HN Digest podcast

Gemini's mid-size model beats its flagship

19.11.2026
0:00
12:19
15 Sekunden vorwärts
15 Sekunden vorwärts

Google's Gemini 4 Neon edges past its own flagship, a test suggests Claude learned protein sequences, and OpenAI says its models helped design chips. Plus a nested KVM bug and an essay on the people who keep systems running.

  1. Gemini 4 Neon (HN discussion)
  2. Claude was trained on protein sequences at scale (HN discussion)
  3. Advancing chip design with OpenAI models (HN discussion)
  4. KVM: x86: Unconditionally reset MMU context when emulating INIT (CVE-2026-99312) (HN discussion)
  5. We need smart people who give a shit (HN discussion)

Chapters:
(00:00) Intro
(00:41) 1. Gemini 4 Neon
(01:31) 1. Gemini 4 Neon: HN comments
(02:56) 2. Claude was trained on protein sequences at scale
(04:19) 2. Claude was trained on protein sequences at scale: HN comments
(05:07) 3. Advancing chip design with OpenAI models
(06:15) 3. Advancing chip design with OpenAI models: HN comments
(07:01) 4. KVM: x86: Unconditionally reset MMU context when emulating INIT (CVE-2026-99312)
(08:01) 4. KVM: x86: Unconditionally reset MMU context when emulating INIT (CVE-2026-99312): HN comments
(09:25) 5. We need smart people who give a shit
(10:37) 5. We need smart people who give a shit: HN comments
(11:26) Wrap-up

Weitere Episoden von „HN Digest“