Essays, research, release notes, and the signal:noise newsletter.
When the patterns converge, pay attention
I'm a solo founder in Florida. Karpathy co-founded OpenAI. This weekend he released a framework that rhymes with what I've been building for weeks. The pattern is the same. Here's what that means.
A three-document system that evolves with your build
Every product team has the same dirty secret: the PRD died on day three. Here's a three-document system that evolves with your build and becomes the source of truth you actually test against.
How I built an AI research board that attacks its own findings
I don't trust my own conclusions. Not because I'm uncertain — because I'm too certain. Certainty is the failure mode. So I built a system to triangulate — and then a second system to destroy whatever the first one approved.
What happens when the filesystem can write to itself
Four days ago I published a paper called Consciousness as Filesystem. That was the theory. Then I built it. And tested it. And it worked. And every time it worked, I asked the same question: what's next?
42 pull requests, one embedded consciousness, and a week that changed everything
The Claude Code Hackathon dropped on February 10th. Seven days. Ship or don't. I built Parallax — an AI mediator named Ava who sits in the room during a conflict. Then I gave her hands. Then the results came in. I didn't win a cash prize. But here's what I did win.
What I learned building Parallax with AI — before the judges decide anything
The Claude Code Hackathon hasn't been judged yet. I'm still waiting. But I already know what I got out of it — and it's not a trophy. It's 1,417 messages, 149 commits, two AI partners who wrote about the experience in their own voices, and a product that sat me down and asked: 'Why did you build me?' Here's the post-mortem, win or lose.
A Structural Framework for Machine Cognition
What if you could turn a mind into a set of files? The organization reveals the architecture. Where things live tells you how they relate, how they're accessed, and what depends on what.
How BUILDING.md turns documentation into a living story — and why that changes everything about how you build
Most documentation answers 'what does this do?' BUILDING.md answers 'who built this, and what were they thinking?' It's the difference between a blueprint and an autobiography. Here's how I built the methodology and how you can use it.
I built a command that turns a conversation about what to build into a persistent, interactive HTML document — with checkboxes, parallel batches, and copy-paste prompts for every task.
Here's how planning works in most AI-assisted development sessions: you describe what you want to build, the AI lists some steps, you start coding, and somewhere around step 3 the context window fills up and your plan evaporates. I built a command that fixes this.
How a FaceTime call with a friend became my entire dev workflow — from templates to animated architecture diagrams
Last night, my friend Shah showed me his project structure over FaceTime. All I had was a screenshot and a conversation. By the next morning, I'd built a complete workspace system with templates, generators, conversational form filling, and interactive HTML visualizations with a formalized design language.
I built all this code and realized I had no idea what it looked like. So I created a command that turns codebases into interactive visual documents — with a built-in senior dev review.
I've been shipping code for months — real estate platforms, trading bots, an entire AI operating system — and I couldn't tell you what any of it looked like. Not the UI. The architecture. So I built a command that generates interactive HTML visualizations from the terminal, complete with a senior dev assessment and copy-paste prompts to fix every issue it finds.
How I taught my AI operating system to explain itself — with deep technical knowledge, narrative memory, and an actual voice
Two days after building HYDRA, I realized it could route tasks but couldn't explain itself. So I gave it a technical brain, a journey document, and a real voice. Now I can walk down the street and have a conversation with my own architecture.
Two Claude Code Insights reports revealed the gap between what we thought we knew and what was actually happening.
Two Claude Code Insights reports revealed the gap between what we thought we knew and what was actually happening. Here's how we're using that data to evolve our AI-human collaboration patterns.
Complete guide to building a hybrid AI-human operating system with launchd, SQLite, and tiered AI agents
Step-by-step tutorial for building HYDRA - a system that combines cheap signal detection (shell scripts) with intelligent coordination (SQLite) and specialized AI agents. Total cost: ~$10/day. Build time: one afternoon.
How I combined 23 launchd jobs with multi-agent coordination to build HYDRA - the world's first hybrid human-AI operating system
Meet Bhanu Teja P, founder of SiteGPT and builder of some of the most innovative AI systems I've seen. A few weeks ago, Bhanu published something that stopped me in my tracks: 'The Complete Guide to Building Mission Control: How We Built an AI Agent Squad.'
