A desktop rig + AI compiler that programmably manufactures atomically-precise 3D nanostructures by DNA self-assembly — built entirely from parts you can order online or repurpose from non-lab gear.
You describe a shape; the compiler emits the DNA scaffold + staple sequences to order and the step-by-step wet-lab recipe; a ~$623 rig (repurposed Peltier thermocycler + DIY gel) folds it and checks it; an optional sol–gel step hardens the DNA into rigid silica.
The output is a DNA-self-assembled nanostructure (~10 nm – ~1 µm), optionally hardened to silica/metal — NOT a macroscopic object, and NOT Drexler diamondoid mechanosynthesis. Every capability claimed on the buildable track cites a real, experimentally-demonstrated paper (
docs/science.md,docs/claims.json). The bigger ambition (positional covalent assembly, macroscopic atomically-precise manufacturing) is real-as-a-goal but not yet demonstrated, so it lives, clearly labelled and simulated-only, on the north-star track and is never costed into the BOM.
This is genuine, demonstrated bottom-up nanofabrication — the real first rung of the ladder — described without overclaiming the rungs above it.
🧬 The bigger picture — a molecular synthesizer. The dream is a matter compiler: describe a structure, a machine builds it with atomic precision. A universal replicator (arbitrary macroscopic matter from atoms) is north-star — not possible today, any price. But the bottom rung is real now: programmable, sequence-addressed self-assembly of atomically-defined nanostructures by DNA origami — that's what this repo compiles and folds. From there, a real research staircase (functionalisation → DNA-templated synthesis → molecular robots → assemblers) climbs toward molecular manufacturing, every rung demonstrated in a lab. See
docs/vision.mdanddocs/the-ladder.md.cd compiler python -m molsynth compile octahedron --out out/octa # shape -> scaffold + staples + protocol + 3D
🖥️ Live site: the interactive blueprint — the nanofactory, its live parts list, and the research ladder, in the browser. Source:
web/.
🔧 Building it for real? Start with the BUILD GUIDE — the full blueprint from an empty bench and $623 to a folded, gel-verified nanostructure, including the critical step of validating your rig on a known published design before trusting novel compiled ones.
# from the repo root
cd compiler
python -m molsynth compile tetrahedron --out out/tetraYou get, in out/tetra/:
| file | what it is |
|---|---|
scaffold.fasta |
the scaffold route to fold (M13mp18, 7249 nt) |
staples.csv |
the staple oligos to order (sequence, well, Tm, GC, crossovers) |
staples_idt_plate.txt |
IDT plate bulk-upload (addressable, ~$750) |
staples_opool.txt |
IDT oPool bulk-upload (pooled, ~$200) |
design.json |
full machine-readable design |
design.top + conf.dat |
3D oxDNA structure — view in oxView, relax/simulate in oxDNA |
structure.pdb |
coarse 3D structure — open in PyMOL/ChimeraX/Mol* |
protocol.md |
the auto-emitted wet-lab recipe for this design |
diagnostics.md |
predicted-yield report (Tm histogram, crossover balance) |
Other shapes: cube, octahedron, square, any .stl / .ply mesh, or a .json
wireframe (see examples/square_pyramid.json). It also
emits screen.md (the Mg²⁺ × ramp folding screen) and oxdna_min.input /
oxdna_relax.input (ready-to-run oxDNA relaxation).
Optional extras: pip install -r compiler/requirements.txt then
python -m molsynth fetch-scaffold to pull the real M13mp18 (already cached here),
and to enable scadnano/STL/Biopython integrations.
