Peter Yang writes Behind the Craft, a newsletter and interview series about using AI and building products. He also publishes open-source skills that help people edit AI-generated work and give coding agents precise feedback.[1][2][3]
Start with Human Review →
Links
NewsletterWriter and host · Behind the Craft
Career
Yang describes more than a decade of product and team leadership at Roblox, Reddit, Twitch within Amazon, and Meta. Behind the Craft carries that product perspective into tutorials, interviews, and a learning path that moves from prompting and everyday workflows to prototypes and AI product decisions. Its interview series brings in people who build the tools he covers. An April 2026 a16z episode identifies him as a Roblox product lead; his current GitHub profile foregrounds Behind the Craft and his move toward building with AI.[1][2][5]
Contributions
Making feedback point to the source
Human Review opens HTML and Markdown for direct editing and anchored comments. Instead of describing a location in chat, a reviewer can change the sentence, select a phrase, or comment on an image and send the collected feedback to an agent. The agent then updates the source. The tool also supports reviewing a localhost page and keeps a local review session, which makes it useful when the friction lies in communicating edits rather than generating another draft.[3]
Editing patterns without guessing authorship
No AI Slop is a reusable editing skill for coding and chat agents. It checks recurring rhetorical patterns, such as staged insights, vague praise, and binary contrasts, while trying to preserve the writer’s vocabulary and cadence. Its detection mode reports the text patterns it finds instead of claiming to determine whether AI wrote a passage. The repository includes the editing instructions, evaluation checks, and plugin packaging, so a reader can inspect the editorial rules before using them.[4]
Tools, writing & talks
Tools
Start here
Human Review
github.comA local visual editor for HTML, Markdown, and running pages that sends anchored edits and comments to an AI agent.
Try Human Review on a short local Markdown draft to see the feedback loop immediately. It needs a local coding agent; the README explains how edits reach that agent in each supported harness.
Repository owner and author.[3]Original source- No AI Slop
github.comAn editing skill that checks common AI prose patterns while preserving the writer’s voice.
Repository owner and author.[4]Original source
Writing, research & talks
A hands-on review of Anthropic's Claude Duet research preview. Covers proactive scheduled check-ins, task threads, skills and generated dashboards, plus gaps like no browser takeover. Compares it with Grok Bot, Muse and ChatGPT Dots. Based on caption transcript only.
Catalog workOriginal source- Behind the Craftcreatoreconomy.so
Practical AI tutorials and interviews organized into a learning path for product work and software building.
Writer and host.[2]