[Figure CEO Launches Browser AI: Ranking First is True, 'Strongest in History' is Overhyped] According to monitoring by Beating, the humanoid robotics company Figure's CEO Brett Adcock has launched an AI company called Hark, which recently released its first AI Agent product, Handoff. This tool can independently operate web pages, helping users order takeout, shop, book flights, and fill out forms. Handoff scored 97.7 in the browser operation evaluation Online-Mind2Web, surpassing GPT-5.4 and Claude Opus 4.8 to rank first. This evaluation involves AI completing tasks on real websites, with results manually verified. However, Handoff only outperformed the second-place Yutori by 0.4 points. Adcock quickly labeled it the 'strongest internet model in history.' That said, the ranking did not include GPT-5.6 or Claude Opus 5. Additionally, in Hark's own WebTailBench v2 test, Handoff scored 68.6, falling short of GPT-5.5's 72.3. The product is also not yet publicly available, so external parties cannot retest it. Ranking first is true, but the claim of being the 'strongest in history' is clearly exaggerated. [Original Link]