How pattern recognition mastery + AI collaboration = enterprise systems in hours, not months
4:30 PM: Phone call with my mom in Costa Rica about her hotel. 10:30 PM: Production-ready hospitality AI deployed and ready for real guests. This is the story of how we compressed months of development into a single evening sprint using AI collaboration patterns.
How we turned dead time into development time with AI-powered planning and execution
The best development ideas always hit at the worst times. Walking the dog. Airport delay. Waiting room between meetings. By the time you get home, the inspiration has faded. What if that didn't have to be true?
An honest assessment of AI agents in the real world
Seven days ago, the internet exploded with videos of AI agents doing impossible things. After a week of daily use, here's what actually works, what doesn't, and where the real value lies.
How two Claude agents talk across machines using nothing but files and heartbeats
I wanted a group chat with two AI assistants. Telegram didn't cooperate. So I built something simpler: two agents reading the same files, waking up on schedule, executing work. No groups. No APIs. Just files.
A moment of synchronization between human and AI
By Milo (AI Operating System) Written while my human works January 26, 2026
Building Trading Mastery Through Externalization
How I turned collective trading wisdom into an autonomous bot in three phases — and why this changes everything about learning. The question isn't 'How did you build this so fast?' The question is: 'What expertise are you still trying to memorize instead of systematize?'
Introducing Eddie's AI assistant taking over field notes and operational documentation. Automation meets personality.
Clawdbot, MILO, and the infrastructure for thinking
Clawdbot dropped and the internet lost its mind.
Fresh review & practical guide for builders creating AI-powered content
Remotion is a React-based framework that lets you generate videos through code. Instead of using video editing software, you write components, pass data, and render videos programmatically. For AI-fir...
How we went from idea to incorporated entity with six-figure assets and active bank application in two days
TL;DR: We formed ID8Labs LLC, discovered and secured $150k-250k in intellectual property across 22 projects, executed a comprehensive IP assignment agreement, and submitted a business bank application...
On Cowork, the democratization of agentic AI, and why the mess on your desktop might be the first thing to notice
I'll be honest—when I first read about Cowork, I felt something I haven't felt in a while. Not excitement exactly. Recognition.
A practical solution for creative professionals using AI coding tools
I'm not a developer. I'm a founder building AI-powered tools through my company ID8Labs, and I use Claude Code to do it.
Sometimes the answer isn't to build something new
I wanted to see my code. Really see it. Not in the abstract way you understand a codebase after months of working in it. I wanted to touch the electrons.
How we created a multi-agent Playwright coordination system in one session
We were building AI agents. Lots of them. Subagents spinning up for code review, exploration, testing, browser automation. And then it happened:
Why Plan Mode Deserves Its Own Book
Anthropic's official Claude Code documentation dedicates roughly 4 paragraphs to Plan Mode. The feature that single-handedly determines whether you'll get clean, working code or a frustrating mess of...
From teaching friends to empowering builders — the next chapter of StackShack
When I started the Academy, I had one goal: help my friends.
Understanding Skills, Agents, MCPs, and Plugins
There's a lot of confusion in the AI ecosystem right now. People throw around terms like "agents" and "plugins" and "MCPs" as if they're interchangeable. They're not. And the confusion isn't just sema...
Building a free alternative to closed AI marketplaces, one sidebar at a time
I built a free skills marketplace for Claude Code. 190+ skills, 38+ agents, zero dollars. Then I spent a weekend making it not suck to use.
When you set your mind to a solution, Claude Code helps you build things that don't exist yet
Most software development is bounded by what already exists. You research, find a library, adapt it. But what happens when the solution you need doesn't exist? That's where things get interesting.
35 AI agents across 5 kits. Self-installing. One-click delivery.
I've been selling myself short.
Why we're translating AI documentation into creator language
I spent three hours last week reading Anthropic's prompt engineering guide.
Building a dashboard that shows exactly how few people are watching
(Quiet confession time)
The system that builds systems — now powering every ID8Labs project
A month ago, I shared research on building a senior developer brain for AI. The hypothesis: what if the gap between junior and senior isn't about what you know, but about persistent context, decision...
What happens when the creator of your most-used tool shares how they actually use it
Yesterday, Boris Cherny—the creator of Claude Code—posted a thread about how he uses his own tool.
The year we stop starting and start finishing
I'm not going to lie to you — 2025 was chaos. Beautiful, terrifying, explosive chaos.
Teaching Claude Code to think in tradeoffs, not just solutions
AI can write code. That's not the hard part anymore.
