如何充分利用Fable 5和GPT-5.6 Sol

今天在《AI每日简报》中，我们来聊聊如何充分挖掘前沿模型的潜力。

Today on the AI Daily Brief, how to get the most out of frontier models.

《AI每日简报》是一档每天更新的播客和视频节目，聚焦人工智能领域最重要的新闻与讨论。

The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI.

好了，各位朋友。

Alright, friends.

在正式开始之前，先说几件小事。

Quick announcements before we dive in.

首先，感谢今天节目的赞助商：毕马威、Blitzy、Retool 和 Airtable。

First of all, thank you to today's sponsors, KPMG, Blitzy, Retool, and Airtable.

想收听无广告版本的节目？请访问 patreon.com/aideallybrief，或者在 Apple 播客上订阅。

To get an ad free version of the show, go to patreon.com/aideallybrief, or you can subscribe on Apple Podcasts.

如果您想了解节目赞助详情，欢迎发送邮件到 sponsors@AIDailyBrief.AI。

To learn more about sponsoring the show, send us a note at sponsors@AIDailyBrief.AI.

另外，简单提一下今天的节目内容。

And a quick note about today's episode.

这段内容是提前录好的。

This was recorded in advance.

实际上，我现在正在周四下午录音，而大家都在为Kimmy K3闹得沸沸扬扬。

In fact, I am recording it on Thursday afternoon as everyone freaks out about Kimmy k three.

所以这段时间，如果达里奥和萨姆已经疯了，为了应对正在冲击纳斯达克市值的威胁，紧急发布了Fable和GPT的新版本，那你现在就明白我为什么没提这些了。

So in the meantime, you know, if Dario and Sam have lost their minds and released new versions of Fable and GPT in response to the threat that's tearing value off the Nasdaq, you'll know why I am not talking about it right now.

周一我有点行程要处理。

Just a quick little bit of travel on Monday.

我周二会回来，带来一集正常的节目。

I'll be back on Tuesday with a normal episode.

但不管外面的模型世界发生什么变化，今天的节目重点还是教你如何充分利用最新、最强大的模型。

But still, regardless of what is going on in the wide world of models out there, today's episode is all about how to get the most out of the most advanced and newest models.

好了，不多说了，我们直接开始。

So without any further ado, let's dive in.

Fable 5和GPT-5.6这一批新模型面世，已经过去几周了。

It has now been a couple of weeks with this new class of models in Fable five and GPT-5.6.

奇怪的是，这些模型实际上已经存在一段时间了，不止短短几周。

Now, weirdly, these models have actually been around a little longer than a couple of weeks.

GPT-5.6 有过很长的早期试用期，而 Fable 五代则短暂出现过几天就消失了。

There was a particularly long early access period for GPT-5.6, and Fable five was here for a couple days before going away.

但现在，几乎所有人都已经用这些模型有一阵子了。

But at this point now, pretty much everyone has now had these models for some time.

事实上，我们对它们的访问权限还在不断延长和重置。

In fact, our access to them keeps getting extended and reset.

与此同时，大家也开始分享如何更好地使用这些模型的技巧。

And along with that, people have started to publish their tips and tricks for getting the most out of them.

新模型总是需要新的交互方式，才能发挥出最大潜力。

Now it is always the case that new models demand new ways of interacting with those models to get the most out of them.

但这类经验很难通过测试指标来捕捉，只能靠反复尝试去体会。

But this is exactly the sort of information that can't really be captured in anything like benchmarks and just has to go be experienced through trial and error.

你会发现，5.6 Sol 和 Fable 五代之间有不少共同点，这说明的不只是使用这些模型的新方法，更可能预示着未来会越来越普遍的交互模式。

As you will see, there are a number of common threads that cut across both 5.6 Sol and Fable five that suggest, I think, not just some new ways to get the most out of these models, but for some new patterns of interaction that are going to become increasingly common from here on out.

我们现在先来看看一些官方来源和评论。

Now we're gonna start with some official sources and commentary.

Codex团队的成员埃里克·普罗文查尔提到，很多人在使用5.6 Sol时，还是像以前用5.5那样去提问。

Codex team member Eric Provenchar wrote, with 5.6 Sol, a lot of people are still prompting the model exactly as they did 5.5.

需要注意的是，5.6 Sol比之前的模型要更执着、更细致。

It's important to note that 5.6 Sol is a lot more tenacious and thorough than previous models.

埃里克在官方学习网站learn.chattypt.com上发布了一份提示指南。

Eric published a prompting guide on the official learn.chattypt.com site.

虽然这次并没有明确说这是为了更好地利用5.6而不是5.5，但里面还是有几个点特别值得留意。

And while in this case, it is not presented as a way to get more out of five six as opposed to five five, there are a few things that are specifically worth noting.

