HackerNews中文版

我使用每月 20 美元的 Claude 套餐，主要用于研究、“橡皮鸭调试”以及作为想法的“声音板”（我之前使用 Fable 来做这件事，但现在只能使用 Opus 4.8 或 Sonnet 5），最近我注意到 Opus 4.8 在高上下文情境下的响应质量严重下降。为了尝试解决这个问题，我禁用了记忆系统（现在标记为“遗留”功能，尽管据我所知并没有替代品）。禁用此设置后，尽管每次都需要输入必要的上下文来继续任务，这有些繁琐，但我注意到响应质量和准确性有了显著的提高——我是那些在 Opus 4.7 发布后抱怨过的人之一，我的理论是，为了增加记忆上下文而进行了一些更改，结果导致实际响应质量下降。总的来说，我怀疑最先进的思考模型的“高上下文”能力是否被夸大了。作为简短的轶事，我曾进行一项研究任务，输入了 30 多个提示。切换到一个新的聊天，并将讨论的要点浓缩成一个单一的新提示，其结果优于继续在之前的聊天中进行提示，尽管（或者我可以说，正是因为）之前的对话在主题上拥有丰富的上下文。附注：我希望“项目”和“聊天”的记忆是分开的设置。禁用记忆会增强聊天的能力，但会严重削弱项目的能力，因为它禁用了项目的内部记忆，导致模型每次要求引用文件时都要费力地查找。我想一个临时的解决方案就是为每个任务创建一个项目，但这似乎很繁琐。这是一个已知现象吗？最近的更改是否加剧了这种情况？任何想法和建议都将不胜感激。谢谢！

查看原文

I use Claude on the $20/mo plan, mainly for research, "rubber duckying," and as an idea soundboard (I was using Fable for this but am now relegated to Opus 4.8 or Sonnet 5), and recently noticed a severe downturn in Opus 4.8 response quality in high-context situations. To attempt to remedy this, I disabled the memory system (now labeled "Legacy" even though to my knowledge there is no replacement).<p>With this setting disabled, aside from the slight tedium of inputting the necessary context each time for task continuations, I have noticed a drastic improvement in response quality and accuracy—I'm one of those shmucks who whined a bit after Opus 4.7 came out, and my theory is some change was made to increase memory context that resulted in a decrease in actual response quality. In general, I wonder if the high-context abilities of SOTA thinking models are overstated.<p>As a brief anecdote, I was working through a research task with upwards of 30 prompts. Switching to a new chat and condensing the gist of the discussion into a single new prompt yielded superior results than continuing to prompt the prior chat, despite (or due to, I would argue) the prior conversation's wealth of context on the subject.<p>Aside: I wish that "project" and "chat" memory were separate settings. Disabling memory boosts the capabilities of chat, but severely nerfs projects by disabling their internal memory, leading the model to grope around for files each time you ask it to reference one. I suppose a hacky solution could just be to create a project for everything, but this seems tedious.<p>Is this a known phenomenon? Have recent changes exacerbated things? Any thoughts and tips are appreciated. Thanks!

HN 提问：你们是否注意到禁用记忆功能后，AI 的回复有所改善？