WorkBuddy

Headlines · 3

1. 2h
   Flash-model test rerun: deliverable quality tracked the agent workspace, not the model

   A 36Kr piece by Qidian Research Society revisits a Flash-model comparison in which DeepSeek-V4.1-Flash and Yunzhisheng U2-Flash ran a phone-industry research report and a match-3 game inside the AI office workspace WorkBuddy, while Qwen-3.8-Flash answered from a solitary chat box on Qwen's AI platform and produced wrong 2026 phone line-ups, misjudged data and a game with frozen scoring logic. Re-running Qwen inside the same WorkBuddy environment, the article says, yielded an eight-chapter report with self-correction and a working game via a locate-bug, edit-code, recompile-and-verify loop. The author argues AI office competition is shifting from the 'model entrance' to the 'task entrance' and onward to enterprise AI infrastructure and multi-model governance.

   36氪 人工智能 · 🔥 10
2. 1d
   Why OpenRouter’s Rankings May Not Measure Global Model Preference

   An analysis argues that OpenRouter’s rankings measure token volume routed through one intermediary rather than global developer preference, and that prices, discounts, free access, batch jobs and self-testing can heavily influence the results. It also questions the comparability of Artificial Analysis scores and ARC-AGI results when test composition, weighting, scoring anchors, interfaces and versions change.

   钛媒体 · 🔥 5
3. 2d
   AI keeps getting stronger — so why aren't companies shipping faster?

   This is a commentary piece on organizational collaboration and AI adoption: starting from the gap between solo developers and team delivery, the author argues that AI surfaces demo-ready output earlier, while the real finishing work — permissions, data, exception handling and acceptance criteria — has not disappeared and is easily undercounted. It cites "Shifting Work Patterns with Generative AI" (66 firms, 7,137 knowledge workers, a six-month randomized field experiment), which found users of the tool cut roughly two hours of email time per week but detected no change in task volume or composition, plus a September 2026 WorkWorlds preprint (192 paired evaluations, 76.7% vs 68.0% pass rates) and a 2025 Procter & Gamble field experiment with 776 professionals. The author urges firms to track end-to-end task time from request to acceptance and to separate missing expertise from missing decision authority.

   虎嗅 AI · 🔥 9

Experience and discussion from the community

Share my WorkBuddy experienceAsk about WorkBuddy