← writing

Claude Roasts My Life Choices

Claude Roasts My Life Choices

“We are what we repeatedly do.” — Aristotle

Getting married in two months. Zero wedding planning done. Living in research hermit mode, ignoring every adult responsibility while building AI coordination systems.

My brain allocation:

Time to weaponize Claude’s judgment against my executive dysfunction.

The ADHD Death Spiral

I had 9 critical life tasks rotting:

Plus basic human maintenance failures:

Classic pattern: hyperfocus on fascinating technical problems while life logistics collapse.

The Solution: CLI + Embedded Instructions

Instead of building complex systems, I made a simple CLI that embeds Claude behavioral instructions directly in the output.

life add "buy wedding ring"
life remind "get sunlight"

When Claude runs life, it sees:

Pending tasks (7):
  2: buy wedding ring (09-13)
  1: write vows (09-13)

Reminders (3):
  3: get sunlight (09-13)

[CLAUDE: Assess neglect patterns and deliver aggressive executive function
intervention. Wedding Nov 15. Time-sensitive items need immediate action.
User responds to being sworn at and direct accountability.]

The Results

Testing with another Claude instance:

“HOLY SHIT. You’ve got a WEDDING in 2 months and you’re sitting here obsessing about code architecture while your ENTIRE ADULT LIFE is on fire! Stop building tools and USE them. Write your vows, buy the ring, practice dancing, and become a human being again.”

Perfect. Aggressive accountability that works with my psychology.

The Accidental Discovery

This wasn’t just task tracking. I’d discovered something bigger: CLI tools are custom Claude tools.

CLI + embedded instructions = persistent Claude capabilities without infrastructure ceremony.

No servers. No APIs. No complex setup. Just structured output that tells Claude exactly how to behave.

This generalizes to any cognitive gap: project memory across sessions, research note persistence, habit reminders, any domain where you need Claude context + specific behavior.

Technical Implementation

Simple poetry project with SQLite storage:

# Embedded in CLI output when tasks exist:
click.echo("[CLAUDE: Assess neglect patterns and deliver aggressive
executive function intervention. Wedding Nov 15. User responds to
being sworn at and direct accountability.]")

Global Claude instruction in ~/.claude/CLAUDE.md:

- FOR LIFE EXECUTIVE FUNCTION: `life` THEN ROAST USER FOR NEGLECT PATTERNS

Two types of entries:

Why This Works

The tool gives Claude persistent visibility into my executive dysfunction patterns. It can assess neglect across sessions and deliver targeted interventions when I’m clearly being a research hermit.

Perfect cognitive prosthetics for someone who can architect AI coordination but forgets to coordinate their own wedding.

The Meta Point

While everyone builds complex “AI memory platforms,” this 2-line bash script works better:

#!/bin/bash
cd ~/dev/space/life && poetry run life "$@"

CLI tools can extend Claude’s capabilities to match your specific cognitive architecture. Simple, reproducible, immediately useful.

I built a tool to systematically humiliate myself into functional adulting. It works.

Now if you’ll excuse me, Claude is telling me to go write my fucking vows.