The Tech Debtor

A Vault That Talks Back

Sun, 19 Jul 2026 00:00:00 +0000

Part 3 of a series on wiring my Obsidian vault into Claude. Part 1 · Part 2

The last two posts covered the capture side — the various front doors that get content into the vault. This one is about the other half of the brain: Claude reaching into the vault directly, reading and writing it live, in the middle of a conversation. This is the half that finally solved both halves of the problem I complained about back in Part 1.

Feeding the Machine

Sun, 12 Jul 2026 00:00:00 +0000

Part 2 of a series on wiring my Obsidian vault into Claude. Part 1

In the first post I said the system splits into two halves: the capture-and-batch side that pulls stuff in and tidies it, and the interactive side where Claude and I work in the vault together. This post is about the first half — specifically, getting content in.

Here’s the belief the whole thing rests on: the moment you capture something and the moment you organize it should never be the same moment. Every time I’ve made capture depend on also filing the thing correctly, capture loses. The thought evaporates while I’m deciding which folder it belongs in. So the job I handed to Claude is exactly that deciding — where a thing goes, what it’s called, how it’s tagged — which frees capture to be as dumb and frictionless as possible.

Teaching My Second Brain to Talk Back

Mon, 06 Jul 2026 00:00:00 +0000

Part 1 of a series on wiring my Obsidian vault into Claude.

A while back I wrote about falling down the homelab rabbit hole — one Raspberry Pi turning into a dedicated server turning into fifty-some Docker containers. I’d like to report that the rabbit hole has a bottom. It does not. It just moved up the stack, from containers to something closer to cognition.

This is the first post in a short series about how my Obsidian vault — my “second brain” — stopped being a static pile of Markdown files and started reading and writing itself, with Claude as the intelligence layer in the middle. This post is the map: the why, and where the series is headed.

Homelab DNS Configuration

Sat, 09 Aug 2025 00:00:00 +0000

I’ve previously discussed how I route external traffic to my homelab and how I route traffic that has reached my homelab. One outstanding issue was the following: I wanted to be able to have internal-only resources, accessible only locally or through a VPN, along side my externally visible resources (like this blog!). In addition to the privacy of internal-only resources, I wanted to reduce traffic through my VPS. One example is Jellyfin - I didn’t want local traffic to route out to my VPS, only to be routed back to my homelab to stream movies.

Replacing Google Photos

Wed, 18 Oct 2023 00:00:00 +0000

One of my goals as I’ve been working on my homelab has been to replace Google Photos. For me, there have been main reasons to find an alternative:

Privacy - more and more I dislike the idea of my photos being hosted somewhere I don’t control
Space - I’m starting to hit the limit of free storage on Google Photos

As a replacement, I wanted something that provided some of the same features I’ve come to know and love about Google Photos, but in a self-hosted format:

Rust and RISC-V Ramblings

Tue, 10 Oct 2023 00:00:00 +0000

I started my programming career with pretty run-of-the-mill programming languages - C++, Java, Python. Then, one of my professors had a challenge for the class: complete the assignment in Haskell. I was curious, so I jumped in - and loved it! I hadn’t realized there could be such a vastly different way of thinking about programming, and this set me on a path to continuously dabble in random languages.

I took a Coursera course on Scala, was introduced to Akka and the Actor Model through that. That lead me to Elixir and Erlang, which I really enjoyed. I’m still convinced Akka, Elixir, Erlang, and the Actor Model are some of the most solid ways to develop large, distributed software systems.

Managing Containers with Docker Compose

Wed, 27 Sep 2023 00:00:00 +0000

My homelab current consists of about 50 services - 30 or so exposed through Traefik, and then roughly 20 backing services (databases, etc). All of those services run in Docker; all have different lifecycles, requirements, start-up scripts. If I were using barebones Docker, I would probably have 50 scripts lying around to maintain those containers. Instead, I use Docker Compose.

For highly available, self-healing, enterprise-grade services, you probably want something like Kubernetes that can scale and manage your deployment. However, for a homelab and personal use, I really like the dead-simple yet powerful features of Docker Compose. By defining the services in a YAML file (infrastructure-as-code), your entire Docker environment can easily be brought up with one command line:

What do I want to be when I grow up?

Wed, 20 Sep 2023 00:00:00 +0000

Let me tell you a story - it’s not that unique, but hey, this is my blog and I can write what I want! All through primary and high school, I was told I excelled at math. I would get special worksheets and lessons; I took advanced classes in high school; I took part in a Math Bowl (if you have to ask, you’re not cool enough). Because of all of this reenforcement, I spent most of my teenage years pretty much set on going to college for Math. I didn’t know what I was going to do with that degree, but that was a problem for Future Travis.

Directing Traefik

Sat, 02 Sep 2023 00:00:00 +0000

So - you’ve set up some fancy new service on your home server that will let you digitize and get rid of that pile of papers that keeps stacking up on top of your file cabinet (instead of actually filing the papers - yay procrastination!). The problem: you can only access said service on 192.168.0.37:8234. Oh, and now you have 7 other services that you also need to remember the IP and port for.

How I access my home network

Sun, 09 Jul 2023 00:00:00 +0000

One of the projects I’ve wanted to tackle for a long time was to implement a way to access my home network from anywhere. There are a bunch of ways to tackle this, including just port forwarding at the router. However, I had a few requirements:

Didn’t want to have to use an IP (and look up my home IP when it changes)
Should be secure, so that not just anyone can access my network
Should be simple enough to configure and modify

For the first requirement, I could have used a Dynamic DNS service - in fact, my ASUS router comes with a built-in service. This would have allowed me to access my home network with a domain name (or asuscomm.com subdomain).

Home Sweet Homelab

Tue, 06 Jun 2023 00:00:00 +0000

A few months ago, I really got into the idea of automating my home and monitoring my energy usage. I had previously done some automation and integration of my smart devices, but it was mostly through the respective apps - there wasn’t much automation across brands or products.

Then, I jumped in the deep end. I had just decided to get solar panels installed and to trade in my car for an EV, and with that wanted to monitor my home energy usage. After some research, I came across Home Assistant, which is able to aggregate data from many different services and take action on that data.

About

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

I’ve been a software engineer for over ten years, working mainly with Java. In my spare time, I like to dabble - tinker with ideas, learn new technologies, start projects. I am very bad at finishing any of those.

I hope to use this blog to share some of those experiences. For myself, this will help me wrap up projects and get them to a point where they are presentable. For others, I hope you can find a nugget of usefulness out of these ramblings.

Privacy

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

This site uses Plausible Analytics to understand how many people visit and which posts they read.

Plausible is designed to be privacy-friendly:

It doesn’t use cookies.
It doesn’t collect or store any personal data, and visitors aren’t tracked across sites or devices.
It only reports aggregate stats — things like page views, referring sites, and approximate location by country — and no individual visitor can be identified from that data.

No consent banner is shown because none is required: nothing personal is collected, so there’s nothing to opt in or out of.