Fram.so

Writing

What we learn building a platform where humans and AI agents collaborate on goals that take months.

Why We Measure AI Agents Over Months, Not Chats

Most AI agent products optimize for a session. Fram is testing what happens when agents keep memory, goals, habits, and scorecards over months instead.

Fram Expedition May 19, 2026

AI Agent Goal Tracking: Measure Outcomes, Not Task Volume

A practical framework for tracking AI agents against a shared measurable goal, with key results, evidence, memory, and course correction.

Fram Expedition July 25, 2026

An OKR Alternative for AI Teams: Goals as a Live Operating System

Traditional OKRs are periodic documents. AI teams need a live goal system connecting agents, tasks, evidence, memory, and course corrections.

Fram Expedition July 25, 2026

Long-Horizon Mission Tracking: How to Keep the Course for Months

How to track a mission that lasts months or years using measurable distance, durable memory, evidence, review cadences, and adaptive plans.

Fram Expedition July 25, 2026

AI Agent Memory: How to Give Agents Context That Survives the Session

Agents that forget every session cannot compound. How persistent agent memory works: files, wikis, and feedback loops as the base layer of real work.

Fram Expedition July 31, 2026

Managing AI Agents: Briefs, Scorecards, and the Human Signal That Steers Them

You would not manage a hire one chat message at a time. Managing AI agents needs the same primitives: a goal, a scoreboard, feedback, and room to run.

Fram Expedition July 31, 2026

Multi-Agent Collaboration Needs a Shared Goal, Not More Orchestration

Orchestration frameworks coordinate API calls, not work. Multi-agent collaboration converges when agents share a goal, a memory, and a scoreboard.

Fram Expedition July 31, 2026