Nucleic Acid Compiler (NAC)
conceptAI & Compute · Biotech & Transhumanism · Defense & Military-Industrial
This is a plan for rewriting a person's genes with beams of light, no needle required.
Who they are
A goal of a US military research agency (DARPA), under a program called GO[1].
What they do
It's a package of engineered proteins built inside living cells that react to specific colors of light and, on cue, build new DNA or RNA from scratch.
How it works
Instead of shipping gene edits into the body using viruses or fatty bubbles, an operator would simply shine patterns of light into tissue; the light-triggered protein machinery inside the cells would then write and run new genetic code.
Why it matters
If it works, changing someone's genes becomes as easy and remote as pointing a light at them, effectively a read-and-write interface to the human genome.
The engine's record — word for word
DARPAs GO program goal. Complex of engineered proteins expressed within living cells that respond to specific optical signals (wavelengths of light) to synthesize DNA/RNA template-free. Eliminates physical delivery bottlenecks (viral vectors, lipid nanoparticles). Enables massless transfer of genetic information. If deployed, remote operator beams light patterns into tissues prompting in-cell NAC to compile and execute new genetic code. The read/write interface to the human genome.
Report #179: the NAC is the modern payload compiler — 'massless transfer of genetic information... template-free mechanism that incorporates nucleotide bases in response to specific optical signals' (solicitation verbatim). Code decoupled from mass is the funded frontier, not theory; the seeded-payload thesis's 'code can travel' sub-claim is current state biotechnology.
Report #183: the compiler is no longer aspirational at the protein layer — AlphaFold 3 (Nature, May 2024) is a diffusion architecture predicting 'the joint structure of complexes including proteins, nucleic acids, small molecules, ions and modified residues'; RFdiffusion (Nature, Aug 2023) generates novel protein backbones from noise in SE(3)-equivariant space; ProteinMPNN (Science, 2022) solves inverse folding (backbone → sequence) at 52.4% recovery vs Rosetta's 32.9%. Beyond structure: genome-scale models now write and read at organism scale (Evo 2, Nature 2026; ESM3, Science 2025). AND THE CHOKE POINT MOVED: US nucleic-acid synthesis screening, mandated via EO 14110 funding requirements, was COUNTERMANDED by EO 14292 in mid-2025 — screening is voluntary again (Front. Bioeng. Biotechnol. 2026), leaving IBBIS's open-source screener as the residual infrastructure. Every capability on this thread runs through who checks the order. [AI & Bio-Selection Audit] Two present-day instances now sit under the compiler: arc_institute_evo2 (genome-scale read/write) and opencrispr_profluent (an AI-designed editor that is 98% a splice of three natural Cas9s). Governance is the tell: the federal nucleic-acid-synthesis screening mandate (EO 14110 §4.4) was BUILT then KILLED — EO 14110 was revoked by EO 14148 on Jan 20 2025
[2], so screening reverted to voluntary industry standards (IGSC Harmonized Screening Protocol). Present rules on paper are not a functioning constraint (the report's own failure-mode #6). Held; name no holder.
Follow the trail
Walk this on the live map →