The last day I didn’t make a commit to a GitHub repository was November 13th. That’s a 37-day streak. My GitHub contributions for 2025 now stand at 5,335. That heat map is getting dense.

I haven’t updated the log in a while, but I’ve been heads-down on Rentizens. If momentum holds, it launches in January as a controlled “friends and family” release. Real users, real feedback, real learning. That’s the only way to test assumptions.

The Custom Workflow Stack

Here’s what’s actually driving the output: custom slash commands that let me methodically roll through PRDs and stories without losing context.

The workflow looks like this:

• /intake - Takes a PRD or idea and structures it
• /plan-story - Creates an implementation plan with TDD phases
• /implement-story - Executes the plan, tests included

A now.md doc gets continuously updated through hooks, capturing progress. This is the key insight: I’m trading tokens for consistency and continuity. Every checkpoint costs tokens, but it means I can restart any session without losing context.

The screenshot below shows Story 014-1 being planned. It’s a pure data model slice—perfect for TDD. The plan breaks down into five phases: write tests first (TDD Red), create migration, implement methods (TDD Green), verify, update docs. Even small stories get this treatment.

Below that, you can see the custom prompt commands I’ve built. Each one serves a specific purpose in the factory.

The PR Velocity

Look at the screenshot below. That’s 164 closed PRs in Rentizens. The list shows PRDs being sliced into stories, stories being implemented in sequence: 016-1, 016-2, 016-3… 017-1, 017-2… 018-1, 018-2…

Many of those PRs were merged within hours of each other. A few days ago, I was averaging maybe 8-10 PRs per day. This velocity would have been unthinkable six months ago. Not impossible—unthinkable. The mental model of what’s achievable has shifted.

The UI Consistency Problem

I still haven’t cracked great UI consistency. I’ve been experimenting with MagicPath.ai, Google’s Stitch, Gemini 3, and Claude Code’s frontend-design skills. They all serve different purposes, but I haven’t found the right combination.

Part of the problem: I don’t want to slow down mainline development to figure it out. That’s probably flawed thinking. Sometimes you have to slow down to go fast.

One thing I did figure out: I can host mock indexes for MagicPath designs on GitHub Pages. Link them together so they kinda-sorta flow. It’s scaffolding that won’t see production, but it helps me move forward on design decisions.

This is the “factory IP” idea—building tools and scaffolding that compound yourself, even if they’re throwaway.

Plenty of SaaS

Speaking of factory IP, I built out plentyofsaas.com as a holding company site. Maybe 2-3 hours total, mostly iterating on content.

That’s the leverage: infrastructure that used to take days now takes hours. Projects that used to take weeks now take days. The compounding is real.

What’s Next

Rentizens heads toward a January launch. Controlled release. Learn fast. Iterate.

The custom workflow stack continues to evolve. Every friction point is a potential hook. Every repeated task is a potential slash command.

Trade tokens for consistency. Build the factory. Ship the product.