to test out a new component in your system because we had standardized interfaces or rapid/low cost tech transfer mechanisms? What if the best-in-class methods and assays were published and kept up to date so that it was possible to compare candidates side by side in a more robust way?

6. If you’re just getting to know me and don’t know about my famous macarons, the brief history is that I was kind of bored in my first job out of college so decided to spend six months researching and perfecting French macarons (and found my soulmates who had taken this a few steps farther!). In any event, my recipe which is largely foolproof in my hands was the “official” Michael’s macaron recipe when they did a little French themed bakery endcap one year. Sadly, they no longer have a blog so this is what remains!

Dear Annie Wu, I’ve never met you, but I’m a fan:

Sadly, no macarons for me anymore now that I’ve given up added sucrose/fructose… Although now I’m sort of curious what would happen if I tried to make these with dextrose… Dextrose is hygroscopic so it MIGHT actually work OK… hmmmmmmmm….

7. Folks playing around biology “standards”:

• Consortia / Standards Bodies:

  • Allotrope Foundation: Founded in 2012, Allotrope Foundation is an international consortium of pharmaceutical, biopharmaceutical, and other scientific research-intensive industries that is developing advanced data architecture to transform the acquisition, exchange, and management of laboratory data throughout its complete lifecycle.

  • Pistoia Alliance: Incorporated in 2009 to enable precompetitive collaboration, many of which include or are directly focused on data ontologies and standardization.

  • BioNet, MITRE’s Synthetic Biology Moonshot: Launched at the end of 2023, MITRE’s BioNet is working primarily with academic groups with a long history in the synthetic biology space to develop a common language and ontology for researchers to collaborate.

  • Earlier efforts include the Synthetic Biology Open Language (SBOL) and the BioBricks Foundation which focus on making interchangeable biological components a reality

• Lab Automation Standards: as companies have raced to adopt lab automation in order to try to get scale economics out of experimental execution, companies like TetraScience, Synthace, and Strateos (beware, their web site has been getting attacked) in addition to the next-gen automation companies themselves (e.g. Ginkgo, Automata, and Medra) have offered software/service solutions to help standardize lab protocol creation and translation to robotic standards.

• Standardization in lab notebooks / bioinformatics: companies like Benchling, Dotmatics, Genedata, and TetraScience provide tools for coherent data models and interoperable systems, ensuring that experiment results (and associated metadata) can flow seamlessly between instruments, LIMS/ELN systems, and analytics tools.

8. Some cool companies to watch here (again, A→Z, not playing favorites!) are:

• A-Alpha Bio: protein-protein interaction data at scale

• Basecamp Research: literally scaling mountains and diving under ice sheets and stuff to collect and learn from novel biology, cool stuff!

• bitBiome: clever approaches (both biological and commercial) to accumulating novel biological sequences

• Digital Biology: proprietary tech links cellular and tissue phenotypes to underlying molecular structures, allowing for functional screening of genetically encoded systems at scale

• Ginkgo Bioworks: large automated lab that can generate huge amounts of lab data (e.g. functional insights) at scale

• Ultima Genomics: ultra high throughput sequencing

• Zafrens: bead-based ultra high throughput phenotypic and molecular screening technology

9. There are stories of pharma execs literally flying around the world to find the “magic hands” within their companies - those elusive scientists who can actually reliably reproduce experiments - that is CRAZY!