I was on the bus when Claude asked whether to commit the refactor. I tapped yes on a Telegram keyboard. The code was merged before my stop.
I'm Nav. I design infrastructure at national scale during the day and build the tools I wish existed at night. Everything I make, I use.
Building a multi-agent robotics setup where agents coordinate to gather and analyse real-time data. Working on an air-gapped home AI that controls the house without touching the cloud. Running and evaluating local models on a cluster of Mac Studios. Shipping OpenDray, a terminal cockpit for AI coding agents.
Containers, orchestration clusters, multiple cloud regions, a mesh network tying it all together. I run a scaled hybrid lab because I refuse to recommend architecture I haven't broken myself. If I'm going to stand in a steering committee and say something works, I want to have broken it first.
Read this post →You need a kill switch that doesn't depend on you being awake.
go:embed bakes the Flutter web build into the Go binary. One file to copy, one process to run.
I added a Telegram bridge to OpenDray so I could check on Claude sessions from my phone. It was supposed to take a Saturday.
ShopSerp watches 250+ stores across 17 countries. It started because I was shopping for a camera lens.
I kept doing the same infrastructure tasks every week. So I built agents to do them instead. There are 14 of them now.
I was designing liquid-cooled GPU processing farms 14 years ago. The conversation has changed. The physics hasn't.
User provisioning, license assignment, mailbox permissions, security policies. Every tenant slightly different. Every change a potential landmine.
The hard part wasn't the automation. It was testing it. You can't simulate a real data center failure in a staging environment.
Every streaming platform stores metadata differently. Song titles don't match. Artist names have variations. ISRCs are unreliable.
Key management, transaction signing, cold storage. No established patterns existed. No frameworks.
Custom rigs, custom cooling loops, power distribution, monitoring. The rigs are long gone but the knowledge stuck.
Built for a friend whose warehouse ran on spreadsheets and group chats. Now it tracks shipments, scans barcodes, manages returns, and handles serial numbers across vendors. The agents handle receiving and dispatch workflows so the humans can focus on exceptions.
Last week, DynaCore's identity agent offboarded a user across an entire M365 tenant in under two minutes. Disabled the account, set up forwarding, converted the mailbox, released the license. I found out from the summary it sent to Telegram.
A framework for giving any business its own AI employees. You define the domain rules and connect the tools. The agents learn the repetitive patterns. Humans stay in the loop for judgment calls. Started because I kept rebuilding the same agent scaffolding for different industries.
I was shopping for a camera lens and checking five different websites. That annoyed me enough to build an agentic crawler that watches 250+ stores across 17 countries, tracks price history, and tells me when the best deal appears.
I wanted to start a Claude session on my server and check on it from my phone. That turned into a full terminal cockpit with a Telegram bridge, plugin system, and support for six different AI agents. The bridge reads Claude's structured output and renders permission dialogs as inline Telegram keyboards.