WhoBuilt

QM

1 Talk5 Facts5 Sources

Key facts

Built
Y Combinator
Capabilities
uses a grind tool to set minimum time or token budgets for goals
supports human-reviewed bulk database upserts
centralizes agent conversations in Postgres and exposes them to the agent
personal employee context, sandbox files, crons, Slack access, and a web UI

Timeline 5

Dated events, newest first. Open the sources under any item.

How we verify
Built1Capability4
20265 events
  1. as of Sep 2026Capability
    Uses a grind tool to set minimum time or token budgets for goals
    “we set budgets on goals. The agent isn't allowed to give up on its task before a certain amount of wall-clock time” 57:22
    Unverified1 source· Talk
    +2 more capabilities on the full timeline
  2. as of Sep 2026Capability
    Supports human-reviewed bulk database upserts
    “we do allow writes through human-reviewed bulk upserts” 54:43
    Unverified1 source· Talk
  3. as of Sep 2026Built
    QM built Y Combinator
    “We're working on QM, YC's open-source agent harness for work.” 45:58
    Unverified1 source· Talk

Talks 1

1:00:11
Why The Harness Matters More Than The Model
Seth, John Sadavallone, Josh

The talk argues that an agent’s harness: the software layer that connects an LLM to tools, memory, compute, and the outside world, can matter as much as the model itself. It traces harnesses from simple prompting and tool use to systems that manage context, coordinate persistent subagents, and improve their own prompts or code. The speakers present Prime Agent and OpenJarvis as examples of harness design for long-running research and on-device personal AI, and describe QM as YC’s open-source work assistant. Across the examples, they emphasize expressive agent-controlled tools, long-horizon evaluation, efficient context management, and careful permissions.

Sep 7, 2026· 1 h 0 minDiscussed 13:14