Anyone can use Claude Code. Boondogglers conduct it.
Claude solves faster than any human. That gap will not close — it will widen. What will not change: Claude cannot verify its output against domain reality, cannot reframe a poorly formulated problem, cannot supply accountability when the build ships.
The human’s job is not to type less. It is to decide more precisely — what to build, in what order, and what counts as done. That is conducting. Boondoggling is the practice of learning how.
Boondoggling, frick-fracking, noodling.
While Claude thinks, the status line cycles through silly words — boondoggling, cogitating, lollygagging, spelunking, meandering, pondering. This site got its name from those terms. We liked the vibe.
Some of them turned into real concepts:
- Boondoggling
- A boondoggle is a structured build plan that Gru generates alongside a detailed Software Design Document. It splits the work into Minion steps — the exact prompts that Claude Code or a similar tool needs to run, in the required order — and Gru steps — a step-by-step recipe of what the human needs to do and how to check and verify each step ran correctly.
- Frick-fracking
- The process of making small, iterative changes to a working app. Right now I’m frick-fracking — adding this very text to the site.
- Noodling
- Dreaming up new features over coffee, at the gym, or listening to music. Noodling is where builds start — before a single prompt gets typed.
Software engineering, done well, is conducting.
The five things only you can do.
Plausibility Auditing
Hear the wrong note before you recompute it. Claude produces fluent output that may be structurally unsound — you catch what passes every syntax check.
Problem Formulation
Rewrite the question so the answer becomes findable. Claude optimises within the frame you give it — you decide whether the frame is the right one.
Tool Orchestration
Choose which tool runs, in what order, with what constraints. Claude executes a single turn — you sequence the pipeline across turns, tools, and contexts.
Interpretive Judgment
Decide what the output means for this situation, this user, this deadline. Claude returns text — you supply the judgment that text alone cannot carry.
Executive Integration
Hold the whole build in your head. Claude sees one prompt at a time — you hold the architecture, the tradeoffs, and the accountability across all of them.
Meet Gru.
Gru is a system prompt for Claude that builds Software Design Documents and generates a Boondoggle Score — a sequenced, dependency-ordered plan that separates what Claude builds from what you build.
You run it in your own Claude Project. No account on this site, no API key, no paywall. The videos show you how.
Real builds. Real SDDs. Real Boondoggle Scores.
Every doc in /dev is a real build produced using Gru. Browse by category — websites, agents, games.