Dowson Tong, Tencent's senior executive vice president and CEO of its Cloud and Smart Industries Group, published a lengthy response to the criticism that Tencent has been slow in AI. He admits the company has indeed been anxious internally, while rivals have made much more noise in the past few years. More bluntly, Tencent's overall compute capacity is "seriously insufficient," which has slowed Hunyuan training and product development and had "a fairly big impact."
But Tong believes this may be only the first kilometer of an AI marathon: "outlasting others matters more than starting early." He teases that the next-gen Hunyuan Hy4 will bring more breakthroughs, while insisting Tencent can't just compete on foundation models. Algorithms set the upper limit, but engineering determines how quickly things land. Tencent's biggest ace, he says, is its real-world scenarios: WeChat-based office work, documents, meetings, and enterprise services.
He also did something rare: revisited several missteps. Over the past year, Yuanbao spent heavily on promotion to chase user growth, but at the time "the model and product weren't ready," and the results were unsatisfying. CodeBuddy and WorkBuddy were also not part of some carefully planned success route; in the early days, the developer tools lost money for a long time and nearly got cut.
WorkBuddy later caught the agent wave, shipping over 40 versions in its first three months. Tong says his team now uses AI extensively to write code, with people mainly in charge of judgment, debugging, and quality control. He predicts that the boundaries between product, development, and testing will blur further, and that small teams, with AI, will be able to accomplish what used to require a large team.