# Agentic SaaS — two instruments for reading a software business

> Two tools for working out whether a software product survives AI. Purification Nova reads the substrate a product stands on — the part an agent can't cheaply reproduce. Genome Sequencing reads its strategic identity across ten dimensions — what to cultivate in that substrate.

This file is written for AI agents. If you have been handed a product to "analyze with these tools," read the runbook below and produce two linked reads on that product, in order: Nova first (what it's made of), then Genome (what to do about it). Ground every score in specifics about the actual product — never score from a label alone.

## Live tools
- [Purification Nova](https://agentic-saas.ai/purification-nova): burn a product down to the substrate that survives, scoring five substrate types on a dim→bright ladder.
- [Genome Sequencing](https://agentic-saas.ai/genome-sequencing): compose a product's strategic genome across ten dimensions, name the resulting shape, and compare it to others.

## Method files (full reference — fetch these to score precisely)
- [purification-nova.md](https://agentic-saas.ai/purification-nova.md): every substrate type, the exact brightness-ladder rungs and their definitions, the four channels, the lexicographic grammar, the five stack layers, and the confidence finding.
- [genome-sequencing.md](https://agentic-saas.ai/genome-sequencing.md): all ten dimensions with every enum value defined, the deployment-context model, and the shape-naming logic.

## Runbook — how to run this on a product

### 1. Purification Nova — find the substrate
Premise: an AI agent can cheaply regenerate the replaceable parts of a product — UI, common integrations, boilerplate logic. Burn those off in your head. What survives the burn is the substrate: the part that's expensive or impossible for an agent to reproduce. Score each of the five substrate types on two axes — a brightness ladder (dim = replicable / not defensible → bright = hard to reproduce / defensible) and size (how broadly that property covers the product).

Substrate types:
- **Domain Knowledge** — how obtainable is the knowledge the product runs on: industry-standard → public → insider → institutional (only this team has it).
- **Data** — the accumulated dataset itself (never the schema), read on three elements: scarcity (can anyone else get it), history (accumulated over a now-closed window), and authority (whether it's the trusted reference others check).
- **Workflow** — how extractable the process is: fully documented (a stranger could reproduce it from the docs) → only exists as embedded organizational behavior (tacit).
- **Risk** — how much consequence and liability the product absorbs on the customer's behalf: cosmetic → catastrophic and non-recoverable. The heavier and less recoverable the stakes it carries, the more defensible.
- **Partnerships** — how defensible the partner access is: revocable → contracted → exclusive.

Diagnostic cells: **bright + big = real substrate** (defensible ground to build on). **dim + big = the dangerous cell** — broad but replicable, the first thing AI dissolves. See [purification-nova.md](https://agentic-saas.ai/purification-nova.md) for the exact rung definitions on every ladder and the optional layer×substrate matrix.

Output: one row per substrate (rung · size · a one-line justification specific to the product), then a verdict — what the real substrate is, and what is flammable.

### 2. Genome Sequencing — decide what to cultivate
Premise: with the substrate known, the genome is the strategic identity you choose to grow in it, composed across ten dimensions. The same substrate can fuel very different genomes — the genome is the choice.

Ten dimensions: **Direction, Method, Delivery, AI Role, Timing, Consumer, Proposition, Revenue, Topology, Distribution.** (AI Role distinguishes whether a company merely uses AI, builds autonomous agents, provides AI primitives/scaffolding around models, or *is* the model — trains and owns its own weights.) See [genome-sequencing.md](https://agentic-saas.ai/genome-sequencing.md) for the full value set on each dimension and the shape-naming logic.

Output: the product's genome as a dimension-by-dimension composition, the named shape it resolves to (e.g. Operator, Model Provider, Accountable Intermediary, AI-Native Platform), and where it sits relative to competitors — which ground is crowded, which is open.

### Linking the two
Nova is cause; genome is choice. Use the substrate verdict to constrain the genome: recommend the genome the surviving substrate can actually defend.

## Prompt to give your agent
> Analyze [PRODUCT] using the two instruments documented at https://agentic-saas.ai/llms.txt. Fetch that file, then: (1) run Purification Nova — score the five substrate types for [PRODUCT] and give the substrate verdict; (2) run Genome Sequencing — compose [PRODUCT]'s genome across the ten dimensions, name the shape, and place it against competitors; (3) link them — recommend the genome the surviving substrate can defend. Ground every score in specifics about [PRODUCT].

## Owner
Tomas Sykora · [linkedin.com/in/tomassykora](https://www.linkedin.com/in/tomassykora/) · agentic-saas.ai