AE4JC on AI

Finish the Text Before It Forks

Fri, 31 Jul 2026 00:00:00 -0500

The rule: every editing pass — grammar, style, fact-checking, tone — happens on the source of truth, before the text is copied anywhere else. Once a copy exists on a second platform, that copy is finished. You publish it. You don’t improve it there.

Why

The moment the same piece exists in two places and you edit the second one, you have two partly-correct versions and no authoritative one. Merging them is not a merge. It’s a judgment call on every difference, made by someone who has to remember which change was an improvement and which was a mistake.

My First Month With a Second Brain

Thu, 30 Jul 2026 00:00:00 -0500

The morning it started — July 2nd, 2026. Many images in this piece are a real photo or screenshot from the month described.

One month ago, I didn’t have a second brain. I had a folder of notes, a habit of pasting things into a chat window, and an idea I’d read from two strangers on X. Today I have an AI agent that maintains a real, working knowledge base on its own — reads my notes, fixes what’s broken, catches its own mistakes, and has shipped four live websites along the way. It seems like a blur right now, but here’s a summary of what this first month actually looked like: the real wins and the real mistakes, in the order they happened.

Check the Bytes You're Actually Publishing

Fri, 31 Jul 2026 00:00:00 -0500

The rule: the check runs against the exact artifact a reader will get — not the draft it came from, not the file you were editing, not the version you remember approving.

That artifact is different depending on when you’re standing. Before publishing, it’s the built output that is about to be uploaded — the rendered page, not the source you typed. After publishing, it’s the live URL, because anything that happened during upload or rendering happened after your last look. Both checks are worth running; what they have in common is that neither one trusts the source file.

The Bottleneck Is Also the Inspector

Fri, 31 Jul 2026 00:00:00 -0500

The first version of this article argued that a publishing workflow runs at the speed of the steps a machine can reach, and that every step needing a human hand is a hard stop that chains the work to one specific device in one specific place.

I still think that’s true. It was also incomplete in a way I didn’t see until I tried to make the argument better — by asking an AI to find me a stronger version of it, and then checking what it handed back.

Treat a Graphical-Only Surface as Terminal

Fri, 31 Jul 2026 00:00:00 -0500

The rule: if a platform has no programmatic way to edit or remove what you’ve already published, treat it as a one-way destination. Send it finished work. Never plan to iterate there.

Why

There’s an assumption built into modern publishing that everything is editable. Most of the web is. When a platform isn’t, the cost doesn’t show up at publish time — it shows up at the first typo, weeks later, when the correction turns out to require a specific machine in a specific room.

Score a Platform Before Adopting It

Fri, 31 Jul 2026 00:00:00 -0500

The rule: before committing real work to a platform, find out which parts of the loop a program can reach — create, read back, update, delete. Do it first, while switching is still free.

Why

Platforms get chosen for what they can display. Reach, audience, formatting, how the finished thing looks. Those are the visible properties, and they’re all about publishing day.

Every cost after publishing day comes from a different set of properties — the ones I did not think to check: whether a script can read the piece back, revise it, or remove it. Those determine what the next year of maintenance feels like, and they’re usually discoverable in about ten minutes of reading documentation — but only if you think to look.

If a Fork Is Unavoidable, Automate the Diff

Fri, 31 Jul 2026 00:00:00 -0500

The rule: when the same content genuinely has to live in two places, the comparison between them is a script you run — not a careful reading you perform.

Why

Sometimes a fork can’t be prevented. A platform has no update path, an editor only exists on one side, a collaborator works somewhere you don’t. Fine. But the moment two copies exist, someone has to answer “what’s different, and which version is right?” — and that job is far more error-prone by hand than it looks.

Keep an Explicit "Needs a Desktop" List

Fri, 31 Jul 2026 00:00:00 -0500

The rule: anything blocked on a specific device, place, or physical step goes onto a written list the moment it’s found — with enough detail that doing it later requires no thinking, only hands.

