I haven't been absent. I've been building.

It has been nineteen days. I went into a work induced fugue state advancing my personal projects. Here's what I was working on:

Alpha-Lab is a quant research pipeline I designed to answer a consistent and broad set of recurring questions. So far it covers most of the pipeline: optimization, backtesting, attribution, feature engineering, regime study, execution and system diagnostics. The core problem it solves is one I kept stumbling over. Every new research idea spawned a new repo, a new set of conventions, a new architecture, and repeat work over the same core components. Alpha-Lab unifies them under one umbrella. I work on it daily and keep adding features. Expect to see more content where I demonstrate its use, and limitations.

Sports Prediction Machines, similar in spirit but smaller in scale in comparison to Alpha-Lab, it was designed to implement a similar pipeline except for sports. I started with NFL, College Football, NBA, and now MLB. It's currently on version 2. Version 1 had most of what I wanted: strategy selection, backtesting, portfolio selection and analysis. However the biggest problem was trust. Too many suspect outputs and bug fixes revealed too many gaps in the audit chain. A result that I couldn't audit was a result I could not trust. Garbage in Garbage out with extra steps. Version 2 was a rebuild from the foundation up focusing on auditability and correctness from the start.

Untilt is the most unique of the three. It is a behavioral fitness app for discretionary traders. The premise being that the high failure rate of discretionary traders is not due to lack of knowledge or strategy, but a lack of repeatable, measurable processes that form a true system. Sending buy and sell orders are only one part of a successful trading operation. The greatest discretionary traders have already encoded these structures that makes their judgement reliable. Most traders have not. Untilt is that structure: a trading archetype that maps your likely strengths and weaknesses, and a daily scorecard that tracks the habits most likely to undermine your edge. Just went live on the android store. More on this one later in the week.

Three different projects, one thesis underneath all of them: reliable judgment requires infrastructure. Whether the domain is quant research, sports prediction, or discretionary trading, the edge is not just the idea. It is the operating structure that makes the idea measurable, repeatable, auditable, and improvable. I'll unpack that thesis more this week.

These three systems birthed the structure of the Forge suite. It wasn't a curriculum I designed in the abstract, but a distillation of lessons learned when I built with greater understanding and focused intention. Forge Learn is the deployed pipeline. Forge Ship is the framework that keeps it trustworthy.

One note on Forge Learn: early bird pricing is open through Friday. I did not promote it the way I planned. I was building. That is where the time went. If you have been on the fence, this is the window.

→ Forge Learn — six sessions, live cohort, deployed pipeline in your own AWS account
→ Forge Ship — production skeleton, engineering guardrails, deploy configs. No sessions, just the stack.