AI For Dev

simon-scrapes

3 min read

Simon Scrapes

YouTube creator focused on agentic systems for business, with practical Claude Code workflow content. Runs the Agentic Academy on Skool — a paid community with a “full Claude Code track” that teaches building with Claude Code, sub-agents, skills, and design patterns. The wiki tracks Simon as the primary source for the llm-design-patterns thread — his Claude Code workflow taxonomy is the seed source for the new patterns library page.

Stub — first appearance via the workflow patterns video. Will grow as more sources land.

Channels

  • YouTube: Simon Scrapes — agentic systems and Claude Code
  • Community: Agentic Academy on Skool — paid, with a Claude Code track

Content in This Wiki

  • Every Claude Code Workflow Explained (& When to Use Each) — Walkthrough of 5 named Claude Code workflow patterns (Sequential / Operator / Split & Merge / Agent Teams / Headless) with explicit when-to-use and when-not-to-use guidance. Anchors the wiki’s new llm-design-patterns concept page. Also documents: the claude -w git work-trees flag, the claude -p headless flag, the 10 concurrent sub-agent limit, the Agent Teams experimental research preview shipped with Opus 4.6 (CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1), and the 3 always-on built-in sub-agents (Explore / Plan / General-purpose).

Key Ideas

  • The pattern selection ladder — five Claude Code workflow patterns aren’t alternatives but a progression. Climb only as high as the task requires: Sequential → Operator → Split & Merge → Agent Teams → Headless. Most users default too low (sequential one-conversation flow) and never explore the higher rungs.
  • “If you’re still using Claude Code one conversation at a time, you’re using it wrong.” — opening framing
  • Hub-and-spoke vs peer-to-peer is the architectural distinction between sub-agents (Pattern 3) and Agent Teams (Pattern 4). Sub-agents can only report to the main agent; agent teams can message each other directly via a shared task list.
  • Agent Teams are 4–7× more tokens than a single session. Only worth it for genuine cross-discipline coordination (front-end + back-end + test that have to talk as they build). Most users should never reach for them.
  • Headless (claude -p) becomes powerful when scheduled. Plug into cron / Mac Task Scheduler / Windows Task Scheduler and Claude runs autonomously on any cadence. The bridge from tool you sit with to team member that works independently.

Editorial framing the wiki applies to Simon’s content

  • Simon plugs his own Skool community at the open and close of his videos. First-party self-promotion, not undisclosed affiliation. Treat capability claims as descriptive (the video is content-dense and explicitly emphasizes when not to use the more advanced patterns), and the Skool plug as the conversion call-to-action.
  • His framing is action-oriented and tool-pragmatic — closer to Nate Herk in tone than to Nate B Jones’s strategic analysis. The wiki uses Simon as the pattern taxonomy source rather than the strategy source.

See Also

Graph View

Backlinks