Superdense · Blog
Loop engineering, in the open
Notes on outcome-driven agents — pointing AI agents at a real-world metric and getting them to improve on it, run over run.
- Jul 15, 2026·6 min read
Agent reward loops: the missing layer between memory and real-world improvement
AI agents don't improve just because they remember. They improve when each shipped action is tied to measured reward and used in the next run.
- Jun 30, 2026·7 min read
Agent memory vs. agent feedback loops: why remembering isn't the same as improving
Most "AI agent memory" just gives an agent recall of the past. It doesn't make the agent better. Here's the difference between memory and a feedback loop.
- Jun 30, 2026·7 min read
Evals vs. outcomes: why your AI agent should loop on a real-world goal
Evals tell you if an agent passed a test. Outcomes tell you if it moved the number you actually care about. Here's the difference, and why it changes how you build.