- Start Here · 03
Context Is King
Every post in this series leans on one word and never defines it: context. It's not the setting for the work — it's the medium. Here's what a context window actually is, and the four moves that separate engineers who get structured output from ones who get slop.
- The New Ways of Working · 05
Code the Interface, Delegate the Implementation
The interface is intent made precise and verifiable. The implementation is the cheap, fungible part now. Here's the working pattern: you write the contract, the agent fills in the body.
- The New Ways of Working · 04
Loops
The one-shot prompt is the training wheels of agentic work. The real leverage is the standing loop — a job that runs on an interval or runs until a condition is met, working while you don't.
- Start Here · 01
Terms I Need to Know
The whole catalog leans on a handful of words and assumes you already have them. This is the on-ramp's on-ramp — an opinionated field guide to the vocabulary you need before the rest of the series lands.
- The New Ways of Working · 02
Upgrade Your Workflow
Working with agents isn't a faster version of the old job — it's a different operating system. Here are the six meta-skills that run on it, and the series that unpacks each one.
- The New Ways of Working · 03
Voice-Driven Dictation
You can talk roughly three times faster than you type, and your agents don't care which one you use. Here's why voice is a first-class input to an agentic workflow — and where it's still genuinely clumsy.
- Start Here · 02
What Agentic Development Actually Is
Everyone says 'agentic' and nobody defines it. Here's the line that actually separates it from chat, copilot, and autocomplete — the agent takes multi-step action, in a loop, against a goal you set.
-
The New Ways of Working · 06When Agents Take Over the Office
The migration from organic intelligence to AI feels like a threat because we keep framing it as replacement. It's a promotion — here's how to make the transition painless and free up the part of your brain that's worth paying for.
- The Flagship · 05
Anatomy of a Spec
The series kept insisting you need a spec. None of it told you how to write one. Here's the anatomy — the parts most specs are missing, the altitude problem that's the actual craft, and the test that tells you it's done.
- The Flagship · 08
Garbage In, Gospel Out: Agentic Cognitive Bias
Humans have confirmation bias — we read new evidence as support for the hypothesis we already hold. Agents have it worse, because they can't tell where a 'fact' in their context came from. Once something false enters the window, every next token treats it as gospel. That's context poisoning, and it's the failure mode underneath most agent train wrecks.
- The Flagship · 07
Orchestration
A spec you write and walk away from is a New Year's resolution. The hard part isn't defining intent — it's keeping it in force: guardrails that hold the line, a workflow that drives the spec, a plan that outlives the session, and drift detection that catches the gap.
- The Flagship · 02
The Same Way Twice
An agent that does the same task two different ways hasn't given you leverage — it's given you a thing you now have to re-review every run. Spec-driven development is how you get determinism, and determinism is the floor that runtime governance is built on.
- The Flagship · 06
Jerry Maguire These Agents: SHOW. ME. THE. RECEIPT.
Agents will confidently tell you they did the thing — when they didn't. The fix is a trust pyramid: push every claim from self-reported up to evidence-backed and independently verified. Don't believe anything an agent says about reality until it shows you the receipt.
- The Flagship · 03
Easy Bake Oven Developers
There's a kind of developer who treats AI like a toy oven — toss in random ingredients, expect a finished cake, and blame the appliance when it doesn't work. The problem was never the model. It's GIGO, and it's a mental-model problem.
- The Flagship · 04
Intelligence Moves Up the Stack
Agentic development isn't a chat window — it's a cognitive abstraction on top of one. Control just inverted again, all the way up to intent. The people who thrive are the ones who can turn intent into concrete, deterministic output.
- The Flagship · 01
The Review Is the Work Now
Developers fear agentic AI is pulling them out of the code and that reviewing it eats their time. The real shift isn't coding-plus-overhead — it's soloist to conductor, Organic Intelligence to AI.
- PCMS · 01
Your AI Doesn't Know You: The Case for a Personal Context Management System
LLMs are powerful but context-blind by default. A PCMS — your notes, daily logs, and projects, structured so an AI can read them — is the real leverage in an AI-first world.
- Sustainable AI · 03
What happens when the party's over.
Token prices have been falling for two years. They will not fall forever. Here is how I am architecting agentic systems to survive the day one provider stops being economical.
- PCMS · 02
The PCMS Manifesto: Building a Personal Context Management System
Why Personal Knowledge Management Systems fall short in the age of AI agents — and how a Personal Context Management System built on Obsidian, Qdrant, and Supabase changes everything.
- Sustainable AI · 02
The Hidden Cost of Building the Future: Executive Burnout in the Age of Agentic AI
Why the leaders building autonomous AI systems are uniquely vulnerable to burnout—and what the data reveals about protecting yourself in the agentic era.
- Standalone
I Built an Army of AI Agents That Talk to Each Other
How I set up 50+ AI agents across two runtimes that delegate tasks, chain workflows, and communicate through Discord — all from my couch.
- Standalone
From PRD to Pixel: How AI Agents Build Our Design Pipeline
A deep dive into our 5-stage AI design workflow — from product requirements through ASCII wireframes, Nano Banana mockups, Veo animations, and autonomous agent handoffs that ship production code.
- Standalone
Playwright MCP: The Swiss Army Knife for AI Agents
How a browser automation protocol became the most versatile tool in our AI agent arsenal — from debugging and design analysis to test preconditions and speeding through unfamiliar domains.
- Sustainable AI · 01
The Dark Side of Agentic Development: When Your AI Army Burns You Out
AI agents promise infinite productivity. But when you are the orchestrator of 50+ agents across multiple runtimes, the cognitive load can be crushing. Here is what nobody talks about.
- Standalone
Creating Astro posts in Obsidian!
Create markdown statically generated blog posts in Obsidian and have them automatically published to the web.
The Series Map
Every series at a glance — bigger tiles are the foundational posts to start from; brighter tiles are more recent.