Archives

2026 / 07

"OpenAI found that 43.5% of occupation-specific ChatGPT messages concern tasks associated with another occupation. The early signal is task redistribu
"PrismML 称 1-bit Bonsai 27B 只需 3.9GB、在 iPhone 上达到每秒 11 token,并保留 FP16 平均分的 89.5%。这几个数字分别意味着什么?"
"PrismML says its 1-bit Bonsai 27B occupies 3.9GB, runs at 11 tokens per second on an iPhone, and retains 89.5% of FP16 performance. Here is what each
Netflix 约 300 个片目使用过 GenAI 工作流。本文校准这一数字的证据边界,并把公开指南转成可执行、可审计的三段式生产门槛。
Netflix AI guidelines explained: which GenAI uses need written approval, what 300 titles reveal, and how production gates should work.
用单 Token 样本为 LLM API 建立低成本漂移传感器,并严格区分变化检测、模型身份和根因归因三层证据。
Build a cheap behavioral drift sensor for LLM APIs, then use an evidence ladder to keep detection separate from identity and root-cause claims.
一套可回滚的 GPT-5.6 Prompt 删减实验方法:逐组删除指令,用同一套 Eval 守住质量、权限、证据和输出契约。
A controlled method for making GPT-5.6 prompts leaner without deleting the constraints that keep production agents reliable.
将 Claude Cookbook 示例升级为有版本、可重复、带权限断言、成本预算与发布门禁的生产 Agent 回归测试集。
Previous Next