A new way to diligence.

// 01 / WHAT IT IS

gutcheck is a diligence platform for agentic replication. Enter a URL and agents research the build, scope the architecture, and write real code to produce each company's hero deliverable. This is then used to drive a cost-to-replicate estimate across compute, human, and go-to-market layers.

// 02 / HOW IT WORKS

The analysis engine scans the target, identifies features and integrations, then scores replication difficulty across four dimensions: Product, GTM, Physical Infrastructure, and Talent & Domain. Low scores mean shallow moats; high scores mean deep ones.

// 03 / WHAT THE AGENTS READ

Each analysis pulls from several sources: the product's homepage and marketing site, its key subpages (pricing, about, features, docs), and external search results. The replication attempt goes further - agents research the stack, scope the architecture, and write a working grain of real code, then derive the cost from what they actually had to build. When a score reflects that build it carries a Build-tested badge.

// 04 / CONFIDENCE & ASSUMPTIONS

Every estimate carries assumptions, and we would rather show them than hide them. The strongest confidence signal is whether we attempted a real build: a Build-tested score was adjusted by evidence from generated code, not analyzer heuristics alone.

• Human cost assumes a fully-loaded San Francisco engineer at roughly $250K / year.
• Engineering build cost is calibrated by what the agents learned attempting the build - a multiplier bounded to within ±50%, not a raw guess.
• AI acceleration is baked into the Product score: code an agent can largely generate scores as a shallower moat.
• Cost is shown as a band, not a single number - cost to compete is uncertain by nature.

// 05 / WHAT THIS DOES NOT COVER

Reading a benchmark across funds? A few things to keep in mind - calling them out is the point, not a disclaimer.

• Subset coverage: we analyze a sample of each fund's portfolio, not every investment.
• Fund-size skew: a large fund may have dozens of recent investments while a smaller fund's sample spans several years - not apples-to-apples without context.
• Vintage differences: older investments predate the current AI wave; cross-vintage comparisons need that caveat.
• Fairer comparisons: where possible, compare funds of similar size and vintage rather than the raw cohort.