AITOPAI热报
IndustryMulti-source

Simon Willison：Claude Opus 5.5 让 Vibe Gaming 成为拐点

向阳乔木@vista8
Post· 11 sources
View the post
Why we picked it

Simon 又发观察文了：他说 Opus 5.5 连 3D、音乐、游戏设计都能一手包办，GPT-6 Astra 暂时还差点趣味感，做游戏的可以看看。

博主引用 Simon Willison 2026 年 10 月的文章，称 2025 年 11 月的 Opus 4.5 与 GPT-5.1 曾把 Vibe Coding 推到拐点。2026 年 10 月发布的 Claude Opus 5.5 在 3D 建模、音乐生成、游戏设计能力上大幅提升，使 Vibe Gaming 成为新拐点。文章提到 GPT-6 Astra 也能做游戏，但趣味感不如 Opus 5.5，预计年底才能赶上。瓶颈从模型能力转移到了创意点子和生态。

Image · 向阳乔木
Original · 向阳乔木

看 Simon 老哥写的最新一篇文章还挺有同感的。 2025 年 11 月，Opus 4.5 / GPT-5.1 让 Vibe Coding 成为拐点。 前几天看 DHH 的访谈，他也是从Opus 4.5 开始真正开始被 AI 能力震撼到的。 今年2026 年 10 月，Claude Opus 5.5 让 Vibe Gaming 成为拐点。 主要是能力大幅提升，3D 建模、音乐生成、游戏设计。 瓶颈已转移到“点子”和生态。 GPT-6 Astra 能做游戏，但趣味感不如Opus 5.5，估计年底才能赶上。 simonwillison.net/2026/Oct/28/vi… 💬 18 🔄 1 ❤️ 22 👀 5930 📊 20 ⚡

#Claude Opus#GPT-6 Astra#Vibe Coding#游戏生成

Keep reading

Recent stories with the same tag or section
ModelsUS vs ChinaMulti-sourceSelected01:57
CritPt v2 Benchmark Launches with Agentic Code Execution

The 54-point gap between GPT-5 and GPT-6 Astra in about a year shows how quickly frontier models are absorbing research-level physics when given code execution. Tracking these scores before the late October Intelligence Index update helps researchers and benchmark watchers pick models for scientific workflows. Compare against Claude Opus 5.5 at 78% when choosing a frontier model for agentic computation tasks.

#GPT-6#Claude Opus 5.5#CritPt#Artificial Analysis
Event
官方账号Artificial Analysis@ArtificialAnlys11 sourcesSource
PapersBuildersMulti-sourceSelected22:25
Anthropic's Agent4Sci Benchmarks Coding Agents on Scientific ML

The scaffolding, not the model, roughly doubled the improvement — a design lesson for anyone building AI-scientist or agent systems. Builders of research agents can copy the three skill modules (ideation, experiment management, scheduled checklists) and benchmark against Agent4Sci's 20 tasks. The paper also exposes predictable agent failure modes, such as wasted compute and narrow local sweeps, that teams can check for directly.

#Anthropic#Claude#Codex#Agent4Sci
Event
DAIR.AI@dair_ai11 sourcesSource