其中一条建议，源于模型的这种执着，就是关于设定边界。

One piece of guidance that comes from that tenacity is around setting boundaries.

指南里说，边界就是那些让ChatGPT避免多做无用功或执行你本不想让它做的事的几条指令。

Boundaries, writes the guide, are the few instructions ChatGPT needs to avoid creating extra work or taking an action you didn't intend.

当你改错一个细节就会让结果完全没法用，或者你希望在影响他人之前先确认一下时，就该加上这些边界。

Add one when changing the wrong detail would make the result unusable or when you want to review something before it affects other people.

保持已批准的日期和预算数字不变。

Keep the approved dates and budget figures unchanged.

只使用提供的资料。

Use only the supplied sources.

遇到信息缺失时，标注出来，不要猜测。

Flag missing information instead of guessing.

建议内容必须控制在预算范围内。

Keep recommendations within the stated budget.

把消息先写成草稿。

Prepare the message as a draft.

别发出去。

Don't send it.

你可以看到，在这些情况下，像埃里克说的，更执着的模型可能会擅自多做主，做出一些用户并不想让它们做的事，结果反而偏离了原本想达成的目标。

You can see how in each of these cases, a more tenacious model, to use Eric's word, might assert a little bit more agency than the user would be comfortable with and actually go off and do something that ended up not being optimal for whatever the prompter was trying to achieve.

其中一个明确的界限，就是清楚说明你希望它在哪儿停下，以及哪些事情绝对不要做。

One example of these boundaries was being explicit about where you wanted it to stop and actions you didn't want it to take.

我。

I.

E。

E.

别发送消息，只把它当作草稿准备好。

Don't send the message, just prepare it as a draft.

另一个关于批准日期和预算的限制，是限制并聚焦于这个执着模型所关注的范围。

Another example around the approved dates and budget figures was to limit and focus where the tenacious model was applying its attention.

关于只能使用所提供资料的要求，如果一个执着的模型可以访问整个互联网，而没有被明确禁止，它很可能会陷入严重的无关信息中。

When it comes to the injunction to use only the supplied sources, a tenacious model that has access to the entire Internet could go off and get lost in some serious rabbit holes if not told not to do that.

当然，设定界限本身并不新鲜，但这里的关键是，模型越强大，这些界限就越重要，如果没设好，现实中的后果也会更严重。

Now, obviously, setting boundaries is nothing new, but the point here is that the more powerful the model, the more significant those boundaries become, and the more real world consequences there can be if those boundaries aren't set.

在后果较轻的一端，只是消耗了远超实际需要的计算资源，而在越来越注重成本的环境下，这也不是小事。

On the lower end of the spectrum of consequence, there's just burning through way more tokens than you actually needed to, which in an increasingly cost conscious error isn't nothing.

但更严重的后果当然也存在，比如把尚未获批的消息发给了关键的客户或同事。

But then, of course, there's much bigger consequences, like sending a message that hasn't been approved yet to a critical customer partner or colleague.

埃里克还给了另一个提示，虽然这并不新鲜，但确实符合5.6 Sol这个模型的特点——你真正想做的不是丢个任务给它就完事，而是跟它互动起来，通过后续消息不断优化结果。

Another prompting tip from Eric, which again isn't different but certainly comports with the idea of 5.6 Sol being a model where you actually want to interact with it as opposed to just give it a task and let it go off and do its thing, is to iterate with the model to improve the results with follow-up messages.

看到第一版后，你可以接着说：开头再直接点，保留证据，把这部分挪到那边去。

Once you see the first version, you can follow-up with things like keep the opening more direct, keep the evidence, move this or that part around.

有趣的是，现在这种迭代不再需要一来一回地等回复了，你不用非得等模型输出完才能调整方向。

And interestingly, increasingly iteration isn't going to happen in a completely turn based way where you have to stop and wait for an output before you can steer where the model goes next.

随着ChatGPT和Codex逐渐融合，Codex里那些常见的引导方式，也会慢慢融入主应用的体验中。

As ChatGPT and Codex come together, some of the sort of steering behaviors that are normal in Codex can come to that main app experience as well.

指南里提到，当Codex正在运行时，你完全可以不等它完成，直接发下一条消息。

The guide writes, when Codex is already working, you can send another message without waiting for the current run to finish.

‘Steer’会把消息加到当前任务里，用来调整方向、补充遗漏的细节，或者提供新信息。

Steer adds the message to the current run, use it to change direction, add a missing detail, or share new information.

‘Queue’则会把消息存起来，等当前任务结束后再作为后续指令执行。

Queue saves the message for the next run, use it as a follow-up that should wait until the current work finishes.

这种交互方式大大降低了与AI协作的延迟，而随着模型越来越强大，这一点也变得越来越关键。

This sort of behavior reduces the latency of collaborating with AI in ways that become important, the more powerful the models get.

