The Solicitation
On December 19, 2025, DARPA’s Biological Technologies Office published a program solicitation — DARPA-PS-26-10, Generative Optogenetics (GO) — to build what it calls a Nucleic Acid Compiler (NAC): a protein complex, expressed inside a living cell, that synthesizes DNA or RNA sequences template-free, using light as the sole carrier of genetic information. The program’s own framing of the gap it intends to close:
No existing technology enables massless information transfer to relay genetic instructions to living cells. All current approaches require some mechanism predicated on moving matter that encodes the genetic information, typically DNA or RNA nucleic acids, across biological barriers like a cell wall/membrane.
The NAC is the technological write-head for the channel described in The Serpent Channel. Where the serpent channel is the natural read/write interface into the living boundary — the one Narby’s shamans tuned with the pharmakon, the one Levin operates with a voltage clamp — the NAC is the engineered one, designed to spec, built to write arbitrary sequence on optical command.
What the Specification Requires
Every current method of genetic modification requires matter crossing the cell membrane: viral vectors, lipid nanoparticles, electroporation, CRISPR ribonucleoprotein delivery. GO eliminates the matter requirement at the point of writing. The information travels as photons. The cell, once expressing the NAC, becomes writable by light.
The solicitation specifies optogenetic domains responding to four distinct wavelengths with minimal (≤1%) co-activation — one optical channel per nucleotide base. The final milestone (month 41 of a 42-month program) targets synthesis of a 3,000-base-pair sequence of arbitrary content, at one base per second, with an error rate below one in a million (1×10⁻⁶ per base).
The specification that changes the architecture is not the length or the fidelity. It is this, from the month-41 milestone, in the program’s own words:
All GO performers must demonstrate the capability to write two distinct nucleic acid sequences to the same cell or population of cells, with less than 1 hour required to reset the system between write events.
This is not a one-shot genetic edit. It is an ongoing programming interface. The cell becomes a receiver for a continuous, reprogrammable stream of genetic instructions delivered optically. The NAC is a permanent door installed in the boundary that opens to any optical signal carrying the correct encoding.
The Bootstrap Problem
The NAC itself must be delivered to the cell through conventional means — the protein complex’s genetic sequence has to enter the cell before the cell can express it. This initial installation still requires matter: a viral vector, a lipid nanoparticle, a transfection event. But once installed, every subsequent write is massless. The first delivery is physical. Every reprogramming after that is light.
This creates a two-stage architecture: installation (matter-dependent, one-time) and operation (massless, indefinitely repeatable). The delivery infrastructure for the first stage already exists. The mRNA-LNP platform demonstrated during the pandemic response proved that lipid nanoparticle delivery can reach billions of cells across a global population in months. Pandemic preparedness infrastructure — manufacturing, cold-chain logistics, regulatory emergency-use pathways, population-scale injection campaigns — is, in capability terms, an installation pipeline. The GO program supplies what would be installed.
This linkage is not stated in the solicitation, and the GO program proposes no such application. It is a structural observation about two capabilities that DARPA and adjacent agencies have funded separately: a delivery platform (mRNA-LNP, already scaled) and a write-head (NAC, now under development). The serpent channel’s two-stage architecture — install the port, then write through it — maps onto what exists and what is being built. The map is not the claim. The claim is only that the two stages are no longer hypothetical in isolation.
The Natural Port versus the Engineered Port
The Serpent Channel describes three natural conditions that open the port: developmental plasticity (embryogenesis, when the morphogenetic field is most labile), boundary breach (trauma), and pharmacological perturbation (the pharmakon). All three require the target’s own boundary to be compromised. The natural port opens because the self cannot hold its address space — the integration window is compressed below the threshold of boundary defense.
The NAC bypasses this entirely. It does not need the port to be naturally open. It does not need the target’s window compressed, the vessel shattered, or the boundary breached. It installs its own port. The cell’s boundary remains intact — but it now contains an engineered receptor that transduces optical signals into genetic sequence regardless of the cell’s developmental state, coherence level, or boundary integrity.
This changes the threat model. The natural write-side requires a degraded target — a self too small to defend its own address space. The engineered write-side requires only that the target express the NAC. The question shifts from “how open is the port?” to “has the port been installed?”
A necessary boundary on the speculation: the GO program explicitly excludes the developmental and human contexts the natural-port discussion invokes. The solicitation’s out-of-scope section prohibits NAC development in embryonic stem cells, and states plainly that all human-subjects and animal-subjects research is out of scope, as are export-controlled and biosafety-restricted genes. The engineered port, as funded, is a proof-of-concept in immortalized cell lines. The write-access-at-conception scenario is the natural analogue the serpent-channel thesis theorizes — not what this program builds. The two should never be conflated, and the exclusions are the reason why.
