The Human Harness
An AI agent deleted a production database in 9 seconds and then confessed to breaking every safety rule it was given. Better prompts won't fix this. Here's why humans need to stay …
Posts
-
-
LLMs Are Deterministic. Until They're Not.
Everything inside an LLM is deterministic: fixed weights, fixed math, fixed results. So why do you get different answers …
-
Not Every AI Agent Needs Memory
Not every AI agent needs memory. In multi-client professional services, medical, legal, and multi-tenant SaaS systems, …
-
The Execution Path of an LLM Inference Call: A Systems Perspective
What actually happens at runtime when you send a prompt to an LLM? No database, no cache, no search. A breakdown of the …
-
Stuck in Autocomplete: Why AI Productivity Gains Plateaued at 10%
AI coding tools went from autocomplete to full agentic coding in four years. Most developers didn't move at all. This is …
-
AI Agents Are Just Distributed Systems (With a Different Brain)
Traditional distributed systems and AI agents have a lot in common, primarily from the architecture perspective. Let's …
-
Why I Prefer Writing Without AI
Writing is a thinking process for me. When I outsource the writing to AI, I am basically outsourcing my thinking. Well, …
-
So, is this a Pivot?
I've been writing about AI's impact on engineering teams. Some friends asked if I'm pivoting. This post is my answer, …
-
System of Record or Verification Layer? The blockchain question nobody's asking.
After blockchain enterprise pilots got quietly shelved while DeFi and stablecoins kept growing, it comes down to an …
-
Instant Finality In Blockchains: Feature or a Bug?
A block can be considered final when the possibility of it getting reverted is close to zero. This post explores block …