Field notebook · v.01

I build usable systems from what is already there

We start with what already exists (tools, data, workflows, instincts, constraints, the spreadsheet everyone is afraid to delete), find the logic doing the real work, name the assumption causing the break, and leave behind the smallest system the team can actually run without me.

Same brain, different rooms. CPG pricing and margin. AI pilots that quietly underperform. Founders with too much instinct and not enough structure. Communities whose default user was wrong from day one.

See the proof Bring the messy version

Diagram · 01

Built from what was already there

1
Messy room. Tools, data, instincts, spreadsheets, half-built rules.
2
Stated problem set aside. Real one named out loud.
3
Logic already doing the work gets surfaced.
4
Smallest usable system gets built. It survives without me.

If the loop doesn't survive without me, it isn't a system.

§ 02 · The pattern

The stated problem is usually not the real problem.

The work is repeatable because the shape of the failure is repeatable. Different industry, same six moves.

01Set the stated request down. Name the real decision underneath it.
02Inventory what already exists. Tools, data, reports, workflows, people knowledge, the ugly-but-important spreadsheet.
03Find the logic secretly doing the work. Usually it lives in one person's head or one tab nobody opens in front of leadership.
04Name the assumption causing the break. Default user, default buyer, default workflow, default forecast.
05Borrow the method from wherever this shape of problem has already been solved. Not best practice. Transferred logic.
06Build the smallest usable system that survives without me.

§ 03 · Where it shows up

Same brain, different rooms.

What I usually see →

Pricing, margin, and market logic
CPG pricing files, promo plans, cost flow, buyer logic, commodity moves. One outdated assumption costs real money every week.
AI where the workflow is already weak
AI does not fix a broken workflow. It speeds it up. Before you automate the workflow, make sure the workflow isn't stupid.
Founders with too much instinct and not enough system
The signal is already in the founder's head. It is not yet in a form anyone else can run, fund, or repeat.
Communities built around the wrong default user
Mission-driven work usually knows the friction. The operating structure to make that friction usable, fundable, and buildable is the missing piece.

§ 04 · Proof, by pattern

The useful part was usually already in the room.

See all five proofs →

Proof / 01 · Real system behind the stated problem

The $120K study that was already sitting in the company's own data

Context: A roughly $1B protein processor was quoted a licensed elasticity study to answer a pricing question.
Hidden system: The inputs were already in house. The dependency was on a vendor's method, not on missing data.
What changed: Derived the answer from first principles, landed within 0.05% of the licensed result, and folded it trade spend calculators.
Public result: The model is no longer the artifact. The pricing process it seeded is.

Proof / 02 · Scattered knowledge turned into usable logic

Bacon pricing across a roughly 95M-lb portfolio, run on stale assumptions

Context: Pricing ran on static cost thresholds and lagging forecasts against a belly market that did not care about either.
Hidden system: Buyers, planners, and finance each held a piece of the logic. The market was treated as something to react-to vs for.
What changed: Pulled scattered buyer logic, cost-flow rules, and promo behavior into one inventory-weighted, market-reactive pricing process on a weekly cadence.
Public result: Bacon pricing increased by $0.19/LB Value translated across the portfolio, repeatable inside a loop the team ran long after me.

§ 05 · In public

The same engine runs in writing.

An oil thesis that called $100 Brent ahead of consensus. A beef piece that named producer-level incentive failure before the trade press. A commodity analogy drawn from astrophysics. Same move every time: read the signal in public data that other people were already looking at.

Read on Substack →

§ 06 · Start here

Bring the ugly spreadsheet.

The pretty one is probably hiding the problem. Bring the half-built model, the contradictory dashboard, the AI pilot that kind of works, the slide nobody wants to present. That is the raw material.

Start with the messy version →