This Week's Video

In this week's video I dive into how to give AI agents memory. Back in November of 2025 Google published a white paper titled "Context Engineering: Sessions and Memory". It's a fantastic read to learn the pillars of giving agents memory. One of OpenClaw's most powerful features is how it can remember and learn from its interactions with you. Which is what inspired me to make this video. In it I walk you through how you can give an agent memory and walk through OpenClaw as a real example of it in practice. Watch it here →

Live Stream 2/11/2026

Towards the end of last week, Anthropic released Opus 4.6. I'm excited to put it to work later this evening during my next live stream. I'll be continuing on building CreatorSignal, a Ruby on Rails app that I started fresh last week. I'm hoping to dive deeper into my typical AI workflow when building software, covering:

1. Git worktrees in practice
2. Trying out Claude Code's new Agent Teams feature.
3. Kicking off the development of some AI features in a Rails app.

If you're interested, be sure to tune in. If you can't make it, the video will be available on demand afterwards.

Claude Code on the Go

Sometimes I have an idea I want to act on before it slips away. Having Claude Code accessible from my phone makes that possible.

Claude Code on the web works great for smaller projects like my personal website or buttondown-mcp. I can push tasks from my phone and have it open a PR to review later. Unfortunately it hasn't worked for newer Rails apps like CreatorSignal. I'm running Ruby 4.0 and haven't found a way to customize Anthropic's container environment.

Home server + Termius is what I use for everything else. I have a VPN at home, Termius on my iPhone, and tmux to manage multiple Claude Code sessions. It's been rock solid.

And it's not just for coding. This newsletter was edited and published through Claude Code over SSH. Reviewing drafts, updating Notion, and pushing to Buttondown, all from my phone.

Thank you

I also wanted to take a moment to say thank you to all the support everyone has been giving. I'm amazed by how much positive feedback I've gotten. The YouTube channel just hit 2,000 subscribers! I look forward to making more content on what I've been learning. Also, if you have any topics you'd like to see me cover reply to this email.

Quick Hits

• If you're building applications with the Anthropic API and your tools aren't being called, check out the tool_choice parameter. This post breaks down the gotcha nicely.
• Ever used Claude Code to help you with Claude Code itself? You might've noticed it makes tool calls to fetch documentation. I recently asked it to cite its source and it pointed me to code.claude.com/docs/en/memory.md. Interesting pattern. Anthropic exposes markdown docs specifically for LLM consumption. Could be worth borrowing for your own projects.

Want to go deeper on any of this? I work with senior engineers and tech leads 1:1 on AI-augmented workflows, system architecture, and leadership. Learn more about coaching. Or just reply. I read every response.