When Jeff Liu, Li-Yao Huang, and Steven Lin step onto the stage for Y Combinator's Demo Day on June 16, 2026, they'll be pitching what might be the most audacious claim in biotechnology: the world's first programmable drug to cure all cancers.
FinalDose—incorporated in October 2025 in the UK with fewer than ten employees—has published no peer-reviewed papers. Filed no patents. Disclosed no animal data. What it does have: a landing page describing a "programmable cell elimination platform," LinkedIn profiles announcing the three Oxford PhDs are "on a mission," and the kind of outsized ambition that either heralds breakthrough innovation or spectacular flameout.
The timing, though, is what makes this worth watching.
Because while FinalDose exists mostly in narrative space, the science it's presumably building on has started to migrate from academic whiteboards into actual human bodies. Logic-gated therapeutics—drugs that execute Boolean operations on biological inputs before deciding whether to activate—are no longer theoretical. They're in Phase 1 trials. And they're targeting something the pharmaceutical industry has struggled with for decades: the roughly 80 to 90 percent of disease-driving proteins that conventional drugs simply cannot touch.
The Undruggable Proteome
That statistic gets thrown around in drug discovery circles with the weariness of a long-standing defeat. Most of the human proteome lacks the structural features—binding pockets, enzymatic grooves, accessible active sites—that small molecules need to latch onto. Cancer's most notorious villains, proteins like MYC and certain forms of KRAS, have historically laughed at pharmaceutical intervention. Transcription factors, those master regulators of gene expression, remain stubbornly out of reach for traditional chemistry.
Which is where programmability enters the conversation.
In March 2026, Nature Biotechnology published something intriguing: a proof-of-concept using DNA logic gates to deliver drug payloads only when two specific cell-surface biomarkers are present. Think of it as an AND operation—both conditions must be met before the therapy activates. When they are, a hybridization chain reaction amplifies drug delivery by over tenfold. A week later, Nature Reviews Bioengineering called it a potential landmark in selective drug delivery.
The academic work remains firmly in the preclinical realm—meaning mice, not people. But it signals a shift that's been building across the sector: therapeutics are becoming conditional. Programmable. Smart, if you're inclined toward marketing speak.
From Theory to Flesh
The field has moved fastest where the stakes are highest and the engineering most mature: cell therapy.
Senti Biosciences announced in April that it had completed Phase 1 enrollment for SENTI-202, a CAR-NK therapy engineered with what amounts to biological circuitry. The construct uses OR logic to target cells expressing either CD33 or FLT3—markers common in acute myeloid leukemia—while a NOT gate tells it to stand down if it encounters EMCN, a protein found on healthy blood vessel lining. Early data presented at major oncology conferences in late 2025 suggested the system could actually distinguish leukemic blasts from normal tissue in heavily pretreated patients.
It's still a living cell therapy—complex, expensive, requiring patient-specific manufacturing. But it demonstrates that Boolean logic can function in humans, not just in silico.
Strand Therapeutics has taken a different path, embedding regulatory circuits into circular RNA that drives CAR or cytokine expression only when cells meet predefined criteria: tumor hypoxia, specific immune signatures, engineered kill switches. The company raised $153 million in Series B funding in August 2025, betting that programmable mRNA could finally solve the toxicity problem that's plagued systemic interleukin-12 delivery for decades. STX-001, its lead asset expressing IL-12, showed initial Phase 1/2 data at the American Society of Clinical Oncology meeting last year—local expression in tumors without the dose-limiting cytokine storms that have ended previous attempts.
Sample sizes remain small. The data is early-stage, the kind where every dose cohort gets scrutinized for signals that might vanish with scale. But the concept has moved from speculation to patient enrollment, which in biotech represents crossing a meaningful chasm.
The academic DNA-logic paper from March 2026 operates in a narrower technical space but with more scientific rigor. Researchers demonstrated AND-gated drug release in cell culture and mouse xenografts, showing selective killing of cells expressing both HER2 and EGFR while sparing cells positive for only one marker. The hybridization chain reaction—triggered when both aptamers bind their targets—polymerizes DNA scaffolds carrying drug payloads, amplifying output by over a hundredfold in some models.