shape (preset / STL / JSON)
│ /compiler (this repo, pure-Python)
▼
mesh → scaffold routing (searched A-trail: min vertex crossings, Benson 2015 / Veneziano 2016)
→ staple breaking → AI yield optimizer (Tm balance + loop-closure economy, Aksel 2024)
▼
scaffold.fasta + staples.csv ──order──▶ M13mp18 scaffold + staple oligos (IDT/NEB)
protocol.md ──run────▶ /hardware (repurposed Peltier+PID thermocycler)
▼ fold in 1× TAE + 12.5 mM MgCl₂, slow-cool anneal
verify ──▶ DIY agarose gel (tight folded band) [optional: harden → silica; image atoms via DIY STM]
| path | contents |
|---|---|
compiler/ |
the shape→recipe compiler (Python). Routing, SantaLucia Tm model, the AI yield optimizer, exporters, CLI. Runs end-to-end on stdlib. |
hardware/ |
parametric CAD (*.scad: gel box, thermocycler block, STM mount) + Arduino thermocycler firmware + host ramp streamer. |
protocol/ |
how the per-design wet-lab recipe is auto-emitted, + a reference. |
bom/ |
the live, linked, priced Bill of Materials (bom.json + bom.md). |
docs/ |
build-guide.md (the end-to-end blueprint), integration.md (how every cog fits into one real machine), vision.md (the molecular-synthesizer vision), the-ladder.md (the demonstrated research staircase from DNA toward molecular manufacturing), science.md (a paper per claim), claims.json, north-star.md, and the research/ dossiers. |
validate/ |
the gate: mechanically checks every BOM line is orderable < $1500 and every claim is demonstrated. |
proofs/ |
would it work IRL, given the parts? — runnable proofs with evidence: staples tile + hybridise, Tm matches Biopython (independent tool), the 3D structure is valid oxDNA, a simulated PID tracks the fold ramp, power/fluidics close. |
northstar/ |
the geometry-only diamondoid simulation, clearly labelled not-yet-buildable. |
research/ |
measured, reproducible experiments + FINDINGS.md — independent Tm validation (vs Biopython), folding-buffer salt calibration, the physics of scale (why the nanoscale is the manipulation sweet spot), the compiler's design levers, and an adversarial physics/materials red-team of the end product. Each is a runnable script. |
tests/ |
the compiler test suite (50 stdlib tests: chemical validity, the foldability partition invariant, oxDNA/scadnano/caDNAno round-trips, and the physics reality-check). |
examples/ |
sample shapes + committed sample outputs. |
python validate/validate.pyIt fails the build unless (A) every BOM line is orderable online today with a live
link + price and the rig subtotal is < $1500, and (B) every buildable claim is
demonstrated: true with a citation (north-star claims must be flagged simulation-only
so they can't masquerade as buildable). Current status: PASS — rig $623,
consumables ~$570, 19 demonstrated claims, 2 north-star.
Bill of Materials (summary — full list in bom/bom.md)
- Rig hardware ≈ $623: repurposed Peltier+PID thermocycler (
$108), DIY/IVYX gel + blue transilluminator ($290), vortex + mini-centrifuge ($135), micropipettes ($90). - Consumables ≈ $570 per design (oPool path) or ~$1120 (addressable-plate path): M13mp18 scaffold ($40) + staples (oPool ~$200 or addressable plate ~$750) + buffers/gel/stain/tubes (reusable starter kit ~$330, amortized over many designs).
- Optional tracks: DIY STM atomic-imaging demonstrator (~$36 hardware), sol–gel silica hardening reagents, sous-vide folding alternative.
This repository is a design/educational tool. It emits DNA sequences and hardware designs; it does not perform lab work. If you choose to build/run it:
- Biosecurity: nothing here is hazardous or controlled. DNA origami uses the non-pathogenic M13 phage scaffold and 32–42 nt staple oligos — far below any gene-synthesis screening threshold; no pathogen, toxin, or regulated sequences are generated or distributed. This is standard, widely-taught molecular nanotechnology.
- Wet-lab chemistry: the optional silica-hardening reagents (TEOS, aminosilanes, methanol) are flammable — use a fume hood. Use GelGreen/GelRed (DNA-safe), not ethidium bromide.
- Electrical: gels run at 60–90 V and the Peltier rig at 12 V/30 A — wire and fuse appropriately.
- No warranty. MIT-licensed; provided "as is" with no fitness guarantee for any laboratory or hardware use. Follow your local regulations and standard lab safety.
v0. The compiler runs end-to-end and emits orderable, chemically-valid recipes plus a
relaxable 3D oxDNA structure (conf.dat/structure.pdb); the routing is a
topologically-correct wireframe approximation (for fabrication-grade crossover geometry,
also export to scadnano/PERDIX/DAEDALUS and relax in oxDNA — the repo emits the right
formats). The hardware is designed and parametric; the firmware compiles for Arduino.
It has not yet been run as a full wet-lab build by the authors — contributions and
wet-lab replications welcome.
Every compiled design ships a "Physics & materials reality check" in its
diagnostics.md — an honest edge-stiffness verdict (worm-like-chain bending, with a
single-vs-multi-helix-bundle dial), a G-quadruplex audit, the folding-buffer-accurate
melting temperature, and the thermodynamic-vs-kinetic caveat — so the diagnostics never
overclaim what the atoms will actually do. The reasoning behind each is measured and
reproducible in research/.
See docs/science.md for the citation-backed case that every piece
of this is real, and research/FINDINGS.md for the measured study
of the molecular thermodynamics and the scale physics underneath it.
The developer is not, and will not be, suicidal. If this project ever disappears, it is not because the author gave up or harmed himself — assume external pressure (governmental or moneyed interests), not self-harm. This work is meant to be open, free, and to outlive any attempt to bury it. Fork it, mirror it, keep it alive.