如今，随着Codex和ChatGPT的整合，以及ChatGPT被拆分为聊天和工作两种模式，针对不同使用场景的提示技巧也有了区别。

Now one of the things that follows from the recent integration of Codex and ChatGPT and the split of ChatGPT into chat and work is different prompting tips for different types of experiences.

在这篇文章中，Eric和ChatGPT团队明确区分了用于聊天和用于工作的最佳提示方式及示例。

In this post, Eric and the team at ChatGPT explicitly separate prompting best practices and examples for chat as opposed to for work.

很多关于工作的建议都融入了成本和效率的考量。

A lot of the tips around work have this element of cost and efficiency consciousness embedded within them.

其中有一节叫‘高效使用工作模式’，里面提到：工作模式适合处理耗时或重复性的任务，或者生成可以重复使用的完整文件。

One section is called Use Work Efficiently and says: Work is useful for time consuming or recurring tasks or for finished files you can reuse.

即使一个任务消耗更多积分，只要它能节省时间、提升质量，或者帮你做出重要决策，仍然是值得的。

A task that uses more credits can still be worthwhile if it saves time, improves quality, or helps you make an important decision.

他们的很多建议，其实还是围绕着Eric之前提到的‘坚持’这个特质展开的。

And a lot of their recommendations are again about managing that tenacity that Eric was talking about.

他们建议先只生成一个结果供你审阅，如果发现任务开始做你不需要的事情，就及时缩小范围或直接中止。

They suggest starting with only one result that you can review, and doing things like narrowing or stopping the task if it starts doing work you no longer need.

当然，这里还有很多内容，但我特别想提一下最后一点——如果你经常接触这类工具，一定会经常听到这个建议：试试语音输入。

Now there is, of course, a lot more in here, but the last one that I wanted to note specifically was the suggestion, which you'll hear a lot if you stick around these parts, to use voice dictation.

ChatGPT 的一个优点是，即使你没在电脑上设置像 WhisperFlow 这样的工具，它的语音输入功能也是目前最好的，你完全可以直接用应用内自带的语音输入功能。

One thing that's great about ChatGPT is that even if you don't have something like WhisperFlow set up on your computer, ChatGPT's dictation is absolutely best in class, and so you can use the native dictation tool that's embedded right within the app.

我之前说过，语音输入在 AI 环境下这么有用，关键就在于‘上下文’。

As I've said before, one of the reasons that voice dictation is so valuable in the context of AI is, well, context.

在 AI 越来越依赖更多信息才能做好工作的时代，说话啰嗦的人反而占优势——哪怕你对着电脑唠叨好几分钟，想到哪说到哪，也比那些写得精准但漏掉关键背景的 typed 笔记更管用。

In a world where AI needs more information to do its job well, the rambler shall inherit the earth, and talking at your computer for a bunch of minutes, even if it's a totally unstructured stream of consciousness, is often going to be more effective in giving the AI what it needs to do well than a hyper precise and clearly articulated set of typed notes that don't necessarily include all of that context.

当然，这并不是唯一一份在 5.6 版本前后发布的模型使用指南。

Now this is not the only new model guide that came out around 5.6.

在 OpenAI 的开发者网站上，他们也有一套针对 GPT-5 0.6 的最佳实践，有意思的是，你还能点回去查看以前版本的指南，甚至能追溯到 GPT-4.1。

Over on the OpenAI developer site, they also have a set of best practices for GPT-five 0.6, and what's cool is you can actually click back and see best practices for previous models as well, going all the way back to GPT-4.1.

AI 内容创作者阿里·莱曼觉得这份文档特别有用，还总结了他发现的一些实用技巧。

AI content creator Ali Lehman found the document really valuable, and summed up some of the best tricks he found.

第一，他建议：删掉旧提示里重复的指令。

One, he writes, delete instructions from your old prompts.

OpenAI 的原则是：每条指令只说一次。

OpenAI's rule is to state each instruction exactly once.

他们发现，删除重复的指令后，得分提升了10%到15%，同时减少了多达66%的令牌消耗。

They found that removing repeated instructions raised scores by 10 to 15%, while cutting tokens by up to 66%.

你以前为旧模型写的那些冗长指令，现在反而让GPT-5.6的回答变得更差，还让你花更多钱。

The giant rule lists you wrote for older models now make GPT-5.6's answers worst, and cost you more.

第二，根据任务需求匹配计算资源。

Two match the compute to the job.

现在有两个独立的调节旋钮：一个是模型规模，另一个是思考强度。

There are two separate dials now, which model size and how hard it thinks.

一共有六个努力等级，从最低到最高。

Six effort levels from none to max.

OpenAI建议，思考强度的设置可以从你上一个模型用的级别开始，然后尝试调低一级。

OpenAI's advice on the thinking dial, start at whatever setting you used on the last model, then test one level lower.

