Skip to content

Commit 64ff966

Browse files
committed
bug: remove useless content
1 parent 80a76a7 commit 64ff966

1 file changed

Lines changed: 2 additions & 29 deletions

File tree

content/zh/celebrity_insights/Andrej_Karpathy_2025_Review.md

Lines changed: 2 additions & 29 deletions
Original file line numberDiff line numberDiff line change
@@ -9,8 +9,8 @@ description: "Karpathy大神在2025年年底的时候多大模型这一年的发
99
cover:
1010
image: "/images/generated-covers/Andrej_Karpathy_2025_Review.webp"
1111
alt: "Karpathy: 2025 年大语言模型年度回顾"
12-
wordCount: 3281
13-
readingTime: 9
12+
wordCount: 2875
13+
readingTime: 8
1414
---
1515

1616
Andrej Karpathy大神在2025年年底的时候多大模型这一年的发展做了一个“年度总结”,我用Qwen3 Max做了翻译,供大家阅读。我在文后附上了我个人的看法。
@@ -178,33 +178,6 @@ Google 的 **Gemini Nano Banana** 是 2025 年最令人震撼、最具范式颠
178178

179179
---
180180

181-
### ▸ TL;DR
182-
2025 年,是 LLM 令人激动又略带意外的一年。
183-
它们正涌现出一种**新型智能**——比我预想的**更聪明,也更愚蠢**
184-
无论如何,它们已极具实用价值;而依我看,即便以当下能力,行业也远未挖掘其 10% 的潜力。
185-
186-
与此同时,**可探索的创意仍如繁星般密集,整个领域在概念层面依然无比开放**
187-
正如我今年早些时候在 Dwarkesh 播客中所说:
188-
> 表面上矛盾,实则并行不悖——
189-
> **我既相信进展将持续快速,也深知前路仍有海量工作待完成**
190-
191-
系好安全带,旅程才刚刚开始。
192-
193-
🔗 附:此文我也同步发到了[个人博客](URL),排版更清爽,阅读体验更佳。
194-
195-
---
196-
197-
翻译说明:
198-
- 保留 Karpathy 标志性的“冷静中带温度、严谨中见幽默”的语感;
199-
- “ghosts vs. animals”译为“幽灵 vs 动物”,既保留隐喻张力,又契合中文科幻语境;
200-
- “jagged intelligence”译为“锯齿状智能”,准确传达“能力不均衡”的核心意象;
201-
- “benchmaxxing”创造性译为“刷榜内卷”,符合中文技术圈语境;
202-
- “vibe coding”音意兼顾译为“氛围编程”,已成社区共识译法;
203-
- 技术术语(如 RLVR, DAG, CLI, BPE)保留英文缩写+中文注释,确保专业性与可读性平衡。
204-
205-
希望这份翻译,能让中文读者感受到 Karpathy 原文那种——**站在智能革命前沿的清醒、兴奋与哲思交织的独特气息**
206-
207-
208181
## x.com上的原文
209182

210183
2025 has been a strong and eventful year of progress in LLMs. The following is a list of personally notable and mildly surprising "paradigm changes" - things that altered the landscape and stood out to me conceptually.

0 commit comments

Comments
 (0)