About

You might be here because one of these is eating your evenings

• You run a one-person media operation, and the pipeline — sourcing, writing, illustrating, publishing, distributing, tracking — has quietly taken over every night after dinner.
• You wired up an AI automation flow that runs fine twice and breaks on the third try. No one tells you which link in the chain is actually fragile.
• You're making video or short-form content, and a single post costs you scripting, B-roll, cuts, voiceover, captions, uploading — a full working day per clip.
• You're running cross-border e-commerce or growth funnels, and every channel — listings, keywords, creatives, ads, landing pages — demands attention before the other channels will start working.
• You're still cleaning up Excel, PPT, PDF, scraped data, weekly reports by hand, and you've started wondering how a person your age ended up doing data-entry for a living again.

I've crashed into all of these. AWP is where I write down what's actually left standing after those crashes, so you don't have to run every experiment yourself.

What you'll find here — five tracks

Track 1 — The one-person media stack
The full pipeline a solo creator actually needs: sourcing, long-form and short-form writing, AI illustration, multi-platform publishing, distribution, and the analytics feedback loops that tell you what to write next. The point isn't "save time." The point is: run a media company without hiring.

Track 2 — Video and short-form at scale
Explainer videos, storyline mashups, product ads, TikTok and YouTube shorts — from script to export, end-to-end. Not "how to use CapCut." The pipeline.

Track 3 — E-commerce and growth engines
Cross-border listings, product videos, keyword matrices, Google / Meta / TikTok ad systems, landing page conversion — the full stack from sourcing a product to shipping your first order.

Track 4 — Office automation and data collection
Batch Excel, PPT, and PDF processing, all-platform scrapers (YouTube, Twitter, 小红书, WeChat, Douyin, Reddit, and more), automatic report generation. The goal is to take the repetitive work off your hands entirely.

Track 5 — End-to-end AI automation with Claude Code, Skills, and multi-agent
Which processes Claude Code can actually replace. Which judgment calls it's trusted with. Which setups blow up once you scale past three runs. Deep teardowns with the complete SKILL.md and configs — not a five-minute demo.

What's shipping next

The work-in-progress pile. Released one piece at a time:

① The AWP Skill library — around 100 production-grade Skills. Claude Code Skill teardowns that span every track above. Each one is a Skill I've actually run on real tasks: the complete SKILL.md, the step docs, the scripts, and the bugs I hit on the way. Not hello-world examples.

② The AWP CLI toolbox. Python and shell utilities you can drop into real work — cross-platform publishing, all-platform data collection, media processing (images, audio, video, OCR), email automation, cloud storage (R2 / COS / S3), ad platform controls (Google / Meta / TikTok), CMS management (Ghost, WordPress, FlowUS). Each tool ships with the deploy guide, the credential model, and the traps I fell into.

③ Multi-agent coordination, from scratch. Going from a Claude Code solo operator to a ten-agent internal "company": department structure, task pooling, agent memory, handoffs, cross-machine deployment. Not a framework pitch. An actual working coordination backbone, documented from two agents up to ten.

④ The Claude Code deep-dive series. One production workflow per post, torn down to the bones: SKILL.md design, Hooks orchestration, SubAgent calls, Worktree isolation, MCP integration, plugins, cost control, cross-model validation. The 30,000-word Skill development guide that's already out is the first entry in this series.

⑤ Research breakdowns, made actionable. MIT, Stanford HAI, Anthropic papers — translated out of academic register and into "here's what this actually changes in your workflow this week."

⑥ The solo-operator playbook. The top-level operating system that ties the six tracks together: editorial strategy, content factories, brand laddering, distribution patterns, data feedback loops, growth leverage.

Release cadence: each piece ships when it's actually tested, not on a calendar. Some weeks it's a short tool note. Some weeks it's a 30,000-word field guide. Whichever Skill ripened first that week is the one that goes up.

Who this site is for — and who it isn't

For you, if: you're a technical writer, an indie creator, an indie maker, a marketer tired of fighting for each keyword, a cross-border e-commerce operator, a solo media operator, or anyone building their own knowledge-driven business.

Not for you, if: you want "What is ChatGPT" explainers, a "10x your output" sales pitch, or someone to outsource the actual work to. That content lives everywhere else on the internet. It doesn't live here.

Why my take is worth the time

I've been chasing the same question for years: how do you make a thing run without sitting here clicking through it every time? Chasing that question took me through four generations of tools — SaaS low-code, self-hosted open source, AI coding, and now the early shape of multi-agent coordination. The tools keep churning. The question underneath them never does.

Every wave left me with something that looked impressive at the time and looks naive in hindsight. What's worth reading isn't the impressive parts. It's what I threw out, why I threw it out, and what's still standing.

Two rules I write by

• If I haven't personally broken it, I won't write about it.
• What lands in a post is what survived in my setup last month — not what everyone on Twitter is doing, not what the launch-day demo promised.

Where to go next

Subscribe to the newsletter — one email a week. Whenever you've built enough of your own workflow across these tracks, unsubscribe with no hard feelings. Clean break.

Something wrong, or I missed a case? Email me at [email protected].

— Leo