AutoPhi V21 — PhantasmPhi
CRI-ONE · AutoPhi V21 · PhantasmPhi Series
Ghost-Math ICs
The AutoPhi voxel with only the math strand lit — a witnessed compute lineage.
The math itself, silicon-resident. Deposition corpus in mask ROM. QBeam-only I/O. Quantum branching via ExitPhi. And — on 2026-08-22 at 00:29:51 UTC — the first verifiable placed-and-routed silicon in the AutoPhi portfolio's history.
Enter the RonnaPhantasmPhi listing →What "ghost-math" means
A PhantasmPhi die carries a set of attributed mathematical depositions — each one a formula from the CRI-ONE alchemy calculator corpus — as mask-programmed microcode ROM. The chip idles dark; nothing runs; the math lies dormant. When a QBeam classical frame arrives with arguments, the requested deposition materializes, computes on the AutoPhi seed color-math ALU, and beams the answer back in ten bytes. Deposition #10 is always the handshake — the phantom answers by name.
The functions cannot be flashed, patched, or exfiltrated as source. They can only be invoked. The math is protected the way a physical constant is protected: it is embedded in the material.
The derivation tree — one voxel, four products
Read across the CRI-ONE portfolio (Projects 02, 05, 18/19, 76, 83, 84), one architecture wears four products. Not four separate inventions: subtractions and selections from a single voxel.
| Product | The voxel with… | Project |
|---|---|---|
| AutoPhi Future (V18/19) | all nine technologies lit — compute and energy work | 18/19 |
| Quantum Battery | the processing strand dark; energy-cycle-only mode | 05 |
| PhantasmPhi (this series) | only the math strand lit; deposition ROM on the color ALU; QBeam-only I/O | 83 |
| PhantomCore | fetch, memory hierarchy, interrupts added back — a CPU by design | 84 |
The seed-matrix formula that scales AutoPhi Future across nodes scales
PhantasmPhi identically, with the per-voxel function field swapped from
FLOPS per voxel to deposition invocations per voxel.
The ratio formula is the same; only the per-voxel deliverable changes.
The witnessed record — 2026-08-22 00:29 UTC
| Correctness (saturation battery) | 256/256 in order |
| Measured speedup vs single phantom | 15.8× (352 vs 5,587 cyc) |
| Peak concurrency | 64 / 64 phantoms active |
| Sustained rate at 100 MHz | ~73 M invocations/s per die |
| Routing violations (final) | 0 (converged 25,622 → 0) |
| Magic DRC | 0 |
| KLayout DRC (independent engine) | 0 |
| Netgen LVS | CLEAR (0 errors, 0 unmatched) |
| Setup + hold slack (all 9 corners) | WNS = 0, TNS = 0 |
| Total power | 150 µW |
| Fabrication GDS | 102,054,112 bytes |
| SHA-256 | D61C07EB…D3264697 |
The six editions
Each edition is the same die strapped with a different deposition-pack ID
(PACK_ID = 1–6). All six ship with 12/12 proof-battery pass,
official sky130_fd_sc_hd mapping (~40,514 cells / 1,142 flops /
278,120 µm² — matched digit-for-digit across four independent reproductions),
and a verified GDSII floorplan artifact.
| Edition | Software sibling | Functions in sibling |
|---|---|---|
| PHM01-MED | AlchemyCalc Medicine (V3·MED) | 98 |
| PHM01-NRG | AlchemyCalc Energy (V3·NRG) | 96 |
| PHM01-C | AlchemyCalc Carbon "C" (V3·C) | 95 |
| PHM01-OMNI | AlchemyCalc Omnia (V4·OMNI) | 207 |
| PHM01-FORT | AlchemyCalc Fortuna (V5·FORT) | 227 |
| PHM01-INV | AlchemyCalc Inversa (V6·INV) | 244 |
The Compute Ladder — measured · projected · quested
Per Project 19's MIRACLE_CALCULATION.md: "the value of a capability you never had is not a number." Each rung below is labeled by its evidence class.
| Rung | OPS coordinate | Evidence |
|---|---|---|
| Base camp — TODAY | 2.1 × 10⁹ OPS / die | Witnessed: routed silicon, LVS clear, GDS hash |
| V2 nine-tech compute group | ~4× base (measured RTL 15.8× dual-fabric) | RTL-proven; forge run in progress |
| + Node shrink to 1.5 nm | ~1.6 × 10¹³ OPS / die | Projected — 18 lineage growth path |
| + Clock 100 MHz → 5 GHz, wider datapath | ~8 × 10¹⁵ OPS / die | Projected — measured pipeline × leading-edge clocks |
| + Fleet (2,500 facilities × wafers × licensees) | ~10²⁷ OPS — RONNA-class | Projected on defensible arguments |
| QUETTA (10³⁰) | the flag on the summit | Quest coordinate — no identified physical path |
PhantasmPhi is priced from the base camp, planned to the Ronna-class summit, and named for the Quetta quest that the record does not yet support. The Generational Law guarantees each future generation lists at a lower cost per unit of capability than its predecessor.