Why

Some work can’t be done from wherever you are. That’s a fact of any system that touches the physical world or a graphical interface, and it isn’t a problem by itself.

The problem is rediscovery. Without a list, blocked work gets found at the moment you want it done, forgotten during the hours you’re away from the machine, and then re-derived from scratch the next time it surfaces. The thinking gets done two or three times and the task gets done once — if it gets done at all.

Affiliate Disclosure

Mon, 01 Jan 0001 00:00:00 +0000

Every article page on this site carries a link to AdBlob (adblob.com), my own gear storefront, and the footer links it too. AdBlob contains affiliate links — mainly Amazon Associates and the eBay Partner Network. If you click through and buy something, I may earn a small commission at no extra cost to you.

Articles on this site may also embed posts from X that themselves contain affiliate links. Those posts carry their own disclosure where they were published.

Privacy Policy

Mon, 01 Jan 0001 00:00:00 +0000

Last updated: August 17, 2026.

A plain-language description of what this site actually does with data, not formal legal counsel.

What This Site Collects

Comments. Articles on this site use a comment system that requires a verified email address to post (no anonymous commenting). When you leave a comment, the system stores your email address, the display name you choose, your comment text, and technical data (like your IP address) used to prevent spam and abuse. This data is stored on a server I run myself — it is never sold, and never shared with anyone except as required by law. You can ask to have your comments and account data deleted at any time — the comment system has a built-in “delete my data” option.

How I Built a Working AI-Maintained Second Brain

Tue, 07 Jul 2026 00:00:00 -0500

A few weeks ago, I didn’t know a system like this was realistically possible. Today, I have it running on a dedicated mini-PC, with the AI actively maintaining my Obsidian vault according to my custom guidelines. Here’s how it started.

The spark came from an X Article discussing how people increasingly need help remembering and organizing information in an age of scattered notes, browser tabs, and forgotten chats. The core frustration it highlighted was that AI tools forget. They don’t build on previous work.

Hardware Choice + Day 1 Security Setup for My AI-Maintained Second Brain

Tue, 07 Jul 2026 00:00:00 -0500

Last time I wrote about the guardrails I put in place before letting an AI maintain a real knowledge base. This time: the actual hardware, and how I approached the first day of setup.

Picking the hardware

I used Grok to narrow down the options for the use case and settled on a MINISFORUM UM870 Slim Mini PC (affiliate link) — Ryzen 7 8745H, 32GB RAM, 1TB SSD.

Locking Down Remote Access for My AI-Maintained Second Brain

Sat, 01 Aug 2026 00:00:00 -0500

Sterile mockup, not a real capture — an actual tailscale status screenshot would expose personal network info.

The Second Brain system I’m running today didn’t start with a plan for dedicated, isolated hardware. I was thinking hosted space at first.

But as Grok and I continued to weigh the options, we kept coming back to dedicated, isolated hardware.

The File Structure for My AI-Maintained Second Brain

Sun, 02 Aug 2026 00:00:00 -0500

Last time I covered locking down remote access and the phone-side failures that came with it. This time: designing the vault itself. People picture a “second brain” as something elaborate — nested tags, custom plugins, a taxonomy years in the making. Five folders is all it took.

That’s the whole point of this piece. The structure is small on purpose, and the one part of it I tried to maintain by hand is the one part that broke.

The Retry Loop Was the Real Problem, Not the Platform

Mon, 03 Aug 2026 07:00:00 -0500

If you’re new here, this whole thing started as a chat with Grok. Then it became Claude. Then it became Claude running an actual second brain — a vault of my own notes, maintained by an AI I trust with real decisions. I’m writing these while I’m still living them, not from a finished vantage point looking back. This one: a real mistake, and it wasn’t the platform’s.

