第50期 | Changes in the system prompt between Claud
今日摘要
Simon Willison：Anthropic are the only major AI lab to publish the system prompts for their user-facing chat systems. Their system prom…
Simon Willison：Anthropic's published system prompt history for Claude is transformed into a git-based exploration tool, breaking up th…
Simon Willison：Adding a new content type to my blog-to-newsletter tool - Agentic Engineering Patterns Guides > Agentic Engineering Pat…
Simon Willison：This year’s PyCon US is coming up next month from May 13th to May 19th, with the core conference talks from Friday 15th…
Hugging Face Blog：nvidia/nemotron-ocr-v2 Image-to-Text • Updated 5 days ago • 917 • 121
总结 + 观点：Anthropic are the only major AI lab to publish…｜中文观点：Claude 系统提示词公开最有价值的地方，是让开发者能更具体地比较产品边界、对齐策略和版本变化。
总结 + 观点：Anthropic's published system prompt history for…｜中文观点：把系统提示词做成 git 时间线很实用，因为它把“模型行为变化”从感觉问题变成了可追踪的工程事实。
总结 + 观点：Adding a new content type to my blog-to-newslet…｜中文观点：这类小而完整的自动化案例很有参考价值，说明内容工作流的收益常常来自细碎但可复用的工具改造。
总结 + 观点：This year’s PyCon US is coming up next month fr…｜中文观点：PyCon 把 AI 和安全单独拉出来，说明这两个话题已经从兴趣分支变成了主流开发生态的正式议程。
总结 + 观点：nvidia/nemotron-ocr-v2 Image-to-Text • Updated…｜中文观点：OCR 模型真正值得看的是部署成本和多语言稳定性；如果这两点站住了，它会直接进入文档处理链路。
Changes in the system prompt between Claude Opus 4.6 and 4.7
标签：#ai_engineering_blogs #core
作者：
原文：Anthropic are the only major AI lab to publish the system prompts for their user-facing chat systems. Their system prompt archive now dates all the way back to Claude 3 … 18th April 2026 Anthropic are the only major AI lab to publish the system prompts for their user-facing chat systems.
链接：https://simonwillison.net/2026/Apr/18/opus-system-prompt/#atom-everything
Claude system prompts as a git timeline
标签：#ai_engineering_blogs #core
作者：
原文：Anthropic's published system prompt history for Claude is transformed into a git-based exploration tool, breaking up the monolithic markdown source into granular files and timestamped commits. By structuring extracted prompts … 18th April 2026 Anthropic publish the system prompts for Claude chat and make that page avai...
链接：https://simonwillison.net/2026/Apr/18/extract-system-prompts/#atom-everything
Adding a new content type to my blog-to-newsletter tool
标签：#ai_engineering_blogs #core
作者：
原文：Adding a new content type to my blog-to-newsletter tool - Agentic Engineering Patterns Guides > Agentic Engineering Patterns Here's an example of a deceptively short prompt that got a lot of work done in a single shot. First, some background.
Join us at PyCon US 2026 in Long Beach - we have new AI and security tracks this year
标签：#ai_engineering_blogs #core
作者：
原文：This year’s PyCon US is coming up next month from May 13th to May 19th, with the core conference talks from Friday 15th to Sunday 17th and tutorial and sprint … 17th April 2026 This year’s PyCon US is coming up next month from May 13th to May 19th, with the core conference talks from Friday 15th to Sunday 17th and tuto...
链接：https://simonwillison.net/2026/Apr/17/pycon-us-2026/#atom-everything
Building a Fast Multilingual OCR Model with Synthetic Data
标签：#ai_engineering_blogs #core
作者：
原文：nvidia/nemotron-ocr-v2 Image-to-Text • Updated 5 days ago • 917 • 121
Engineering at Anthropic: Inside the team building reliable AI systems
标签：#ai_engineering_blogs #core
作者：
原文：Agentic coding benchmarks like SWE-bench and Terminal-Bench are commonly used to compare the software engineering capabilities of frontier models—with top spots on leaderboards often separated by just a few percentage points. These scores are often treated as precise measurements of relative model capability and increa...
链接：https://www.anthropic.com/engineering/infrastructure-noise
Scaling Managed Agents: Decoupling the brain from the hands
标签：#ai_engineering_blogs #core
作者：
原文：Get started with Claude Managed Agents by following our docs . A running topic on the Engineering Blog is how to build effective agents and design harnesses for long-running work .
[AINews] Anthropic Claude Opus 4.7 - literally one step better than 4.6 in every dimension
标签：#ai_engineering_blogs #core
作者：
原文：AINews: Weekday Roundups [AINews] Anthropic Claude Opus 4.7 - literally one step better than 4.6 in every dimension The new SOTA model asserts its dominance. Apr 17, 2026 ∙ Paid 65 4 Share Thursday mornings are for prestige AI launches, and while OpenAI put in a valiant effort with GPT-Rosalind and The New New Codex (w...
链接：https://www.latent.space/p/ainews-anthropic-claude-opus-47-literally
llm-anthropic 0.25
标签：#ai_engineering_blogs #core
作者：
原文：LLM access to models by Anthropic, including the Claude series 16th April 2026 This is a beat by Simon Willison, posted on 16th April 2026 . Sponsor me for $10/month and get a curated email digest of the month's most important LLM developments.
链接：https://simonwillison.net/2026/Apr/16/llm-anthropic/#atom-everything
Qwen3.6-35B-A3B on my laptop drew me a better pelican than Claude Opus 4.7
标签：#ai_engineering_blogs #core
作者：
原文：For anyone who has been (inadvisably) taking my pelican riding a bicycle benchmark seriously as a robust way to test models, here are pelicans from this morning’s two big model … 16th April 2026 For anyone who has been (inadvisably) taking my pelican riding a bicycle benchmark seriously as a robust way to test models,...
链接：https://simonwillison.net/2026/Apr/16/qwen-beats-opus/#atom-everything