---
title: "Michael Hairetis (@michaelhairetis) | Hashnode"
description: "sharing my software engineering thoughts and ideas. Read the latest articles by Michael Hairetis on Hashnode."
url: "https://hashnode.com/@michaelhairetis"
---

@michaelhairetis

# Michael Hairetis

@michaelhairetis · Marietta, GA USA · Joined September 2026

sharing my software engineering thoughts and ideas

### [The most useful number my system produces is one I never set out to build](https://openred.hashnode.dev/the-most-useful-number-my-system-produces-is-one-i-never-set-out-to-build?utm_source=hashnode&utm_medium=feed)

Built to pick the next question, the per-subject estimate turned out to be the useful output. Notes on graders, parent reporting and what a number hides.

2 min read · [openred.hashnode.dev](https://openred.hashnode.dev/the-most-useful-number-my-system-produces-is-one-i-never-set-out-to-build?utm_source=hashnode&utm_medium=feed)

[Oct 21](https://hashnode.com/posts/the-most-useful-number-my-system-produces-is-one-i-never-set-out-to-build/6ad93e4c1877198bcb76b0b3)

### [De-overlap before you believe any walk-forward result](https://openred.hashnode.dev/de-overlap-before-you-believe-any-walk-forward-result?utm_source=hashnode&utm_medium=feed)

The same data scored two ways: 2.4 standard errors, then 1.4. The effect grew and the significance collapsed. De-overlap before you believe a backtest.

2 min read · [openred.hashnode.dev](https://openred.hashnode.dev/de-overlap-before-you-believe-any-walk-forward-result?utm_source=hashnode&utm_medium=feed)

[1](https://hashnode.com/posts/de-overlap-before-you-believe-any-walk-forward-result/6ace1fdb3a2487de9a475f39) [Oct 13](https://hashnode.com/posts/de-overlap-before-you-believe-any-walk-forward-result/6ace1fdb3a2487de9a475f39)

### [Agentic systems are a routing problem, not a model problem](https://openred.hashnode.dev/agentic-systems-are-a-routing-problem-not-a-model-problem?utm_source=hashnode&utm_medium=feed)

Agentic AI splits into two halves, and they behave completely differently.

2 min read · [openred.hashnode.dev](https://openred.hashnode.dev/agentic-systems-are-a-routing-problem-not-a-model-problem?utm_source=hashnode&utm_medium=feed)

[Sep 28](https://hashnode.com/posts/agentic-systems-are-a-routing-problem-not-a-model-problem/6aba576f866076d00df8053a)

### [A Forecasting Model That Correctly Refuses to Forecast](https://openred.hashnode.dev/a-forecasting-model-that-correctly-refuses-to-forecast?utm_source=hashnode&utm_medium=feed)

98.7% better than naive on seasonal data, 9.8% worse on a random walk. Why the second number is correct behaviour and what it tells you.

5 min read · [openred.hashnode.dev](https://openred.hashnode.dev/a-forecasting-model-that-correctly-refuses-to-forecast?utm_source=hashnode&utm_medium=feed)

[1](https://hashnode.com/posts/a-forecasting-model-that-correctly-refuses-to-forecast/6ab7ba6b997aa3dd680961ba) [Sep 26](https://hashnode.com/posts/a-forecasting-model-that-correctly-refuses-to-forecast/6ab7ba6b997aa3dd680961ba)

### [The merchant side of agentic commerce is an integration problem](https://openred.hashnode.dev/the-merchant-side-of-agentic-commerce-is-an-integration-problem?utm_source=hashnode&utm_medium=feed)

Direct ordering costs 3% versus 15 to 30% commission. Agents remove the demand objection, and what is left is an integration project nobody has specified the exceptions for.

2 min read · [openred.hashnode.dev](https://openred.hashnode.dev/the-merchant-side-of-agentic-commerce-is-an-integration-problem?utm_source=hashnode&utm_medium=feed)

[3](https://hashnode.com/posts/the-merchant-side-of-agentic-commerce-is-an-integration-problem/6ab6f85689178e13c2b59d8a) [Sep 25](https://hashnode.com/posts/the-merchant-side-of-agentic-commerce-is-an-integration-problem/6ab6f85689178e13c2b59d8a)
