LaserOwl is the work of Joe Taylor — a product engineer and founder addressing the review gap around agent-authored code.

Background

Joe combines frontend engineering with product judgment. He can put a working interface in front of a team inside a week, and is decisive about which interface is worth building. It is what allows the design-partner work to move at the pace it does.

Based in London, with fifteen years of product work across enterprise and startup environments, including time in San Francisco, New York, and Copenhagen.

Joined Cisco Collaboration as engineering manager in 2014, where he co-founded the Collaboration Innovation Labs and led TeamTV, an internal video portal for distributed teams. He also contributed the first user interface for Monica, Cisco’s first virtual AI assistant for Spark and Webex. Earlier work included product engineering at Assemblage from 2012.

In 2018 he raised over $800,000 as CEO of a venture-backed meeting-assistant company. The team released Touchbase, a lightweight video tool that reached more than 4,000 teams during the pandemic, and Nemo AI, a meeting-analysis system trained on thousands of hours of recorded conversations. The products worked; the market did not — a kind of pattern recognition that comes only from running a company through to its conclusion.

Since 2023 he has worked as a product and UX consultant to a number of startups, most recently Harness from 2024 to 2025, where much of the thinking that underpins LaserOwl began to take shape.

Why LaserOwl

Engineering teams are producing more agent-authored code each quarter. The artefact left behind — the pull-request diff — is materially weaker for the purposes of review, debugging, and audit than the human-authored work it replaces.

Claude Code already accounts for around 4% of public GitHub commits, and the share is rising. Anthropic’s own customer case studies report an 80% reduction in incident-investigation time when the agent has full context — yet reviewers of those same pull requests do not. Intent, tool calls, and rejected paths all sit inside the agent session, and the session closes the moment the pull request opens.

Someone is going to build the session-to-PR layer. Joe is convinced it should be him, and is willing to bear the cost of being right about the gap and wrong about the shape — which is precisely what the design-partner programme is for.

The design partnership

Four stages. Four to six weeks. A fixed fee of £20k. Not a subscription.

An engagement involves four weeks of Joe’s time, working alongside the team’s own agent-authored pull requests, building a prototype against a workflow slice the team chooses, and testing whether session-linked context belongs in their review, debugging, and audit flow. At the end, the team has a clear answer — yes, this should become a product, or no, it should not. Either way, the team retains the prototype and four weeks of senior product engineering against its own stack. That is what the £20k buys.

See the full engagement outline