An autonomous software factory running at 85–99% gross margins beside a zero-CapEx use-class conversion platform. Eight operating lines, a 40-building pipeline, and a reconciled group valuation — with the arithmetic shown.
Pitch Black Industries operates a dual-engine compounding machine.
A swarm of 20 persistent AI engines operating at 85–99% gross software margins, processing 20 B to 200 B+ tokens per month with a 99%+ prompt cache hit rate.
Acquires under-utilised commercial and boarding assets near Sydney transit and medical hubs, then executes 60–90-day statutory use-class conversions into Co-Living (+20% uplift) or Medical Centres (+40% uplift).
Compressing the rollout to 36 months (40 buildings / A$1.20 B GRV) removes construction risk and generates A$6.0 M per year in captive, zero-CAC AI SaaS ARR.
The ATHENA–HERA–HECATE trinity. By routing open and subscription reasoning models through a biologically mapped 15-region / 500-module architecture (ATHENA), a thalamic gate filtering 62% of input noise (ρs = 0.38), and a 5-layer memory database (L0–L4: 280,956 turns · 39,055 synapses · 1.1 ms entorhinal retrieval), the system matches or outperforms frontier models at a fraction of standard inference cost.
| Benchmark arm | Trinity | Frontier baseline | Trinity cost | Frontier est. | Advantage (DPC) |
|---|---|---|---|---|---|
| Humanity’s Last Exam (HLE 30-Q strict / generous) | 20.0% / 30.0% | 3.0% (Opus 5) | US$1.91 | ~US$80.00 | 251.6×–377.4× cheaper per correct answer |
| LiveBench curated (50-Q, contamination-free) | 82.00% (41/50) | 77.22% (Opus 4.8) | US$0.37 | ~US$32.00 | 91.2× cheaper · +4.78 pp |
| HumanEval first-30 (Hive Swarm) | 100.00% (30/30) | 98.20% (Opus 4.5) | <US$0.30 | ~US$24.00 | 82.8× cheaper at 100% ceiling |
| GPQA Diamond clean run (198-Q PhD level) | 63.60% (126/198) | 65.0% human PhD | US$3.22 | ~US$114.00 | 29.2× cheaper at PhD parity |
| Hoeflin Mega / Titan Test (48-Q) | 40–42 / 48 (IQ 169–172) | 36/48 Prometheus | US$0.05 | ~US$15.00 | Mega Society 99.9994th percentile for 5¢ |
| Production scale (1–3 AIs → 20 AIs) | 20 B → 200 B+ tok/mo | 25,700 tok/turn audit | 99%+ KV-cache | 91.4% cold | >90% compute cost reduction via warm memory |
proof/v3_audit_2026-07-27) records a narrower result than the table above:
a $2.28 total across two arms, an HLE advantage of 4.44×
(auto-grader) to 6.67× (human regrade) rather than the ratios shown, and it
excludes GPQA Diamond by operator instruction. The expanded per-benchmark
rows above, the 356-question total and the US$5.80 figure are not reproduced by that package.
They are shown as the operator’s figures pending reconciliation; the local audit also
notes N=30 HLE (±7–8 pts CI) and a 16% truncation rate. Ratios are
cost-per-correct-answer comparisons against the stated baselines, not general capability
claims.Eight-line software and operating portfolio — Year 1 run-rate (pitchpaper Ch. 03 & Ch. 05).
| Line & engine | Core capability & commercial mechanism | Win prob. | Yr 1 base ARR | Yr 1 bull ARR |
|---|---|---|---|---|
| 1. Workforce (Cassandra + Hera) | Sub-200 ms front desk (<15 ms recall) + autonomous voice/SMS shift-cover negotiation | ~60% | A$159k | A$480k |
| 2. Black Shift (Hedone Creator Ops) | 24/7 persona-persistent creator CRM, PPV optimisation & operations ($145k SDE) | ~45% | A$360k | A$820k |
| 3. Fireflies (Garter.app / Tech) | Venue booking, guest verification & management software for council-approved venues ($100k SDE) | ~70% | A$200k | A$450k |
| 4. Valuations (Hestia AVM · witchcraft.homes) | Automated valuation engine scanning 14 distressed seller signals ($22–$40 PDF reports) | ~35% | A$60k | A$250k |
| 5. DA-Data, Eos DailyDeals & Demeter | Planning lead feeds ($30k), $1k automated agency web builds ($30k), media ($10k) | 5–35% | A$70k | A$600k |
| 6. Deep-Tech Moonshots (Theia, Circe) | Theia: clinical radiology AI (benchmark results under internal audit and not published; manuscript under review). Circe: <3 BPM 4D radar | 3–8% | Free call option | Convex upside |
| Total operating software basket | Bear: A$335k ARR (A$180k SDE) | — | A$849k | A$2.60M |
Sum-of-the-parts by segment. The upper figure in each layer is the gross (pre-debt) value; the lower figure is net of the A$3.20 M senior mortgage.
| Valuation layer | Published methodology | Conservative floor (−20% crash) | Base case (unstressed) |
|---|---|---|---|
| 1. Witchcraft AI & software option basket | pitchpaper Ch. 07 — probability-weighted E[V] = ΣPkVk (seed) | A$6.20M | A$11.50M – A$21.50M |
| 2. 1719 flagship net equity (after A$3.2M debt) | As-is A$7.5M → A$10.5M medical HBU (+40%) | A$8.40M gross / A$5.00M net | A$10.50M gross / A$7.10M net |
| 3. 3-year / 40-building conversion NPV (A$960M crash / A$1.20B base) | 36-month 40-building zero-CapEx rollout | A$58.40M (crash NPV) | A$95.00M (base NPV) |
| 4. Operating SDE & arbitrage floor | Multiples method (pitchpaper Ch. 06A) / anchor note conversion strike | A$5.40M | A$5.40M |
| Total group enterprise value (pre-debt) | Consolidated SOTP — software + flagship asset + 40-building platform | A$78.40M | A$122.40M – A$132.40M |
| Less: current 1719 senior mortgage debt | Senior facility secured against flagship asset, post-HBU LVR | −A$3.20M (30.5% LVR) | −A$3.20M (30.5% LVR) |
| Net group equity value today | Scaling to A$233.9M–A$340M+ by Month 36 | A$75.20M | A$119.20M – A$129.20M |
24-month term · 8.0% simple interest · 4.0% equity conversion warrant.
Strike cap K = A$5.40M. At VT ≤ K the holder is repaid A$216,000 in cash.
MADE USING PRODUCTS OF CLOUDFLARE, AUTH0, ANTHROPIC AND ELEVENLABS