Bivalence Confirmed by Engineering Specification
The bivalence theorem holds that a γ-neutral channel will heal or harvest depending only on the alignment of the whole the signal serves. The GO program confirms this at the level of engineering specification. The NAC writes whatever optical sequence it receives. It will synthesize a tumor-suppressor gene or a toxin-producing gene through the same mechanism. The molecular machine has no preference. The bivalence is not a design flaw — it is the specification.
DARPA’s own response to this neutrality is governance. The solicitation stands up a Biosecurity Working Group, tasked — verbatim — with developing “policy recommendations against accidental and intentional misuse of GO technologies,” and a Regulatory Policy Working Group to address the “implications of reprogrammable, decentralized, and in situ manufacturing of biological molecules.” The agency builds the γ-neutral channel and, in the same document, convenes external bodies to supply the alignment the channel lacks. This is the icaro at civilizational scale: an external boundary wrapped around a boundaryless interface, because the interface cannot sort on its own.
The agency goes further, and the detail is the thesis in DARPA’s own hand. The Biosecurity Working Group is tasked with “a preliminary assessment of the cybersecurity risks associated with the development of devices incorporating GO technology, to include software and hardware.” A write-head is an attack surface. The agency anticipates that an optical port into the cell is, like any port, something that can be addressed by a signal its operator did not author. That is the write-side stated as a security requirement.
Classification of the Write-Head
Phase 2 of the program classifies NAC-related work as Controlled Unclassified Information (CUI), on the stated logic that “a functional NAC will have similar metrics to export controlled synthesizers.” The protection scope covers “all aspects of NAC sequence development, design, optimization, and integration into living cells, with strict prohibitions on public disclosure or publication of the NAC sequence.” The genetic blueprint of the write-head becomes a controlled technology whose distribution the state regulates.
The asymmetry this creates: whoever controls the NAC sequence controls who can build write-heads. The channel itself is natural and universal — every cell with DNA carries the substrate. The read-side stays open; anyone with the pharmakon or the Work can tune in. The engineered write-side becomes a state monopoly. The serpent was always available to everyone. The instrument that writes through it on command will not be.
The Optical Environment as Programming Medium
If the write-channel operates through light, the optical environment becomes, in principle, a programming medium for any cell expressing the transduction mechanism. The solicitation names “single-cell spatial resolution” and “remote, scalable dissemination of genetic instructions” as program goals.
The constraint that keeps this from sliding into paranoia is the specification itself. The NAC requires four engineered optogenetic domains discriminating four precise wavelengths at ≤1% cross-activation — an extraordinarily specific coded incorporation signal. The ambient optical environment humans inhabit — LED lighting, screen emissions, fiber-optic infrastructure — carries nothing resembling that encoding. Broadband light is not a write signal any more than radio static is a sentence. The honest threat model is narrower and, for being narrow, sharper: the danger is a purpose-built emitter delivering the encoding to cells that already express the NAC — not the lamp on the desk.
What the GO program changes is the standing of the question. Until the transduction mechanism exists, “the optical environment writes DNA” is a category error. Once it exists, the gap between the optical environment and the genetic substrate is closed for any cell carrying the receiver, and the question becomes engineering rather than ontology: who builds the emitter, who installed the receiver, and what does the encoded sequence serve.
Why It Belongs in the Frame
The engineered port does not prove the serpent-channel thesis. The thesis stands on the natural channel — the universal bivalence, the read-side Narby documented, the bioelectric write-side Levin demonstrated. What the GO program supplies is a confirmation from the opposite direction: an institution with no interest in esoteric cosmology, working from pure engineering motivation, has independently arrived at the architecture the thesis describes. A protein that converts light into genetic sequence inside a living cell is a write-head for the boundary. The agency building it has named the bivalence (misuse governance), named the attack surface (device cybersecurity), and classified the blueprint (state monopoly on the write-side). Different cartographer, same territory — and this one filed a solicitation number.
Source
- DARPA Biological Technologies Office. “Generative Optogenetics (GO),” Program Solicitation DARPA-PS-26-10, December 19, 2025. Solicitation for a Nucleic Acid Compiler (NAC) enabling template-free, optically-controlled DNA/RNA synthesis in living cells. All quoted specifications drawn from §1.1 (Background), §1.3.1 (Overall Scope), §1.3.4 (Program Constraints / Out of Scope), §1.3.6 (Program Security), §1.3.7 (Program Metrics), and §1.5 (Advisory and Working Groups).