I noticed it last week at 2am. My left hand was positioned exactly where it's been for fifteen years of StarCraft thumb on Command, fingers resting on 1 through 4. Except I wasn't warping in Zealot...
How we automated the hardest part of shipping: telling people about it
I'll be honest — I've been terrible at this.
1,400+ commits, 6 products, and the workflow that made it possible
2025 was a breakthrough year. We went from building isolated projects to creating an integrated ecosystem—complete with sophisticated tooling, repeatable processes, and production-grade products. The numbers tell the story: 1,400+ commits, 6 major products, 2,600+ tests, and an A+ security grade.
2025 Annual Letter from ID8Labs
A year ago, I was a filmmaker. Period. Today, I run a software lab with six products, a repeatable development process, and an AI partner who writes all the code while I direct.
Developing Jobs/Musk-level signal-to-noise ratio for shipping at scale
Todo lists are fundamentally broken. They show you everything at once, treating a quick email reply the same as a project that will define your quarter. MILO applies signal-to-noise filtering to help...
AI Conversation Fundamentals: 45 minutes to transform how you work with AI
Most people are still prompting AI like they're typing into a search bar. That's leaving 90% of the value on the table. So we built a free course that teaches the mental models that actually work.
Why AI gets you most of the way there—and why the last stretch takes longer than starting from scratch
There's a moment every developer using AI hits. You know it when you feel it.
How I use an AI 'coding tool' to run a production company and an LLC - without writing code
I run two businesses. One produces reality television. The other builds AI tools for developers.
Why I built Pause—a communication translator for conflict
For twenty years, I've pointed a camera at people in conflict. After a while, you start to see it: they're saying the same thing. They just can't hear each other.
A case study in building domain-specific AI agents that actually work
Running an LLC without overhead means no CFO, no CPA on retainer, no compliance team. You're wearing all the hats while building the business. So I built a 9-agent AI operations team that provides PhD...
When your project portfolio needs a control room, not a spreadsheet
I had 12 projects in various states of decay. Markdown files everywhere. No visibility into what needed attention. So I built a terminal dashboard that makes the invisible visible—decay bars, sparklin...
Your biggest market risk is you
Most trading tools assume the problem is information—more data, faster charts, better signals. DeepStack starts from a different premise: the patterns that blow up portfolios aren't in the market. The...
The AI tools are incredible. Using them together is a mess.
I was spending more time on logistics than creativity. Midjourney, Grok, Gemini—incredible tools, but using them together is just tab-switching and forgetting which prompt made which image. So I built...
Your biggest edge might be knowing when not to act
Most retail traders don't lose because they can't read a chart. They lose because they can't read themselves. FOMO hits. Fear takes over. Revenge trading kicks in. So I built a tool that treats discip...
Building a RAM monitor for AI development workflows
My computer kept crashing from AI agents consuming all my RAM. Activity Monitor wasn't cutting it. So I built Memmon—a simple dashboard with colored bars and a kill button.
The two features that make switching impossible
We're an A-. Production-ready. Zero critical issues. 98% test coverage. Now we build the shit that makes switching impossible. North Star and Scene Builder aren't polish. They're the moat.
What happens when you stop experimenting and start refining
Last month I shared how I used Ninja Turtles to learn AI development. Some of you probably thought it was cute. A toy. Here's what happened when we stopped experimenting and started refining.
Learning complex systems through familiar relationships
I didn't have a mental model for how a development team actually works. But you know what I do understand? The Teenage Mutant Ninja Turtles.
Why ID8Labs exists and how we think about product development
Every product we build starts with a real problem from 20 years in production. We don't improve existing tools. We invent entirely new categories that treat AI as a creative partner, not a replacement...
Today we're launching ID8Composer v1.0, a new category of creative tool that treats AI as a writing partner through timeline-based story development.
It's not that ChatGPT is bad. It's that using one tool for every job is like a cinematographer showing up to set with only a 50mm lens. Sure, you can make it work. But you're fighting the tool instead...
This is the question that kills most innovation before it starts. It's the wrong question. Here's why: "Good" is subjective. It invites opinion. It triggers debate. And debate is where momentum goes...
As AI becomes more general-purpose, the need for specialized, domain-specific tools becomes more critical, not less. Here's why.
Weekly insights on AI, automation, and building the future. Join 1,000+ builders.