For a century we found drugs. Now we are learning to write them. Why the shift to programmable, modular, personalized therapy is the defining change in medicine.
By Vincent T. Pham Cancer Biology PhD Candidate, Huang Lab · Co-Founder and CEO, Powerhouse Therapeutics For a century, we found drugs. Now we are starting to write them.
That single shift, from discovery to authorship, is the most consequential change in medicine I expect to witness. The therapy is no longer a lucky find buried in a chemical library. It is a design we compose, edit, and compile. In other words, it is becoming software.
Classical drug development was prospecting. We screened vast libraries of small molecules against a target and hoped one would bind. We should no longer accept this premise. Prospecting for a single small-molecule hit within an effectively unsearchable space of possible combinations is a brute-force search, not a design process. Even our largest screens sample a vanishing fraction of what exists.
Success in that regime depends on luck nearly as much as on insight. Each new disease restarts the hunt almost from zero, because a molecule that worked for one target rarely transfers to the next. This approach gave us extraordinary medicines. It also gave us brutal economics, long timelines, and a failure rate that humbles everyone in the field. The core problem was structural. We were searching chemical space, not engineering a solution within it.
Cell and gene therapies break that pattern. We do not find them; we build them. Their active ingredient is information, and information is written in text. DNA and RNA are sequences, and a sequence is just a string of characters we can read, edit, and rewrite.
This is the heart of the thesis. The medium of the next generation of medicine is the same medium a developer uses to write code. We open a text editor, we specify a sequence, and we hand it to biology to execute. The keyboard has become a therapeutic instrument.
The central dogma is a toolchain we did not design but can now exploit. We write instructions in nucleotides. The cell transcribes them, translates them, and runs the resulting program. DNA is the source. RNA is the build step. Protein is the executing process. Our job is to author code the cell will run faithfully.
Once you see the cell as a runtime, the engineering questions become familiar. How do we deliver the program? How do we control when and where it executes? How do we version it, test it, and patch it when something fails? These are software questions wearing biological clothing.
The most important consequence of writing therapies in text is modularity. When a therapy is code, you can decompose it into parts and recombine them. You build a targeting component, a component that reshapes the local environment, and a payload that does the work. Each is a module with a defined interface.
This is exactly how we think about our IT-ACTS platform. We treat it as an operating system rather than a single drug. The targeting module decides which cells the therapy engages. Other modules decide what happens once it arrives. To address a new indication, organ, or organ system, we swap modules rather than restart discovery.
A platform built this way yields many programs from shared parts. The same core can be retargeted from one disease to another by changing a single component. That is reuse, and reuse is the quiet superpower that let software eat the world.
The deepest implication of writing therapies is that we can write a different one for every patient. The small-molecule era forced a single compound onto millions of bodies and accepted that it would fail in most of them. We tuned the dose and hoped the average held. Software does not work that way, and neither should medicine.
When a therapy is code, the patient becomes the specification. We read a tumor, a genome, or an immune repertoire, and we compile a build that fits that person. The platform stays constant while the parameters change. This is not a luxury layer added at the end. It is the natural behavior of a system whose source is text.
Our lead program makes the point concretely. An autologous cell therapy takes a patient's own cells, rewrites them, and returns them as a treatment authored for that individual. The same logic extends to patient-specific neoantigens, to corrections aimed at a single mutation, and eventually to therapies built for one person who shares a disease with almost no one else. The economics of one-size-fits-all collapse when the marginal cost of a new build keeps falling.
Personalized medicine has been a slogan for decades. The reason it stayed mostly aspirational is that the small-molecule model could not deliver it. You cannot economically design a unique molecule per patient. You can, however, recompile a programmable therapy per patient, because the design work is captured once in the platform and reused on demand.
When medicine behaves like software, the unit of value changes. You stop betting everything on one molecule and start building a codebase. Each program you ship strengthens the platform underneath it. Each failure teaches the system, rather than ending it.
This reframes how we should fund, build, and evaluate biotech companies. An asset is a lottery ticket. A platform is an engineering organization that produces tickets on demand, each one faster and cheaper than the last. We are building Powerhouse on that second premise.
I am not claiming biology is as tractable as code, and honesty matters here. The cell is a noisy runtime we still only partly understand. Delivery remains hard. Manufacturing is unforgiving. Immunogenicity, off-target activity, and biological context can defeat a perfectly written sequence. Personalization sharpens these problems, because a therapy made for one patient must still be manufactured, tested, and trusted at the speed that patient needs.
The abstraction leaks, and in places it leaks badly. Yet every powerful abstraction in computing leaked too, and engineers shipped anyway. They built tools to manage the mess instead of waiting for it to disappear. We will do the same in biology, because the direction of travel is already set.
The scientists who define the coming century will treat biology as an engineering discipline. They will think in modules, interfaces, versions, and platforms. They will move fluently between a sequence file and a clinical hypothesis, because both now belong to the same craft.
We are building that way at the Huang Lab and at Powerhouse. We are not prospecting for the next molecule. We are writing the software of human biology, one programmable therapy at a time.
The views expressed in Perspectives are those of the author and are intended to spark discussion. They do not constitute medical advice.