Four Chinese AI Creators Launch Weekly Podcast 'Next Token' on Model Wars
vista8 · x · 2026-09-05
A new Chinese-language AI podcast, 'Next Token' (词元之外), launched by four creators — Yang Pan, Xiangyang Qiao, Guizang, and Ju AI — covering weekly AI industry news and model commentary. Episode 1 (news window Aug 27-Sep 3) covers:
- Model 'kill-line' wars: GLM 5.3 Flash, Qwen 3.8 Flash Next, and MiniMax H3 releases shifting competition from raw capability to 'cheap and good enough'; benchmarks losing meaning as free trials and real tasks become the selection logic.
- Agent products: native model + native agent combos, always-on cloud computers (Grok Bot), progressive authorization, DeepSeek Harness evals (completion rates and token costs), and the convergence of OpenClaw with coding agents.
- Hardware & local context: Microduck's viral rise, ESP32 as 'new LEGO', local models in browsers and robots, NAS as a home for personal AI context.
- Token economics: top models handle creation while cheap models expand consumption, but compute supply, margins, and org iteration speed decide whether AI apps become real businesses.
More from Companies & People
- MIT's Spring 2026 multimodal AI course by Paul Liang goes public — caglar_ee · 2026-09-05
- Founder-led AI giants dominate: XAI, Meta, OpenAI, Anthropic vs Google and Apple — beffjezos · 2026-09-05
- Notion positions itself as the "Switzerland of AI" as models default to it for memory — ivanhzhao · 2026-09-05
- Fiora, author of viral AI alignment essays, is jobless and looking for work — repligate · 2026-09-05
- OpenAI Wants to Talk About 'The Federalist Papers': Inside Its Constitution Debate — Electronic-Bus-3494 · 2026-09-05
- X wins court order blocking rival startup app from using the Twitter name — Polymarket · 2026-09-05