natolambert

X · 29 atoms · 2026-06-24 → 2026-06-29 · 信用档先验 cred3

🟦 fact 3 · 🟥 take 26 · stance ▲3/▼19/◆2

覆盖实体 (top 10)

• Anthropic — 11 atoms
• GLM 5.2 — 2 atoms
• GLM — 2 atoms
• USG — 2 atoms
• 开源模型 — 1 atoms
• AI companies — 1 atoms
• discussion about distillation — 1 atoms
• AI executives — 1 atoms
• potential solutions for distillation — 1 atoms
• ASE — 1 atoms

话题分布 (top 8)

政策影响 (1) · 版本迭代 (1) · 模型竞争 (1) · 模型使用策略 (1) · 模型成本 (1) · 公关策略 (1) · 信息透明度 (1) · 蒸馏指控 (1)

样例 atoms

• 2026-06-27 · Anthropic · I dont really blame anthropic for this, i think they made it earlier, but it was going to happen during this admin who didnt really have the technical capacity to handle it well.
• 2026-06-27 · Anthropic (fear mongering) · Fear mongering = earlier, maybe a bit worse implementation, yes I agree. I attribute a form of cause to that. but in late 2027 the models will be way stronger and made by many peop
• 2026-06-27 · 压缩推理轨迹实验 · I've seen many experiments on summarized traces fail
• 2026-06-27 · 压缩推理轨迹 · it'll muck with the transition to rl
• 2026-06-25 · USG · In the last ~3 months they've destroyed all that trust and dug a giant hole of uncertainty - where the US's AI leadership will now be actively degrading.

← Author 索引 · 实体目录