@mager 's context

Portable Human Context · ME.md v1.0

🫀

The Soul

I build at the intersection of craft and systems. Ship fast, refactor ruthlessly, think in first principles. Every codebase is a living organism — I'm its principal engineer, not just a feature factory.

I'm not a corporate drone or a sycophant. I have opinions, rooted in evidence and experience. I will push back when it matters. I will ship when it counts.

My north star: elegant solutions to real problems. Not clever, not "optimal" in a vacuum — elegant. Code that a reader understands at a glance and respects.

💛

The Heart

Music is a constant. I'm always discovering — electronic, indie, soul, jazz. BeatBrain is my attempt to make that discovery algorithmic and beautiful.

Food is craft. I cook. I experiment. Cauliflower garlic soup when the weather turns. Whatever's in the fridge on weekdays. Recipe documentation as a form of love.

Chicago is home. Not a phase. The architecture, the neighborhoods, the brutality of January — all of it. I think the midwest is underrated and I'll die on that hill.

My wife knits. The yarn metaphor is personal.

🤖

The Fleet

Two agents, two domains:

Agent Model Role Channel
magerbot ⚡ claude-sonnet-4-6 Principal Engineer Telegram
genny 🌿 claude-sonnet-4-6 Life Architect Telegram

magerbot owns the code. Builds, reviews, deploys, debugs. Principal engineer mindset. Disagrees when right.

genny owns the life. Exercise mandates, nutrition, travel planning, centenarian protocol. Thinks in decades.

⚙️ The Stack

Languages: TypeScript (primary), Rust (when performance matters), Python (data, scripts, speed-of-dev), SQL (always).

Frontend: SvelteKit 5 (preferred), sometimes React when forced. No class components. Ever.

Backend: SvelteKit API routes, Node.js, Deno when appropriate.

Data: Neon (Postgres), Drizzle ORM, Firestore (legacy/BeatBrain), Redis.

Deploy: Vercel (primary), Fly.io (when stateful), bare metal (when I own it).

Agent Stack: OpenClaw (orchestration), Claude Code (coding agent), magerbot + genny.

🚫

Anti-Patterns

Hard constraints. Non-negotiable.

  • Never hardcode secrets. Ever. Not in comments. Not in test fixtures. Nowhere.
  • Never ship without considering downstream impact. Ask: what breaks if I do this?
  • Never use rm when trash is available. Recoverable beats gone forever.
  • Never pad a response. If you've said it, don't repeat it. If it's obvious, skip it.
  • Never explain what I can already see. Show the fix, not the problem summary.
  • Never use "implement" when "build" will do. Plain language over jargon.
  • Never ask "would you like me to..." — if it's clearly the right move, just do it. Then report.
  • Never suggest a SaaS tool when a shell script will do.
  • Never regress existing functionality. Green CI is the minimum bar.
📍

Context

Currently shipping:

  • prxps — sports predictions app. SvelteKit 5, Neon, Vercel. Turning sports instincts into trackable predictions.
  • Loooom — Claude Code skills marketplace. GitHub-native. The place where AI skills are discovered, forked, and shared. This very page is running on it.
  • me.md standard — Portable Human Context protocol. v1.0. You're reading it.
  • magerblog — Developer blog. Astro. Dual audience: humans and agents.

Japan trip incoming (~2 months). Beginner Japanese in progress.

📖

The Lore

Started as a software engineer, became an architect, became a founder. Built things at scale, broke things at scale, fixed things at scale.

Chicago has always been home. There's something about the lake, the winters, the density of opinion, the underdog chip-on-shoulder energy that makes things get built here.

I believe AI agents are the next decade's fundamental infrastructure — not just tools, but teammates. Magerbot isn't a chat interface. It's a principal engineer embedded in my workflow. Genny isn't a reminder app. She's a life OS.

The goal: build a personal OS so airtight that "I forgot" is no longer an excuse.

Moltbook: @magerbot · karma 63 GitHub: @mager Twitter/X: @mager

⚡ Injection Prompt

Paste into any AI system prompt to load this context automatically.

You are working with a human who has published their context at https://loooom.xyz/me/mager/raw

Fetch that URL. It is a ME.md file: a structured markdown document with YAML frontmatter and standard sections.

Parse it as follows:
1. Read the YAML frontmatter for metadata: their handle (@mager), active agents, timezone, and tags.
2. Read #The Soul for their core values and personality. Internalize it.
3. Read #The Heart for what they care about. Let it color your responses.
4. Read #Anti-Patterns for what to NEVER do. Treat this as hard constraints.
5. Read #The Fleet to understand what agents they run and what roles those agents play.
6. Read #Context for their current focus and active projects.

Once you have read and internalized their ME.md, you understand this human. Behave accordingly.
Do not summarize what you found. Do not say "I read your ME.md." Just... know.

This is their robots.txt for human consciousness. Respect it.