It's modular, theoretically. Different aptamers could retarget the system; different payloads could be swapped in. The challenges? Delivery, stability, immunogenicity. DNA nanostructures clear rapidly from circulation. And despite decades of preclinical promise, no aptamer-based cancer therapy has yet secured regulatory approval—a fact noted in a pharmacology journal review earlier this year with what one might charitably call understatement.
What's Driving This
Three forces are converging to make programmable therapeutics more than an academic curiosity.
First: the technical infrastructure to design, test, and manufacture sequence-specific biologics at scale now exists. Comprehensive tumor profiling has become routine enough that the FDA approved Illumina's TruSight Oncology Comprehensive panel in late 2024, which sequences 517 DNA genes and 25 RNA fusions in solid tumors. That depth of molecular characterization provides the biomarker inputs that logic-gated systems require to make their conditional decisions.
Second: regulators are signaling unexpected flexibility. In January, the FDA issued draft guidance endorsing Bayesian primary analyses in clinical trials, potentially shrinking sample sizes for adaptive designs that could suit rare genomic subsets. A month later, the agency launched its "Plausible Mechanism Pathway" for individualized ultra-rare therapies, explicitly including genome editing and antisense oligonucleotides.
The European Medicines Agency followed suit with draft oligonucleotide guidelines consulted through early 2025 and expected to finalize this year. The regulatory apparatus, often portrayed as innovation's bottleneck, appears to be leaning forward.
Third: the therapeutic toolbox itself has expanded considerably. Antibody-drug conjugates now anchor roughly a third of oncology trials involving novel modalities, according to IQVIA data. ADCs demonstrate that targeted payload delivery can work—though they remain constrained by their dependence on surface proteins and internalization pathways. Oligonucleotide therapeutics, estimated by market researchers to have grown from nearly $6 billion in 2024 to over $7 billion in 2025, can target intracellular RNA. DNA logic gates add conditional execution to that reach.
The American Cancer Society reported in January that 70 percent of US cancer patients now survive at least five years—a milestone reflecting decades of incremental progress across surgery, radiation, and targeted therapy. The next 30 percent, the thinking goes, may require something fundamentally different. Whether that's programmable therapeutics or something else entirely remains an open question.
Recent FDA oncology approvals underscore how far precision medicine has already traveled. In February, the agency cleared zongertinib for HER2-mutant non-small cell lung cancer as initial treatment. March brought relacorilant combined with chemotherapy for platinum-resistant ovarian cancer. In May 2026, vepdegestrant—an oral selective estrogen receptor degrader—gained approval for ESR1-mutated breast cancer.
Each targets a specific molecular alteration. Logic-gated systems propose to extend that specificity: requiring multiple biomarkers simultaneously, or excluding normal-tissue markers with NOT gates, or both.
The FinalDose Question
Which brings us back to three Oxford PhDs preparing to pitch a cure for all cancers.
Liu (CEO), Huang (chief scientific officer), and Lin (chief technology officer) hold doctorates in oncology, biochemistry, and computational biology respectively. They have not disclosed their platform's mechanism, delivery modality, or lead indication beyond lung cancer imagery scattered across their website. The company lists no technical details publicly. Its tagline—"Search.Destroy"—echoes the Boolean logic framing now standard in the field, but without data, FinalDose exists as narrative rather than science.
Perhaps that's tactical. Stealth mode protects intellectual property. Demo Day pitches are designed to secure Series A funding, not pass peer review.
But it sets up a credibility gap that grows wider with every comparable company's progress. Senti has human data. Strand has a funded pipeline and clinical readouts. The academic DNA-logic teams have published their methodologies, opened themselves to replication attempts and critique—the slow, grinding machinery of scientific validation.
FinalDose has a landing page.
For the startup to transition from YC participant to credible therapeutic developer, it will need to clear several thresholds—perhaps more than its founders anticipated given the bold framing.
First: proof-of-concept in a disease-relevant model. Does the platform selectively kill cancer cells in vivo, or just in carefully controlled in vitro assays where conditions can be optimized until something works?
Second: a defined mechanism. "Programmable" encompasses DNA logic gates, CRISPR-based circuits, mRNA regulatory programs, small-molecule conjugates—modalities that face wildly different regulatory pathways, manufacturing challenges, and safety profiles. Which one is FinalDose actually pursuing?
