Gabriel Spencer-Harper

Gabriel Spencer-Harper is the co-founder and CEO of Meticulous, a software-testing company that turns recorded interactions with web applications into automated checks. His work addresses a constraint that faster code generation makes more pressing: engineers still need to understand what a change will break before they release it.

From customer feedback to software reliability

After graduating from Warwick in 2016 with a degree in Computer Science and Chinese, Spencer-Harper co-founded Moju with his brother Milo. The product helped brands and marketing agencies find Instagram influencers by subject, audience characteristics, and language. He divided his time between programming, fixing bugs, customer conversations, and user interviews. Moju reached Y Combinator’s final interview but did not receive funding. His early approach to entrepreneurship emphasized releasing a product early and keeping customer feedback close to development.

His subsequent engineering work at Dropbox and Opendoor brought software reliability into focus. Applications could pass unit and integration tests yet fail when users followed paths their developers had not anticipated. At Dropbox, his team spent hours repeatedly checking workflows by hand; automated end-to-end tests required writing and maintaining mock data and responses. Watching recordings of those manual testing sessions get discarded gave him the idea of turning the interactions into reusable tests. That experience became the starting point for Meticulous.

Meticulous was founded in 2021 and joined Y Combinator’s Summer 2021 batch. An early version gave developers command-line tools to record browser actions, simulate them against another version of an application, and compare screenshots. The product developed into a system that selects relevant flows and returns visual differences on pull requests, reducing the work of choosing and maintaining individual test cases. Spencer-Harper announced a $4 million seed round in January 2024. Meticulous announced a $15 million Series A in July 2026, naming Dropbox, Notion, Wiz, and ElevenLabs as customers.

Testing behavior developers did not predict

Spencer-Harper’s approach through Meticulous starts with observed interactions: use the paths people actually take through an application to broaden coverage, then show engineers the consequences of a proposed change.

• Replay testing: His writing on user-driven tests identifies two weaknesses of handwritten tests: developers can specify only scenarios they imagine, and every test requires upkeep. Replaying an interaction against the base and proposed versions of an application exposes differences in behavior, screenshots, or uncaught exceptions. The developer then decides whether a difference is intentional.

• Frontend isolation: Meticulous records network responses and supplies them again during replay. A flow that creates an account or edits data can therefore run repeatedly without repeating those operations on a live backend. Spencer-Harper recognizes the tradeoff: this checks frontend regressions against recorded responses, with narrower coverage than replaying the entire application against a representative backend.

• Deterministic browsers and coverage-guided selection: The system Meticulous builds combines browser determinism with analysis of the code paths exercised by recorded sessions. Repeatable execution reduces distracting test failures, while selecting flows that exercise different paths makes a large collection of recordings useful. Screenshots after individual actions show reviewers changes throughout a journey. These engineering efforts belong to the Meticulous team, rather than to Spencer-Harper alone.

Verification as the constraint on AI coding

In Why AI Didn’t Actually Make You Ship Faster, Spencer-Harper argues that verification becomes the bottleneck when AI generates code faster than people can review it. Handwritten assertions cover selected expectations; his ambition is to make broader frontend checks practical without asking developers to write and maintain each test.

Meticulous’s work on feedback for coding agents extends that approach: agents could inspect the effects of their changes and iterate before requesting human review. The goal follows from the problem that prompted the company—help developers see the consequences of their code while there is still time to change it.