标签: ChatGPT

I wish Sign in with @Google (or sign in with @OpenAI / @claudeai) came with ability to use tokens directly from the customers account.. That way we don’t have to pick between asking them to generate an API key or usin…

I wish Sign in with @Google (or sign in with @OpenAI / @claudeai) came with ability to use tokens directly from the customers account.. That way we don’t have to pick between asking them to generate an API key or usin...

开发者 Nikunj Kothari 在 X 上提出设想:如果“Sign in with Google/OpenAI/Claude”这类第三方登录,能顺带为开发者生成一个仅限当前域名使用的 API Token,那么小型 AI 应用就不必在“让用户去生成 API Key”和“开发者垫付推理成本”之间二选一。这个…

i built a codex-native that maps your psychology about technological change and then tells you how you’d react to different eras of technology change throughout history https://t.co/bMxVKxpAjh

i built a codex-native that maps your psychology about technological change and then tells you how you'd react to different eras of technology change throughout history https://t.co/bMxVKxpAjh

Dan Shipper 构建了一款基于 Codex 的个性化测试应用,它先分析你对技术变革的心理倾向,再模拟你如果身处不同历史技术变革时期会如何反应。这件事的价值在于把大模型能力从“工具使用”延伸到“自我认知与历史情景推演”。

We’ve made some improvements that improve usage on the long tail for power users of Astra when logged in with your ChatGPT account. No change in quality and a pure win that on the long tail can result in up to 3-4X le…

We've made some improvements that improve usage on the long tail for power users of Astra when logged in with your ChatGPT account. No change in quality and a pure win that on the long tail can result in up to 3-4X le...

OpenAI 工程师 Thibault Sottiaux 透露,团队对 Astra 在 ChatGPT 账号登录状态下的长期使用效率做了底层优化,在不改变输出质量的前提下,重度用户的长尾使用最多可减少 3-4 倍订阅额度消耗。这意味着同样的订阅套餐,重度用户实际可用的对话轮次或时长将显著增加。

动态工作流中Claude Code的token消耗削减80%

动态工作流中Claude Code的token消耗削减80%

有开发者在 HN 分享动态工作流优化实践,声称将 Claude Code 的 token 消耗削减了 80%。相比具体数字,这篇讨论的价值在于展示了大模型编码工具在复杂任务下实际可被大幅压缩的推理开销,以及动态上下文管理带来的真金白银的算力节省。

How we monitor internal coding agents for misalignment

How we monitor internal coding agents for misalignment

OpenAI 首次公开了其内部代码智能体的监控系统,用 GPT-5.4 Thinking 实时审查智能体的思考链和操作,目的是在真实工作流中识别“意图不一致”的行为。值得关注的是,这是少数几家前沿实验室公开披露此类内部安全基础设施的案例。