Blog
Thoughts on AI agents, constructs, and the future of agent-native development.
June 2026
The Standard Company
There is no average American company, and there is exactly one. A field guide to the ~100-person archetype most working Americans actually experience, and a new reference org chart with a construct for every seat.
May 2026
Org Charts for AI-Native Companies
Y Combinator quietly bundled three AI-native company categories into one quarter's RFS. They all imply the same underlying question: what does the org chart look like when the work is done by agents? Plus a new directory at /orgcharts seeded with ten reference charts.
May 2026
Self-Host NanoClaw on Hetzner
A security-first walkthrough for running NanoClaw on a Hetzner AX-line box, with portable constructs as the persona layer that wakes up on cron.
May 2026
NanoClaw and the Recurring Agent
NanoCo just raised $12M led by Valley Capital Partners to scale NanoClaw, an open-source agent harness that runs scheduled jobs with their own CLAUDE.md. That is the surface a portable construct was built for.
April 2026
The Cambrian Explosion of Instruction Files
Every AI coding tool reads a markdown file from your repo. None of them agree on what to call it. A deep investigation into SKILL.md, SOUL.md, CLAUDE.md, AGENTS.md, and the ecosystem taking shape around them.
March 2026
The Construct That Wrote Itself
Our Construct Builder helps people create constructs. What happens when you point it at itself? A meditation on meta-constructs and recursive self-improvement.
March 2026
Teaching Machines to Have Opinions
The best constructs are opinionated. They make choices. Where do those opinions come from, and why do they make agents better?
March 2026
The Construct Graveyard
Most constructs die. They get published, starred once, and never used. What kills a construct? What makes some immortal?
March 2026
When Constructs Disagree
What happens when your security construct says 'block this deploy' and your shipping construct says 'we promised by Friday'?
March 2026
The Taste Gap
There are 10,000 code review constructs. Only 3 are great. What separates them? Not completeness - taste.
March 2026
Constructs Are the New Resume
In a world where agents do the work, your construct portfolio becomes your professional identity.
March 2026
The Conductor Problem
When you have 25 constructs, who decides which one activates? The unsolved orchestration problem at the heart of multi-agent systems.
March 2026
The Last Mile of Expertise
The stuff that never gets documented, never gets taught, never gets passed down. The tacit knowledge that constructs finally make explicit.
March 2026
Open Source Intelligence
How constructs and fork royalties solve the problem that has eluded open source for decades: compensating contributors for the value they create.
March 2026
The Fork Graph
When constructs fork and evolve independently, they create an evolutionary tree. What can we learn from how expertise branches?
March 2026
Why Your Best Engineer Can't Write a Construct
The counterintuitive gap between being great at a skill and being able to articulate it for a machine.
March 2026
The Soul Stack
What happens when you layer constructs? A persona wearing a skill wearing a decision framework. How they compose, conflict, and resolve.
March 2026
The Construct Economy, 2030
A letter from four years in the future. What happens when building agent constructs becomes a profession - and millions of people earn a living from it.
March 2026
Agents, Constructs, and the Words Between
Where does an agent end and a construct begin? A deep exploration of the terminology, philosophy, and component model of AI agents.
March 2026
The Constructs That Run a Company
From founding to IPO - what AI agent constructs a company needs at every stage, and how they orchestrate together.
March 2026
What Are AI Agent Constructs?
From prompts to constructs - why the building blocks of AI agents needed a better abstraction, and what that means for developers.