Why not just hire someone to do this in-house?

A full-time AI engineer is a $180K+ commitment and a 3-month hiring cycle, and you still have to know what to point them at. We deliver the scoped roadmap in a week and the first working system in weeks — and the retainer is a fraction of a headcount. When you have enough live systems to justify a hire, you'll hand them documentation and code we built to be owned, not a black box.

Why can't we just use ChatGPT or Copilot ourselves?

You can, and you should — for drafting and Q&A. But a chat window doesn't read your scheduler, reconcile your invoices, or follow up with a customer at 7am without anyone clicking a button. The value is the plumbing between your tools, the edge-case handling, and the monitoring — the part MIT found 95% of companies never finish.

What if the audit finds nothing worth automating?

Then we tell you, and we refund the audit — money back. We'd rather lose the fee than sell you a build that won't pay for itself.

Fixed, not hourly. AI Operations Audit $1,500–$2,500 (credited toward a build). Build Sprint $8,000–$15,000. Ops Retainer $3,000–$8,000/month. You always know the number before we start.

Who owns what we build?

You do — source code, prompts, configuration, all of it, from day one. If you fire us tomorrow, the system keeps running.

50 Commits, One Mission: Making Motus Smarter About Your Body

Sometimes a week looks like 50 small commits but really boils down to one hard problem. This week was that kind of week — I spent most of it fixing how Motus calculates your Total Daily Energy Expenditure (TDEE), which is basically the engine that powers everything else in the app.

The TDEE Calibration Problem

Motus tries to calibrate your TDEE based on real data — what you eat, what you weigh, how you move. But the calibration wasn't pulling your Resting Metabolic Rate (RMR) correctly. It was either using stale data or skipping the lookup entirely, which meant the whole calculation was off.

The root issue: inconsistent RMR resolution across different parts of the codebase. Some places were calling it directly, others were using cached values, and the calibration flow wasn't prioritizing measured RMR over estimates. Classic case of "it worked in isolation but broke in context."

I consolidated all RMR lookups through a single resolver function that knows the priority order: measured RMR first, then fall back to estimates only when necessary. Then I routed the remaining stragglers through that resolver. The calibration now sees your actual measured RMR when it exists, which makes the whole TDEE calculation way more accurate.

This touched a lot of files — hence the 50 commits — but the core fix was about establishing one source of truth for metabolic data.

Recovery Score Tweaks

While I was neck-deep in metabolic math, I also patched a bug in the recovery score calculation. It was overcounting certain activity types and throwing off the daily recovery recommendations. Nothing dramatic, just tightening up the logic so the score reflects actual strain more accurately.

Shipped: Revamped ruhrohhalp Site

Outside of Motus work, I finally pushed the redesigned ruhrohhalp site live. It's been sitting in "almost done" limbo for weeks. Spent a day on final QA — checking responsive breakpoints, fixing mobile nav quirks, making sure all the links actually go somewhere useful.

The new design is cleaner, loads faster, and actually explains what ruhrohhalp does instead of assuming people already know. Small win, but it's nice to have one less thing in the "needs to ship" column.

What's Next

Now that the TDEE calibration is solid, I'm moving on to nutrition logging improvements. The current flow works but has too much friction — I want to cut the time it takes to log a meal in half. That means smarter defaults, better autocomplete, and possibly ditching a few input fields that people rarely fill out anyway.

Also planning to add trend visualizations for the metabolic data now that the underlying numbers are trustworthy. Graphs are only useful when the data is right.