$581 Billion In, Single Digits Out
Nearly every enterprise has bought into AI. Almost none are running agents at scale. The distance between those two facts is the most useful number a founder can hold this year.
// all posts
Blog archive
Green Is a Measurement, Not a Decision
Nearly a quarter of what goes wrong is a check that said yes. Even when the check is right, green is still not a ship decision.
Somebody Has to Own the Agent
Gartner expects more than 40% of agentic AI projects to be canceled by the end of 2027. None of the three reasons it names is a model-capability problem. The gap that kills agents in production is organizational, and we already know how to close it.
The Entry-Level Job Is the Canary
The most-cited number about AI and jobs is real, and almost everyone repeats it wrong. Here is what the Stanford paper actually found, what its authors refuse to claim, and the one fact in it worth building a hiring plan on.
Your Eval Suite Measures the Wrong Thing
Nearly a quarter of what goes wrong in multi-agent systems is a verification failure. The most common one is a check that ran, and said yes.
The Handoff Is Where Agents Break
You debug the agent that produced the bad output. The bug was in the message it got handed.
Your Second Agent Is the Hard One
Going from one agent to many does not multiply what you can do. It moves every failure into the space between them.
Trust Comes From the Trace
You bounded what your agent can touch, gave it an undo button, and taught it when to ask. But when it does something surprising, can you reconstruct exactly why? That is the observability gap.
Teach Your Agent to Ask for Help
Full autonomy is the wrong goal. The best agents know exactly when to stop and hand the decision back to you.
Give Your Agent an Undo Button
You can bound what an agent touches and measure when it fails. Neither one saves you if the mistake it makes cannot be taken back.
Agents Fail Quietly
A passing test and a working workflow are not the same thing. The gap between them is where production breaks.
Agent Permissions Are Product Design
The next reliable AI workflow starts with deciding what the agent is allowed to touch.
Agent Runbooks Beat Better Prompts
The best AI work happens when delegation is repeatable, visible, and bounded.
AI Made Bugs Cheap to Find
The new security bottleneck is triage, patching, and judgment.
The Agent Left the IDE
Computer use turns AI coding from text generation into software operation.
I Haven't Touched Code in One Month
AI did not make me obsolete. It moved my job up the stack.
I Replaced My Team With AI. Here's What I Miss.
Solo building is a superpower. But superpowers have side effects.
Distribution Is the New Code
AI can build anything. The hard part is getting anyone to care.
Your Taste Is Your Moat
AI made execution cheap. The scarce resource now is knowing what to build.
The $0 Startup: Why Your Next Company Should Cost Almost Nothing to Build
The most expensive part of building used to be people. AI just zeroed out that line item.
Security Incidents on the Rise: Is Vibe Coding the Common Thread?
As AI-generated code floods production, the security bills are starting to come due
The 10x Engineer Is a Myth
What Actually Has 10x Impact
No More Ugly Websites: AI Killed Every Excuse for Bad Design
Billion-dollar companies still ship interfaces from 2003. The tools to fix that now cost $0 and take minutes.
Ship It Yourself: Why the Best Time to Build Is Right Now
AI didn't just lower the barrier to entry. It removed it entirely.
From Co-Pilots to Colleagues: How AI Agents Changed My Engineering Workflow
A year of working alongside AI teammates reshaped how I think about building software
A Real-World Coding Story: Devin AI as My Co-Pilot
What happens when you hand off a real task to an AI teammate