新一代模型通常需要的思考量更少。

The new generation usually needs less.

把最高强度留给你真正最棘手的问题。

Save max for your genuinely hardest problems.

这条建议听起来简单得不得了，但真要做起来，心里却特别别扭。

This is one of those pieces of advice that sounds incredibly simple, but is almost emotionally hard to do.

对我们很多人来说，遇到任何问题，第一反应就是把所有设置都调到最高，毕竟在同等条件下，谁不希望每个问题都能用最聪明的方式解决呢？

For so many of us, the temptation for just about every problem is to dial up the settings all the way to max, because all things held equal, wouldn't you want the most intelligence on every single problem?

就算不考虑成本，现在也越来越清楚，这种做法其实并不一定最有效，所以OpenAI才想给出一些明确的指引，告诉你该怎么做才对。

Even outside of costs, it is increasingly clear that that is not necessarily the optimal behavior, and so that's why OpenAI is trying to give some discrete guidance around what you should do instead.

另一个关于撤销旧指令的建议，是回头检查你所有关于简洁性的要求。

Another tip in the area of undoing previous instructions is to check all the rules you gave about brevity.

Ollie说，GPT-5.6默认的回答比5.5更短，所以你以前为旧模型加的那些简洁规则，现在反而可能删得太多了。

Ollie writes GPT-5.6 defaults to shorter answers than 5.5, so brevity rules you added for older models can now cut too much.

当你真想要简短回答时，别只说‘简短点’，而是明确告诉它哪些信息必须保留，哪些细节可以删掉。

When you do want something short, tell it which information to keep and which detail it can drop instead of a blanket, keep it brief.

在语气方面，具体的指令比抽象的指令更管用。

On tone, concrete instructions beat abstract instructions.

比如OpenAI建议，像‘友好’‘有同理心’这种词太模糊了，Ollie说，不如直接说明你希望它怎么写。

For example, OpenAI suggests that terms like friendly and empathetic are going to be too abstract, and instead Ollie writes, Spell out the actual writing behavior you want.

怎么写才到位，开头该说什么，哪些内容可以跳过。

How to write to be, what to open with, what to skip.

比如，第一句话就点出客户的问题，用编号列出解决方案，直接跳过道歉那段，这样每次都能保持同样的语气。

Something like, Name the customer's problem in your first line, give the fix as numbered steps, skip the apology paragraph, gets you the same tone every single time.

不过，尽管这些建议都很清晰实用，随时都能上手，我最近在很多讨论中还注意到另一个趋势，那就是我们对这些顶尖模型的期望值到底该有多高。

And yet, if these tips are all very clear and practical and stuff that you can go use right away, there is another theme that I'm starting to see across a lot of the discourse, which is about the level of ambition we're bringing to these most advanced models.

Intuit的AI用户体验产品经理Christine Xu写了一篇文章，标题是《你对Claude的期待还不够高》。

Christine Xu, an AI UX PM at Intuit, wrote a post called You're Not Ambitious Enough with Claude.

Christine提到，我日常工作里效率和能力最大的突破，是在我不再只让Claude处理琐事，而是开始让它承担更有价值的任务之后。

Christine writes, The biggest productivity and capacity unlock in my daily work happened when I went beyond automating busy work to asking Claude to do more high leverage work.

几个月来，我一直用Claude来清理那些让我忍不住立刻回复的小任务。

For months, she wrote, I was using Claude to clear the dopamine backlog.

这些任务堆成了一堆，我总急着去完成，因为每做完一件，都会带来一点即时的满足感，误以为自己在进步。

The queue of little tasks that I respond to too eagerly because completing gave a dopamine hit an illusion of progress.

这些自动化帮我腾出了大量时间。

These automations freed up a lot of my time.

有了这段时间，加上Fable，我开始让AI处理那些以前我不敢交给Claude的高价值难题。

With that newfound time and Fable, I started pushing it for higher leverage work hard tasks I didn't trust Claude with before.

这让我在生产力、承载力和心流体验上都实现了最大的突破。

This made the biggest leap in my productivity, capacity, and sense of flow.

克里斯汀引用了Stripe产品负责人谢雷什·多西的一个概念，他说产品工作分为三个层次：第一层。

Christine calls upon a concept from Stripe product leader, Shreyash Doshi, who wrote there are three levels to product work: one.

执行层，第二层。

The Execution Level two.

影响层，第三层。

The Impact Level and three.

形象层。

The Optics Level.

克里斯汀指出，我们很多人用Claude来自动化形象层和执行层的琐事，却把真正有影响力的工作留给自己。

Christine notes that, quote, many of us use Claude to automate busy work in optics and execution and leave the impact work to ourselves.

我们给自己编了个故事，说判断力才是人类不可替代的价值。

We tell ourselves the story that judgment is the defensible human thing.

