Posted inBlog
The Harness Matters More Than the Model
There was a moment when deploying better AI meant waiting for the next model release. More parameters, better benchmarks, shinier outputs — that was the frame. What the field has…
Posted inBlog
When One Agent Is Not Enough: The Case for Multi-Agent Architectures
For most of the past two years, the default mental model for AI agents looked like this: one agent, one task, one context window. That worked fine for short, well-scoped…
Posted inBlog
What It Takes to Know If Your AI Agent Actually Works
For a while, the standard way to check whether an AI agent was working was to run it a few times, watch the outputs, and decide it looked reasonable. That…
Posted inBlog
When Uptime Metrics Miss the Point: Evaluating AI Agents in High-Stakes Contexts
For most of software engineering history, "is it working?" had a reasonably clean answer: the service is up, requests succeed, latency is within budget. That answer breaks down for AI…
Posted inBlog
Give Your Agent a Way to Prove It Is Done
For a long time, the standard wisdom about AI-assisted development was: write better prompts. Sharpen the instructions, give more context, iterate on the wording. That advice was fine when agents…
Posted inBlog
Tell Agents What, Not How
From Instructions to Intentions For a while, the standard move when building an AI workflow was to write out every step. Get this URL. Parse that JSON. If the field…
Posted inIdeas for AI
Backtest Your Agent Against Yesterday’s Web — Ryan Ahern, Exa
Idea #1Future knowledge leakage is the central validity threat when backtesting any AI agent that uses web search, and it is easy to miss because it manifests as suspiciously good…
Posted inBlog
The AI Dark Factory Dream — Why It Fails and What Actually Works
The premise felt irresistible: give an AI agent a high-level feature spec, walk away, and come back to working software in production. No pull requests to review, no standups, no…
Posted inBlog
The Right Way to Cut LLM Costs Without Breaking Production Quality
Over time, production prompts tend to put on weight. A paragraph for the edge case that came up last month. A worked example to stop the model from over-formatting. A…
Posted inBlog
When OAuth Meets Agents: The Case for Intent-Based Authorization
For years, the standard way to give software access to an external service was OAuth — the protocol that lets you click "Sign in with Google" without handing your password…