DeepSeek before V4: Liang Wenfeng's talent war, org philosophy and AGI obsession

Why it matters

In April 2026, just before DeepSeek V4's release, LatePost published perhaps the deepest insider report on DeepSeek to date. Starting from three core researchers leaving after Chinese New Year, it reveals the real tensions: talent poached by ByteDance, Xiaomi, and Tencent; Liang's uniquely 'no-overtime' management philosophy; the strategic gamble of pivoting to domestic chips; and his non-mainstream AGI vision — not racing benchmarks, but 'ecosystem building + exploring minority directions'.

Full Interview

Key takeaways

• 'A person can only produce high-quality work for 6-8 hours a day' — the only no-overtime core AI lab globally; not laziness, compute is too expensive.
• The root of talent drain: DeepSeek's goals aren't mainstream and pay can't match big tech — Liang bets on 'minority direction' value.
• V4 delays rooted in domestic chip pivot: 400K operator lines rewritten, 35x speedup — trading time for strategic autonomy.
• Liang's AGI vision: not just model intelligence, but 'ecosystem building' and 'original innovation' — directions big tech won't explore.
• 'Those who keep their heads down may not laugh last, but with more DeepSeek-like companies, China can move from copying to leading.'