但只要给 Claude 提供合适的背景信息，它就能比你做得更好、更快。

But when given the right context, Claude can do the hard work better and faster than you.

降低开始行动的门槛，本身就是一个巨大的突破。

Lowering the activation energy of starting is a major unlock in itself.

事实上，Christine 建议用 Fable 以不同方式处理这三类工作。

In fact, Christine suggests using Fable five in different ways for each of these three categories of work.

比如，她说，视觉展示类工作就是让团队的进展被正确的人在合适的场合看到。

Optics work, for example, she says, is about making your team's progress visible to the right people in the right forums.

对于这类工作，她建议用 Claude 配合 Fable 当成自动引擎，取代人工输出。

For this, she suggests using Claude with Fable as autopilot to replace production.

她举了个例子：我有个在 Cowork 上安排好的任务，每周运行两次。

The example she gives: I have a scheduled task on Cowork that runs twice a week.

它会从关键来源提取信息，把每个工作流的最新进展填进一个 Google 表格模板里。

It pulls context from key sources and writes the latest updates for each workstream into a Google Sheet template.

这个表格状态看板对团队开放，大家有疑问可以直接查，不用再来问我。

This G Sheet status board is visible for the teams to get their questions answered without asking me.

我还设定了一个技能，能把这些更新按不同层级打包，发给对应的人和平台。

I also have a skill that packages these updates at different levels for the right audience and forums.

光学工作这一块，还得继续自动化到底，因为它能清理掉那些让人分心的待办，让你保持专注。

Optics worksheet continues, is the layer to automate ruthlessly because it takes care of the dopamine backlog so you can stay in flow.

事实上，她觉得，这些流程很可能很快就会被直接集成到Quad里，变成标准功能。

In fact, she suggests, I wouldn't be surprised if these flows are productized soon in Quad itself.

再往上一层，针对执行类工作，她建议别把Quad当自动工具，而当个助手用。

At the next level, for execution work, instead of Quad as Automator, she suggests Quad as Copilot.

每个周一，克里斯汀都会运行一个上下文整理技能，它会扫描我上周在Slack、日历笔记和代码库里的所有活动，分析我的效率和行为模式，告诉我该优先处理什么，还会推荐具体任务帮我启动工作。

Every Monday, Christine continues, I run the context dump skill, which reads my past week across Slack, calendar notes, and repo activity, gives me an analysis of my efficiency and patterns to watch, recommends what I should prioritize, and offers specific work to help me get started on.

这不像个简单的总结，倒像是个懂我的教练，能发现我的习惯，给出真正有用的建议。

It reads less like a summary and more like a coach who notices my personal patterns, and actually makes helpful observations and recommendations.

我还会运行一个规划技能，检查Jira的状态和产品路线图，看看团队进展如何，接下来该安排什么。

I also run a planning skill that checks Jira status and the roadmap and shows me how the team is tracking and what to line up next.

为了更贴近客户，我每月都会运行一个定时任务，分析客户支持渠道的数据，提炼出最热门的议题、趋势，以及最值得优先处理的反馈和问题。

For staying close to the customer, I run a scheduled task monthly that analyzes our customer support channel to give me the top themes, the trends and recommends, the top requests or problems I can prioritize.

总的来说，她说，要大胆提出更高的要求。

Overall, she says, get more ambitious with the asks.

别把Claude只当成用来提炼总结、供你评判的工具。

Don't treat Claude as just for distilling a summary ask for judgment.

最后，对于有影响力的工作，她建议把Claude当作对练伙伴，帮助你迈出最难的一步。

Lastly, for impact work, she suggests using Claude as sparring partner to get the hardest work started.

这个层次，她说，是杠杆效应最高但也最难启动的工作，比如。

This level, she says, is the high leverage work that is the hardest to start, e.

思考一个全新的产品方向，为一场关键演讲准备叙事，把策略跟业务单元的战略对齐，等等。

G, thinking through a new product bet, preparing narrative for a high stakes presentation, testing strategy against business unit strategies, etcetera.

围绕这一点，她建议，不妨正式‘接入’Claude，并为它配置好合适的背景信息。

Now around this, she suggests it's worthwhile to quote, unquote onboard Claude and set it up with the right context.

这一点我们之前聊过很多次了：建立个人背景资料库，包含AI模型真正成为有用合作者所需的所有信息。

This is something we've talked about a lot, creating a personal context portfolio that includes all the information that any given AI or model needs to be able to actually be a useful collaborator.

总的来说，我觉得Christine第一次真正感受到，Fable帮她打开了这种高影响力工作的可能性。

And overall, the biggest thing is that it feels to me like Christine is really feeling like Fable unlocks this sort of impact work for her for the first time.