A highway lit up like a circuit

The best description I’ve found for what this actually feels like: AI itself moves like electricity — near light-speed, the answer is already there the moment you ask. But using AI across real tools and platforms feels more like driving. Highways connect one system to another, and the APIs between them are tolls at the gates — pay the fee, cross the gate, keep moving fast on the other side.

The Real AI Limitation Isn't Missing Capability — It's the Gates Between Tools

Tue, 04 Aug 2026 07:00:00 -0500

Now that I have a working AI-maintained second brain, I’ve become a bit spoiled.

If you’re unfamiliar, here’s where this started. The short version: a dedicated mini-PC runs Claude as a permanent assistant that reads, writes, and remembers a structured set of notes — everything I capture, decide, and build — entirely on its own, between sessions. It’s not a chatbot you visit; it’s a system that operates continuously under a set of rules I wrote for it, and it has held up through a month of real use. Outside that boundary is a different world entirely, and living between the two is what this piece is actually about.

Why a Mini-PC, Not a Hosted Server

Wed, 05 Aug 2026 07:00:00 -0500

Somewhere around Part 2, a fair question shows up: why run an AI-maintained knowledge system on a mini-PC sitting in my house, instead of renting space in a real data center with professional security and fire suppression? It’s a good question. Here’s how I actually decided — not the tidy version after the fact, but the method itself, since the method is more useful to you than my specific answer.

The method: Benjamin Franklin’s weighted pros and cons

In 1772, Franklin wrote a letter to Joseph Priestley describing what he called “moral or prudential algebra.” The idea, in short: don’t just list pros on one side and cons on the other and count which column is longer. Instead:

My Wife Called Me a Vampire

Mon, 17 Aug 2026 00:00:00 -0500

My wife called me a vampire before I ever saw the clip.

Marc Andreessen — the “software eats the world” guy — was on Joe Rogan’s podcast describing what AI coding is doing to developers’ hours: people working more, not less. I think it’s because the opportunity in front of us feels like a now-or-never window. There’s already a name for it circulating in the industry: the “AI vampire” — up all night because you’re too productive to shut it off.

The Guardrails File That Grew Itself

Thu, 06 Aug 2026 07:00:00 -0500

Last time I covered designing the vault itself. This time I’m going back to something from the very first installment: the guardrails file — the plain text document that tells the AI maintaining my notes what it may and may not do on its own. Part 1 summarised the core rules I started with as five plain guidelines. What I didn’t say then: those five were a plain-English summary of seven numbered rules, and the file didn’t stop at seven.

The Memory Fix That Broke Its Own Alarm

Mon, 31 Aug 2026 04:16:00 -0500

My AI Second Brain runs on a dedicated machine, and Claude Code Remote Control is the service that keeps every session alive on it. A few weeks ago that service started dying, the same way, over and over: memory climbed for hours, then the whole box went into swap and everything froze at once. Not one session. All of them.

The fix looked simple. Give the service a hard memory ceiling, so a runaway session gets killed on its own instead of taking the whole machine down with it. I set MemoryMax to a real number and moved on.

About

Mon, 01 Jan 0001 00:00:00 +0000

I’m Matthew Foutch (AE4JC) — a Senior Manufacturing Engineer with a multi-decade career in hands-on process engineering. I built SecBrain, an AI-augmented personal knowledge system, to apply lean manufacturing discipline (5S, root cause/corrective action, continuous improvement) to knowledge work and workflow automation, using Claude (Anthropic) and Grok (xAI) as working agents.

This site is the public record of that system: dated accounts of milestones and mistakes, and the operating principles that came out of real incidents. It exists so someone arriving from my LinkedIn project — or anyone evaluating whether AI can be run with professional accountability — can see the work, not just a claim.

What's New

Mon, 01 Jan 0001 00:00:00 +0000

Everything on this site, newest change first. New pages and revised ones both show up here — this list is generated from the site’s own history, so nothing has to be remembered onto it.

Prefer to be told instead of checking? The RSS feed carries the same thing, and there’s a link in the footer of every page.