The dual fabric — QBeam classical, ExitPhi quantum branching
QBeam (Project 76 heritage) — the classical fabric.
Every deposition invocation is an 18-byte QBeam frame; every answer is a
10-byte frame. No conventional bus, no memory-mapped host interface, no
attack surface beyond the frame protocol. Unknown depositions are
rejected with status 0xEE.
ExitPhi (USPTO App# 19/722,805) — the quantum branch. A 16-instruction gate stream to the Quantum Execution Unit, QEC-surface- code-corrected, fidelity-scored, beamed on the nanophotonic link. Quantum state branches die-to-die on the photonic wire while classical answers branch on QBeam — one voxel, two coherent fabrics.
All nine AutoPhi seed technologies are instantiated on V2: LED power recycling, vertical threading, chiplet stacking, nanophotonic I/O, QEC surface code, electromagnetic cooling, quantum battery, quantum execution unit, neuromorphic engine — plus the color-math ALU voxel foundation.
The pricing doctrine — priced against compute affordability itself
Today's compute affordability, defined honestly: the market cost to build equivalent deployed performance. A 20 MW QuettaPhantom facility builds for ≈$0.4B; the wrap's franchise ceiling is 2,500 such facilities. Product: $1,000,000,000,000 — that is 100% of today's compute affordability.
The listed price is set by doctrine at exactly 25% of that number (intentionally 75% lower), reduced a further 15% because the wrap is non-exclusive — many licensees, many fleets, no datacenter dependency. Net: $212,500,000,000 per license, 78.75% below today's compute affordability. Fixed by doctrine, not by negotiation.
What is for sale — and what is not
The RTL is the only thing for sale. Netlists, GDSII artifacts, testbenches, flow scripts, witness records, and documentation ship as evidence and support for the RTL, not as separate assets. One asset, one price, no ambiguity.
Patents are retained by the inventor (QBeam PHY via ExitPhi, USPTO App# 19/722,805). This is a documentation and worldwide commercialization-rights transaction, not a patent license or assignment.
Delivery discipline — never free, never naked
- Never free: the RTL is delivered only after payment clears, as an AES-256 encrypted archive, password conveyed separately by postal mail.
- Never publicly hosted: the RTL, netlists, and release zip are never posted to any public repository, CDN, or store download.
- Per-license serialization: each licensee's archive embeds a unique provenance ledger entry (licensee name, date, license number), so any leaked copy identifies its source.
- Safe by silicon: the ROM is mask-programmed and immutable — no flash surface, no code-injection vector — and the QBeam port exposes exactly one behavior (18 bytes in, 10 out, nothing else reachable).
Eligibility
Offered to any qualified buyer except entities of, in, or controlled from the People's Republic of China or the Russian Federation — always subject to US export law (EAR / OFAC). For this SKU, this supersedes the portfolio's default US-only clause by owner instruction (2026-08-21). USD only.
The honesty box
"The value of a capability you never had is not a number." — Project 19, MIRACLE_CALCULATION.md, adopted here as doctrine.
Cycle counts, cell counts, area, DRC/LVS verdicts, and the fabrication GDS hash are measured in RTL simulation and OpenLane2 signoff, and are reproducible from the shipped testbenches and flow. Clock frequencies, Compute-Ladder projections above the base camp, wafer yields, and facility costs are design targets and market estimates from the engineering record — not silicon measurements or foundry quotations. Quetta and Ronna are naming tiers, never performance claims. Formulas and laws of nature are unpatentable; the CRI-ONE curation, microcode mapping, and this silicon embodiment are what is protected.
Contact
Christopher Gabriel Brown — Inventor · Author · Visionary
christopher@cri-one.com · crioneaka@outlook.com
1341 Wellington Cove, Lawrenceville, GA 30043-5255, USA
Communication by email and postal mail only — no phone calls, no brokers, no intermediaries. US companies only. USD only.
Enter
RonnaPhantasmPhi — the defensible summit
64 phantom engines · nine technologies · dual fabric · $212.5B per license
Open the RonnaPhantasmPhi listing →