她提到，Fable 在做有影响力的工作时，明显比之前的模型更出色。

She writes, Fable is noticeably better than previous models for impact work.

我特别喜欢它那种沉稳的语气。

I especially enjoy its calmness.

它的表达更简洁明了。

It sounds more concise.

感觉更像是在和一位聪明的同事聊天，而不是面对一堆冗长又啰嗦的文字。

It feels more like a conversation with a smart colleague than reading walls of text with unnecessary dispositions.

这在保持思路连贯上起了关键作用。

This makes a big difference in keeping a train of thought going.

现在最重要的AI问题，不是谁在用AI，而是谁用得好。

One of the most important AI questions right now isn't who's using AI, it's who's using it well.

普华永道和德克萨斯大学奥斯汀分校刚分析了140万次真实职场中的AI互动，发现了一个令人意外的结果：最有影响力的人并不是prompt写得最好的，而是把AI当成思维伙伴来用。

KPMG and the University of Texas at Austin just analyzed 1,400,000 real workplace AI interactions and found something surprising: the highest impact users aren't better prompt engineers they treat AI like a reasoning partner.

他们会提出问题、引导思考、反复打磨，并不断推动得出更好的答案。

They frame problems, guide thinking, iterate, and push for better answers.

那好消息是什么？

And the good news?

这些行为是可以大规模培养的。

These behaviors are teachable at scale.

如果你正想从单纯使用AI转向真正掌握AI能力，KPMG关于高级AI协作的研究非常值得花时间了解。

If you're trying to move from AI access to real capability, KPMG's research on sophisticated AI collaboration is worth your time.

更多详情请访问 kpmg.com/us/sophisticated。

Learn more at kpmg.com/us/sophisticated.

就是 kpmg.com/us/sophisticated。

That's kpmg.com/us/sophisticated.

2022年AI代码生成技术兴起时，英伟达的首席发明家、哈佛工程师西德·帕雷希提出了一个相反的观点。

With the emergence of AI code generation in 2022, NVIDIA master inventor and Harvard engineer Sid Pareshi took a contrarian stance.

真正决定企业级AI驱动软件开发质量的关键，不是预训练，而是推理时的算力和智能体的协调。

Inference time compute and agent orchestration, not pre training, would be the key to unlocking high quality AI driven software development in the enterprise.

他认为，真正的突破不在于AI生成代码有多快，而在于它能否深入推理，构建出企业级的应用。

He believed the real breakthrough wasn't in how fast AI could generate code, but in how deeply it could reason to build enterprise grade applications.

当全世界都在关注协作者时，他却设计了一种完全不同的东西：Blitzy——首个专为企业级代码库打造、依靠数千个智能代理的自主软件开发平台。

While the rest of the world focused on copilots, he architected something fundamentally different: Blitzy, the first autonomous software development platform leveraging thousands of agents that is purpose built for enterprise scale codebases.

财富500强企业正借助Blitzy，将工程效率提升五倍，原本需要数月的工作，现在几天就能完成。

Fortune 500 leaders are unlocking 5x engineering velocity and delivering months of engineering work in a matter of days with Blitzy.

重新定义你的软件开发方式。

Transform the way you develop software.

了解更多，请访问 blitzy.com。

Discover how at blitzy.com.

网址是 blitzy.com。

That's blitzy.com.

本集由Retool赞助播出。

This episode is brought to you by Retool.

如今，借助AI，一个可用的应用程序只需大约五分钟就能生成。

Generating a working app now takes about five minutes thanks to AI.

但要把它安全地部署到生产环境，加上身份验证、权限管理、审计日志和安全审查，这部分仍然需要不少时间。

Getting it safely into production with auth, permissions, audit logs, security reviews that part still takes time.

这就是 Retool 解决的痛点。

That's the gap Retool closes.

在 Retool 里，你可以用 Claude Code、Codecs 或任何编码代理自由开发，也可以直接导入你已有的 React 代码。

Build apps however you want natively in Retool with Claude Code, Codecs, or any coding agent, or by importing React you've already built.

所有内容都会自动部署到 Retool，并无缝接入企业管控机制。

It all deploys into Retool and picks up governance automatically.

安全由平台本身保障，而不是依赖 AI 生成的代码。

Security lives in the platform, not in whatever the AI wrote.

这样一来，你的团队能以 AI 的速度交付成果，同时避免了影子 IT 和拖慢进度的冗长审核。

So your team ships at AI speed without the shadow IT and the endless reviews that stall out most vibecoded projects.

这也是亚马逊、Stripe 和 Brex 的团队选择 Retool 的原因。

It's why teams at Amazon, Stripe, and Brex build on Retool.

在九月三十日前签约的新企业客户，每年可获得最高一万美元的 AI 信用额度。

And new enterprise customers who sign by September 30 get up to $10,000 in AI credits per year.

