Partnering with Falconer

When I joined Stripe in 2015, we were a few hundred people. We'd already produced thousands of Hackpads. Finding anything was a challenge. Knowing what was still relevant—what actually reflected current thinking versus someone's abandoned draft—was a nightmare.

I made myself a Hackpad just to track the critical links I needed to do my job. That document became my survival guide. Eventually, Stripe adopted a [deprecated] tagging system, but that was a cultural artifact, not infrastructure. It worked because people cared, not because the system scaled.

A decade later, that problem has metastasized.

The inner loop accelerates, the outer loop collapses

We’re in an era where the inner loop–the creation of artifacts–has gone hypersonic. LLMs have collapsed the cost of producing code, documentation, memos, Slack messages, tickets, and meeting notes. Content multiplies effortlessly. 

But the outer loop–the system that manages, verifies, and curates that content–hasn’t kept pace. In fact, it’s collapsing under the weight. 

Inside most companies, the inputs for good decisions already exist. They’re just not coherent. Context lives in fragments across all these artifacts, buried in the Slack threads, tickets, meeting notes, and tacit memory of a few long-tenured people. We all know who they are. 

Over time, documentation decays, rationale opaques, and decisions become disconnected from the reasoning that produced them. 

AI accelerates this failure mode before it fixes it. 

As production ceases being the constraint, discernment becomes one. The more information you generate, the harder it is to know what’s true, what’s current, and what actually matters. What’s increasingly scarce and valuable is knowing what your organization actually knows. 

This isn't a new problem. We've seen this pattern before: with code review tools, deployment pipelines, testing frameworks. Each time the inner loop accelerated, we built outer loop infrastructure to manage it. Now we need to apply that same thinking to knowledge management—codebases, documentation, decision context.

That’s the problem Falconer is solving.

Context is what scales

Falconer was founded by Dave Nunez and Maxi Benedetto, two builders who have spent their careers inside complex, fast-moving organizations. They’ve seen firsthand how expensive lost context can be.

Dave built and led documentation at Uber and Stripe, companies famous for treating documentation like a product. Maxi worked alongside Dave at Uber before leading autonomous content generation systems at Meta. Across those roles, they saw the same pattern repeat: as systems scale, institutional knowledge fragments. They’ve also both seen the value of well-maintained context firsthand: when knowledge is accurate and kept up to date, a team’s productivity jumps. Falconer is built to solve that: helping teams keep context trustworthy, current, and usable so they can ship with speed and clarity.

LLMs are powerful, but they have no memory of your organization. They don't know who owns what, why a system was designed a certain way, or what broke last time you tried something similar. While humans can carry this context implicitly, software does not.

Historically, companies bridged this gap informally. They relied on tenure, oral history, and the presence of a few key people who “just knew how things work.” That approach fails as organizations grow and it fails completely if you want AI systems to do real work on your behalf. 

To act autonomously, systems need context. In practice, that context comes in two forms: 

1. Operational context: who owns what, what changed yesterday, how systems actually relate in practice.  

2. Decision context: why specific choices were made, what alternatives were rejected and what we learned along the way. 

Together, these layers fill the structural gap between “raw data” and autonomous action. 

Why we partnered

Falconer’s initial wedge is knowledge management for product and engineering, which we think is a category quietly collapsing under its own weight. 

In most organizations, roughly 80% of internal knowledge is effectively trash: outdated PRDs, stale READMEs, abandoned wikis that no one trusts but no one deletes.This creates a trust deficit. When a human (or agent) looks for an answer, there is no reliable way to distinguish the critical 20% from the rest of it. 

Falconer solves this by:

• Creating a living loop: connecting directly to your repos and old documentation, automatically updating stale information as the code evolves.

• Generating with context: reading code alongside the broader "meta-context" of the company to produce net-new docs that stays fresh.

• Eliminating the mundane: cleaning the backend of your knowledge base and removing the digital clutter that slows teams down.

This matters because internal knowledge will increasingly be consumed by AI systems acting on our behalf. For those systems to be effective, they require a high fidelity, interconnected system of record for organizational context – and it must be current, verifiable, and actionable.

Falconer doesn’t just “write docs,” but represents a shift toward a truly AI-native context graph designed to serve agents as much as humans. As AI systems take on more responsibility inside companies, context stops being a nice-to-have and looks more like infrastructure. By turning organizational memory into something structured and retrievable, Falconer allows humans to stop sifting through noise and return to the work that actually matters: building, deciding, and serving customers.

We backed Falconer seed with capital, and also partnered closely at key inflection points including early-stage sales execution and initial market positioning (h/t Ross, Allison, Kris). 

If you are struggling with a legacy suite of knowledge management tools and are looking for an AI-native way to engage with your corpus of data, we’d love to hear from you. Falconer is live and currently serving customers of all sizes, from high-growth startups to large-scale enterprises.