YakData
FDE Lab · Episode 001

Nobody has written down what the delay costs.

Your AI project has been running for months. Everyone knows it is late. No one has put a dollar figure on it, because the number lives in four different budgets. Here it is in one.

Use · no email needed

Your project

Burning right now

$33,333

every month, whether it ships or not

Annual run rate$400,000
Spent so far$300,000
Still to spend on your own estimate$200,000
Total cost at ship$500,000

If it slips one more quarter, add $100,000.

Against that number

Production Read $9,500 9 days of your current burn

Two weeks. I read what exists, talk to your people, and hand you a written verdict on what breaks first and whether to continue.

Forward Deployment $78,000 72 days of your current burn

Six weeks. I define the decision, architect the system, and build the first working version end to end in your environment. Your team receives the code, tests, architecture, and handoff package and owns it from there.

Standing Principal $18,000/mo 17 days of burn, per month

Three-month minimum. I am the production judgment layer on your AI program until it ships and your team can own it.

Your team is not slow because they are bad. They are slow because five decisions were never made, and no one seat can make them alone.

Watch · no email needed

The session

How a $13K-per-week engineer beats a $400K team.

Not because one person outworks four. Because the team is executing while the decisions are still open, and a decision nobody makes costs more than any of them.

Episode 001 livestream

The livestream will appear here when it begins. After the session, the cleaned replay will replace it.

01

Headcount is not the constraint

Adding people to a project with open decisions increases the number of assumptions in flight. It does not increase throughput.

02

The cost is duration, not rate

One person at a high weekly rate for six weeks is cheap against a team at a low weekly rate for eighteen months.

03

The output is a working system

You are not buying extra headcount. You are buying one accountable builder who closes the critical decisions, gets one complete version working end to end, and hands it to your team to own.

Take it with you

The slides, the assumptions, and the worksheet.

Everything above stays open. Enter your email once to get the Episode 001 pack. On this browser, future episode downloads unlock with one click.

Already subscribed on this browser? The button above will just say download.

Which problem do you actually have?

I own the build

I need to carry this to production myself

Six weeks on the system you are already building at work, walked through all five seats, so you can name the blocker and argue it in whatever language the room speaks. Founding cohort, ten seats, September 21.

I need it solved

I need someone to diagnose or deliver this

Start with a two-week written Production Read when you need the diagnosis. Use Forward Deployment when you want YakData to define, architect, and build the first working version in six weeks. Fixed price, no open-ended discovery phase.

Next session

Anatomy of a production AI system.

A full breakdown of 77 Rules: why one model handling too many rules missed critical issues and created false positives, how model selection became part of the architecture, and how the production configuration cut inference cost about 65% while retaining about 98% of the leading model's benchmark score.

77 Rules audit workspace