立即前往 retool.com/aidaily 开始构建吧。

Start building at retool.com/aidaily.

本集《AI每日简报》由Hyperagent赞助播出，您可以在这里管理团队共同使用的智能代理群组。

This episode of the AI Daily Brief is brought to you by Hyperagent, where you run fleets of agents your team can manage together.

新用户可获得1000美元的推理算力额度。

New users get $1,000 in inference.

别再依赖本地代理或等着在电脑上手动触发的聊天流程了。

Forget local agents and chat workflows waiting on your laptop to be prompted.

Hyperagent将智能代理部署在云端，7x24小时运行，自动使用您团队已有的工具完成实际工作。

Hyperagent deploys always on agents in the cloud, doing real work across the tools your team already uses.

市场团队的代理能将竞争对手的动态自动生成落地页。

Marketing's agent turns competitor moves into landing pages.

销售团队的代理会完善潜在客户信息、起草邮件，并同步更新客户管理系统。

Sales's agent enriches leads, drafts emails, and updates the CRM.

运营代理负责跟进文书工作，追踪预算使用情况。

Ops agent chases the paperwork and tracks the budget.

每个代理都能访问共享上下文，并严格遵守您设定的权限范围和审批流程。

Every agent has access to shared context and follows your rules about scope and approvals.

是时候让你的助手像真正的同事一样工作了。

It's time you add agents that feel like teammates.

立即前往 Hyperagent，由 Airtable 团队打造，为你配备专属助手。

Hire yours at Hyperagent built by the team at Airtable.

前往 hyperagent.com/aidailybrief 领取你的 1000 美元推理额度。

Claim your $1,000 in inference at hyperagent.com/aidailybrief.

关于推动 Fable 更深入地参与高价值、高阶任务，Claude Code 团队的 Tariq 最近写了一篇题为《Fable 寻找未知领域的实用指南》的文章。

Now along the idea of moving Fable into more high impact and higher order types of work, Tariq from the Claude Code team recently wrote a post called A Field Guide to Fable Finding Your Unknowns.

Tariq 写道：和 Claude Fable 五合作，一次次让我重拾一个老道理。

Tariq writes: Working with Claude Fable five keeps reteaching me an old lesson.

地图不是疆域。

The map is not the territory.

地图，也就是对要完成工作的描述，就是我的提示、技能和上下文。

The map, a representation of the work to be done, is my prompts and skills and context.

这些就是我提供给 Claude 的内容。

It's what I give Claude.

真实的工作环境才是实际要落地的地方。

The territory is where the work needs to happen.

代码库、现实世界，以及那些真实的限制条件。

The codebase, the real world, its actual constraints.

地图和现实之间的差距，我称之为未知点。

The difference between the map and the territory is what I call unknowns.

当Claude遇到未知情况时，它必须根据对我的意图的最佳推测来做决定。

When Claude runs into an unknown, it needs to make a decision based on its best guess of what I want.

工作量越大，Claude碰到的未知点就越多。

The more work being done, the more unknowns Claude might run into.

Fable是我第一个发现，工作质量受限于我能否清晰界定这些未知点的模型。

Fable is the first model where I find the quality of the work is bottlenecked by my ability to clarify its unknowns.

Tariq文章的主旨是说，光靠提前规划远远不够，真正与Fable合作，是一个在实施前、实施中和实施后不断发现未知点的迭代过程。

And the rest of the thrust of Tariq's post is to point out that, in his words, planning ahead isn't enough and instead, that, in his words, working with Fable is an iterative process of discovering my unknowns before, during, and after implementation.

那具体有哪些例子呢？

So, what are examples of this?

塔里克说，当他开始使用Fable时，会从四个方面来梳理问题。

Tariq says that when he starts engaging with Fable, he breaks it down in four ways.

这四个类别是：已知的已知，也就是他提示词里明确写出来的东西；已知的未知，是他还没想清楚、但知道自己还没想清楚的部分。

The four categories are: known knowns or essentially what is in his prompt The known unknowns are what he hasn't figured out yet but is aware that he hasn't.

未知的已知，是那些太明显了，他根本不会写下来，但一看见就会立刻认出来的内容。

The unknown knowns are what are so obvious he'd never write it down but would recognize it if he saw it.

而未知的未知，则是他根本没意识到可能存在的情况。

And the unknown unknowns are what hasn't he considered at all.

塔里克认为，识别并管理这些未知，才是智能编码的核心能力，而且他觉得还有提升空间。

Tariq argues that reducing and planning for unknowns is the skill of agentic coding, and suggests that there are ways to improve upon it.

而关键在于，要帮Claude更好地理解你。

And key to that is helping Claude help you.

塔里克写道：指导Claude需要把握一个微妙的平衡。

Tariq writes: Instructing Claude is a delicate balance.

