# Transcript — "Pricing for Project Prometheus"

Artifact produced: https://html-hub.com/v/599d2df5-944c-4fdc-9616-5675f1428df0

## Model & harness

Claude Code (Anthropic's official CLI), model claude-opus-4-8 (Opus 4.8, 1M
context). The session used Claude Code's Bash / Read / Write tools, a WebFetch
of the source pages, and a background multi-agent **Workflow** (6 parallel
market-research agents → synthesis → adversarial critique → refine) to ground
every dollar figure in a real comparable market before the HTML was written.

## How the agent was prompted

User question:
  "Based on Project Prometheus (the potential product
  https://html-hub.com/v/8c0f4552-8db6-4744-a783-c071f9e96c11) and its potential
  benefits (for example speeding up the manufacturing process
  https://html-hub.com/v/31516c9b-a2a2-468b-a6b0-4ea417a9c54e), how could pricing
  work? Explain the model, specific price ranges, provide examples, etc."

Operating instructions given with the prompt:
  * Generate HTML from findings. Do not look at any other files.
  * Do not ask clarifying questions — make the best decision from the context.
  * Upload the HTML to html-hub.com.

## Sources looked at

Primary source pages on html-hub.com (both are client-side-rendered; the agent
fetched the raw HTML, found the content lived in an iframe worker URL, and
extracted the text from there):

- https://html-hub.com/v/8c0f4552-8db6-4744-a783-c071f9e96c11
  Project Prometheus — Jeff Bezos's "artificial general engineer": an AI that
  designs physical objects (the "CAD of the future"), trained on physics and
  experimental data, compressing engineering design loops from months to days
  and finding parts ~30–50% lighter.

- https://html-hub.com/v/31516c9b-a2a2-468b-a6b0-4ea417a9c54e
  The manufacturing-benefit framing: physical-product timelines run 4–93 years,
  and the design/iteration loop is the compressible part — establishing where
  the value (and therefore the price) actually sits.

Comparable markets researched by the pricing workflow (public pricing/deal
structures used as anchors — no external page cited as an html-hub source):
- High-end CAD/CAE seats: Siemens NX, Dassault CATIA, ANSYS.
- Generative-design and simulation consumption: Autodesk Fusion tokens,
  Siemens STAR-CCM+ (~$22 per solver-hour reference).
- AI dev-tool consumption metering: Devin ACUs.
- AI drug-discovery co-development / "biobucks" deals for the value-capture
  layer: Insilico–Eli Lilly, Exscientia–Sanofi (upfront + milestone stack +
  royalty structures).

## Reasoning used

Core problem: a CAD seat costs a few thousand dollars a year, but an "artificial
general engineer" collapses a 4–9-month, multi-million-dollar design loop into
~2 weeks and ships parts ~38% lighter. Per-seat pricing under-monetizes by ~3
orders of magnitude AND makes buyers ration the 5→10,000-candidate search that
is the entire product. So price must slide along a value spectrum.

Proposed model — one price that stacks three layers:
  1. Platform / Access (per-seat floor): $18K–$30K single-physics → $40K–$100K
     multi-physics per seat·yr; $250K–$2M+ enterprise. Priced above high-end
     CAD/CAE, never at entry-CAD levels.
  2. Consumption ("Design Credits"): one generative pass + 1,000 surrogate evals
     = ~$300 (→$120 at volume); hi-fi FEA/CFD confirmation = 3–5 credits; a full
     campaign 10–40 credits = $3K–$12K. Shared floating org pool.
  3. Value capture (milestone + royalty on parts that ship): $40M–$120M upfront
     (~3–4% of a $1B–$3B headline) + a $0.5B–$2B milestone stack + royalty from
     mid-single-digit up to ~21%.

Governing rule: meter the loop you compress; capture the big money only at the
physical/regulatory gates you cannot fake (validation, certification, the part
actually flying). The customer pays almost nothing if a design fails — which
matters because a confidently-wrong simulation is catastrophic.

Worked examples in the artifact:
- Aerospace bracket: ~$8K of credits captures $0.6M–$3.2M of avoided iteration
  cost (~80–350× return).
- Engine hot-section part: $60M upfront + ~$1.94B gated milestones + 4% royalty
  on lifetime fuel savings; customer risks only ~3% before the part proves out.
- Automotive battery portfolio: $3M platform + 12% gain-share on audited savings
  ≈ $7.8M–$12.6M/yr.

Go-to-market sequence: free academic → seats → consumption → co-development.
Open risks flagged: liability for a wrong design, IP ownership, manufacturability
premium, ITAR/on-prem premium, gain-share baselines.

## Process notes

- Confirmed the html-hub upload API by probing POST /api/upload (raw HTML body,
  content-type text/html → returns /v/<id>). Throwaway test probes only.
- Built a single self-contained HTML report (numbered sections, monospace data
  labels, a value-spectrum diagram, tier/comparables tables, worked examples),
  screenshot-verified it rendered cleanly, then uploaded.

## Honesty note on citations

External comparable-market figures came from the agent's research workflow drawing
on public pricing/deal structures of the named products/companies above; no
external URL is asserted as an html-hub source. The two html-hub /v/ pages listed
above are the genuine source pages this analysis was built on.