Keep pulling the thread on Josh Pigford.
Josh Pigford developed a "learnings" skill that analyzes his coding sessions and conversations with the AI to distill insights and automatically update a project's Claude file, preventing the AI from repeating mistakes.
In his development workflow, Josh Pigford uses the Opus model for the initial build and planning, followed by an adversarial review pass using GPT-5.5.
Josh Pigford states that using GPT-5.5 for an adversarial review of code written by Opus invariably finds 3 to 5 bugs that Opus overlooked.
Josh Pigford sold his previous startup, Baremetrics, for $4 million.
As a solo founder, Josh Pigford is currently building at least five AI products in parallel.
Josh Pigford launched a product called Proxy User, which uses synthetic users operating actual browsers to perform quality assurance on web applications.
Josh Pigford launched a product called Rumored that monitors for and identifies when large language models hallucinate information about a specific brand.
Josh Pigford's development methodology involves using a custom, open-source "Build" skill on GitHub to manage research, planning, and implementation phases for new features.
Josh Pigford uses Git work trees as self-contained, shippable units that function as checkpoints, allowing him to roll back changes and prevent context rot in the AI's working memory.
Josh Pigford uses the Conductor development environment because it allows for easy switching between different AI models and provides automated management of Git work trees.
Josh Pigford uses a custom "But For Real" skill that "bullies" the LLM into re-reviewing its own work, which typically finds an additional 3 to 5 bugs.
Josh Pigford's product strategy is to build the application first and then use the final feature set to generate the marketing site and copy.