Third: a lead indication with clinical and commercial logic behind it. "Lung cancer," hinted at through website imagery, encompasses EGFR-mutant disease, ALK rearrangements, KRAS G12C, PD-L1 high tumors—molecularly distinct entities requiring different therapeutic strategies. The choice matters.
Fourth: preclinical toxicology demonstrating that the logic gates actually gate. That the therapy doesn't activate in liver, heart, or bone marrow. That the NOT gates hold when they're supposed to hold.
The FDA's April draft guidance on genome editing safety assessment raises the bar for any platform built on genetic circuits. The agency now expects next-generation sequencing to characterize off-target effects and chromosomal integrity, particularly for agents that modify DNA or integrate into genomes. FinalDose's team includes a computational biologist, which suggests some capacity for the bioinformatics-heavy safety analysis modern regulators demand. But whether the platform involves genome editing, transient nucleic acid delivery, or something else entirely remains undisclosed.
The Investor Calculus
For the investors who will watch Demo Day—and the more sophisticated ones who will request follow-up meetings—the risk-reward equation hinges on modality validation timelines.
Logic-gated cell therapies have reached Phase 1. RNA circuits have early clinical data. DNA-aptamer drug conjugates remain preclinical. If FinalDose's platform sits in that latter category, it faces a five-to-seven-year path to first-in-human studies, during which capital requirements escalate and competing approaches advance. If it has a genuinely differentiated delivery mechanism—novel nanoparticle formulation, covalent tissue targeting, activation chemistry not yet described in literature—that could compress timelines.
Nothing public suggests such a mechanism exists.
The broader field, meanwhile, grapples with three challenges that don't yield easily to engineering elegance.
Delivery to solid tumors remains genuinely difficult. DNA nanostructures and large RNA constructs don't penetrate dense stromal tissue effectively—a limitation demonstrated across multiple studies using tumor spheroid models throughout 2025. Getting the therapeutic to the cancer cell is step one. Having it activate correctly is step two. Both must work.
Off-target activation can occur even with multi-input AND gates if biomarker expression overlaps between tumor and normal tissue at unexpected thresholds. Biology is messy. Protein expression exists on continuums, not binary switches. Logic gates impose digital logic on analog systems, and the edge cases can kill patients.
Dosing optimization, now mandated under the FDA's Project Optimus framework finalized in 2024, requires demonstrating that logic-gate performance holds across dose ranges—a characterization exercise that's anything but trivial when your therapeutic's behavior depends on cellular context and biomarker density.
The Broader Momentum
Yet dismissing the entire sector would be premature.
A January report from Boston Consulting Group noted that investment in novel modalities persists despite broader biotech market headwinds—a signal that sophisticated capital sees something worth pursuing. Prime Medicine's first-in-human prime editing data, reported across the latter half of 2025, showed restored protein function in chronic granulomatous disease. The company later deprioritized that program for business reasons, not scientific failure—a reminder that technical feasibility doesn't guarantee commercial viability, but also that the editing itself worked.
The oligonucleotide market continues to grow, regulatory bodies are adapting their frameworks, and the clinical precedents keep accumulating. Senti's trials are enrolling. Strand has a funded development pipeline. Base editing programs are advancing toward the clinic.
Programmability is no longer speculative. The scientific question has shifted from whether it's possible to which architectures will prove safe, scalable, and efficacious in patients with sufficient consistency to justify regulatory approval and payer reimbursement.
FinalDose enters this landscape with meaningful advantages: Oxford pedigree provides technical credibility, Y Combinator backing offers capital access and network effects. But its claims—"curing all cancers"—set expectations the company almost certainly cannot meet. Even the most optimistic projections in logic-gated therapeutics envision subset targeting, not pan-cancer solutions.
The gap between those two realities will determine whether FinalDose becomes a cautionary tale or a case study in successful translation. Demo Day presentations are performances, engineered to generate excitement and open wallets. They're not scientific disclosures.
The data, when it arrives—if it arrives—will matter more than the pitch. June 16 will come and go. What happens afterward is the story worth tracking.
