DeepSeek founder Liang Wenfeng, in an investor call, described the company as driven by vision rather than formal structure, stating, "We don't have an organization; we're driven by a vision," with no KPIs, no written vision, and "no organization." The company operates on "goodwill toward the world" and an obsession with AGI, with API pricing aimed only to "recover hardware costs in ten months." Liang emphasized open source as central, saying, "Even the strongest model will be open-sourced," and noted, "I only worry they won't be able to deploy it." The AGI roadmap includes "CoT → Agents → continual learning → 'singularity' → embodied intelligence," with the goal that "after continual learning, it can develop the next version itself." Compute resources consist of about "20,000 'H-equivalent' GPUs," placing the company "12–18 months behind the U.S." and "chasing with one‑twentieth the compute." For domestic chips, "the ecosystem is fine; capacity is the bottleneck." Liang predicted China "may match foreign models this year," but cautioned "that still isn't AGI," and noted "there are too many base-model companies in China; it will converge."