Introducing Ascent Base
June 21, 2026
Teams are adopting AI agents fast — in the editor, in chat, in CI. But each agent tends to work alone, with no idea what the rest of the team already decided. The context is scattered across pull requests, Slack threads, and design docs, so the agent guesses, repeats a mistake from last month, or quietly contradicts a decision from last week.
Ascent Base is our attempt to fix that: make your agents work as a team. Piton and Porter gather your team’s context, Cairn holds it, and every agent — in your editor and in chat — acts with the full picture. Sirdar automates the repetitive, event-driven work, and Basecamp keeps the overview.
Precision over noise
An agent acting on stale or wrong context is worse than one with none — a confident, outdated fact actively misleads it. So our first principle is precision over recall: cast a wide net for context, then cut hard, so what reaches an agent is what’s actually live.
Why an expedition
The hard parts — gaps, staleness, contested decisions — map cleanly onto mountaineering. A cairn is a marker someone leaves so the next climber can find the route. A Sherpa reads the cairns and finds the way. A Porter hauls supplies up from base. That’s the system, almost exactly:
Sherpa reads cairns, and stacks cairns.
More soon. We’re still on the approach.