Unique Research · Insight / Unique Friend Q&A · January 15, 2024 · Guest: Chen Dabo, Founder & CEO, Cobuxi / Fabrie (柯布西)
Inventory type remains insight (Unique Friend series — not partner welcome).
Guest
Chen Dabo — Founder & CEO, Cobuxi; designer & graphics-algorithm engineer. Imperial College / RCA Innovation Design Engineering; Central Saint Martins product design. Tangerine (London), Nike; founded Evercise and Ally Labs in London. Cobuxi builds Fabrie — online collaborative whiteboard for internet, consumer, apparel, games and creative industries; multi-round USD fund backing.
Q&A
1. Product and core scenarios?
Fabrie + GenAI for product/creative/design/ops teams: internal collab, market analysis, design review, AI-assisted exploration. AI lifts design efficiency; whiteboard as material/doc space becomes a core work UI alongside Feishu-class docs.
2. R&D / GTM; biggest 2023 challenge and win?
Whiteboard live ~1.5 years — unique whiteboard + multi-dimensional table form; top-tier performance (~10k images smooth on one board; 1,000+ realtime collab). AI beta >6 months; 100M+ images generated; strong pro-user and enterprise internal wins. Challenge: keep up with tech, find niche moat, and find the era’s breakthrough in a shifting market.
3. GenAI opportunities vs prior waves?
Barrier collapse lets non-technical founders ship — diversity feeds the ecosystem. Prior waves redistributed resources; this one hits productivity itself — 11-, 4-, 2-person companies and solopreneurs; “success” being redefined.
4. China landing challenges?
Hallucination, compliance/IP, thin value — but key China tension: low labor-cost markets may not suit first-wave AI displacement. Replacing ¥1,000 labor may yield <¥100 revenue if the model is only the spread — need new product forms and era-fit business models.
5. Competing with giants?
Dimensional reduction isn’t risk — it’s inevitable. Do markets/scenes giants won’t. Prior first-mover/flexibility advantages shrink when everything accelerates.
6. Are LLMs still too weak?
Like mobile internet’s image then video nodes — don’t wait for full AGI. Split modalities by maturity; today Fabrie watches reasoning more than generation (GPT-4). With time: layout for AGI-state value rather than only patching today’s gaps with fine-tunes — leaps will dwarf fine-tune edges (cites Altman’s GPT-5 hints).
7. Human–machine collaboration?
HCI once assumed fixed machines; AI is a person-varying interface — hence prompt engineering. As AI becomes more “human,” interaction may get harder, not easier — maybe keep some machine traits for balance. Enterprise: “AI replaces tasks, not jobs” — land by extracting AI-able links. Often a design director (or hired analog) threads AI into process — needlework, not sweeping cuts.
8. 2024 tech to watch?
Vision Pro / spatial computing (personally cautious but watch). AI: on-device inference — not just cost; more flexible business models and new scenes.
1 CN image; attach on ingest. Signup CTA in CN omitted.
— END —