Descriptions:
Peter Steinberger, creator of Open Claw, delivers a candid 40-minute YC talk tracing his project from a personal frustration — no easy way to relay phone prompts to a running agent in late 2024 — to a WhatsApp relay, then an open-source Claude interface that peaked at 4.7 million weekly downloads. The talk covers five questions in roughly chronological order: how Open Claw came to exist, what scale actually felt like, the competitive and operational crises that nearly killed it, and whether building it is still fun.
Steinberger is unusually frank about the hard parts. The project bottomed out at 835,000 weekly downloads in May 2026 before rebounding to its all-time high in June — after being declared dead. A key inflection point came when Anthropic gave approximately 24 hours notice of subscription changes that broke Open Claw’s Claude-optimized harness, leading to the observation Steinberger says is worth writing down: “your dependency’s business model is your business model.” He also describes how a well-funded competitor with an aggressive migration campaign and simpler messaging ate into Open Claw’s momentum while he was focused on security patches and infrastructure.
The talk is notable for its psychological honesty — Steinberger describes losing the ability to enjoy his own product around February 2026 as maintenance obligations grew — and for the practical lessons on harness engineering, compaction, and open-weight model readiness. It is a rare unfiltered account of what happens when a solo-maintained AI infrastructure project escapes into the wild at scale.
📺 Source: Y Combinator · Published August 10, 2026
🏷️ Format: Keynote Launch