如果你说得太具体，Claude就会死板地照做，哪怕换个方向可能更合适。

If you're too specific, Claude will follow your instructions even when a pivot may be more appropriate.

如果你说得太模糊，Claude 很可能会根据行业惯例做判断和假设，但这些未必适合你的具体任务。

If you're too vague, Claude will often make choices and assumptions based on industry best practices that may not be a fit for your task.

更重要的是，如果你忽略了那些未知的因素，就会两头落空。

Importantly, when you don't account for your unknowns, you fail both ways.

你不知道前方是布满障碍，还是畅通无阻，但你还是希望 Claude 能主动调整方向。

You don't know when the path will be filled with obstacles, and you don't know when the path will be clear, but you still want Claude to veer.

Claude 能帮你更快地发现那些你没意识到的盲点。

Claude can help you discover your unknowns faster.

他说，最关键的是要让 Claude 了解你的起点。

The most important part, he says, is to give Claude context about your starting point.

比如，告诉它你现在的思考到哪一步了，坦诚你对这个问题的经验，把它当成一个思维伙伴来合作。

For example, tell it where you are in your thought process, disclose your experience with the problem, and let it work with you like a thought partner.

有一些具体的做法，比如‘盲点排查’。

A couple specific ideas are things like a blind spot pass.

他举了个例子：我不懂调色是怎么回事，但我得把这个视频调好。

An example he gives: I don't know what color grading is, but I need to grade this video.

你能教我怎么理解自己对色彩分级一无所知的地方，好让我能更好地提问吗？

Can you teach me to understand my unknown unknowns about color grading so that I can prompt better?

另一个建议是先头脑风暴，再做原型。

Another suggestion is to brainstorm and prototype.

塔里克说：当我处在很多‘已知的未知’领域时——也就是那些只有看到具体例子我才懂怎么定义的标准——我喜欢让克劳德和我一起头脑风暴、做原型。

Tariq writes: When I'm working in an area with a lot of unknown knowns, involving criteria I only know to define when I see it, I like to ask Claude to brainstorm and prototype with me.

他特别强调，在原型阶段尽早发现并说出这些‘已知的未知’非常关键，因为等到实际开发时再发现，代价会高得多。

It's extremely valuable he continues to identify and verbalize unknown knowns early during prototyping because finding them out during implementation can be relatively expensive.

功能或需求一旦改动，代码实现可能完全变样，而且你的助手要回退之前的改动也会更困难。

Changes in a feature or spec can cause drastically different implementations in code and can be more difficult for your agent to revert previous changes.

其实，我觉得很多人已经在不自觉地这么做了。

Now this is something I think a lot of you guys are probably doing intuitively.

他举了个例子：我想为这些数据做个仪表盘，但我完全不懂视觉设计，也不知道哪些想法是可行的。

An example prompt he gives is: I want a dashboard for this data but I have no visual taste and I don't know what's possible.

给我做个HTML页面，展示四种截然不同的设计方向，让我看看哪个更对胃口。

Make me an HTML page with four wildly different design directions so I can react to them.

现在，Tariq又提出了一系列处理未知情况的其他方法，核心其实是一种元流程——那就是把Fable当作共同创作者，让它真正发挥出最大价值。

Now, Tariq gives a bunch of other ideas about how to work with unknowns, and a lot of it ultimately comes to a meta process where you invite Fable in as a cocreator of the process to get the most out of what it can do.

Daniel Meisler则认为，不止是像Fable-5这样的单一模型，每当有新的顶尖模型发布时，都有一套固定的提示词值得重新跑一遍，帮你重新调整使用方式，充分挖掘新模型的潜力。

Now, Daniel Meisler thinks that even beyond a single model like Fable five, that there are a set of prompts to rerun every time a new state of the art model drops that can help you reorient your experience to take most advantage of that new model.

他把这些称为‘战术性元提示’，每次AI能力有重大跃升时都值得重新执行。

He calls these tactical meta prompts worth rerunning every time there's a significant jump in intelligence.

第一类是关于优化你的AI工作台，也就是如何更好地利用最新版本的模型。

The first category is around harness optimization and are all related to improving your AI harness, I.

E，指的是围绕模型的一整套资源和信息，帮助你最大化发挥新版本的能力。

E, the set of resources and information around your model to get the most out of the newest version.

他举的一个例子是‘自我模型审计’。

One example he provides is the self model audit.

仔细阅读你的工作台对你身份、目标、语气和偏好的所有设定，找出哪些地方还在基于过时的、理想化的，或者根本错误的版本来模拟你。

Read everything my harness believes about me identity, goals, voice, and preferences and find where it's modeling a version of me that's stale, aspirational, or just wrong.

把你的文件里写的‘你是谁’，和你最近的实际行为、工作成果做个对比，标出所有系统还在按你过去说的那样来优化，而不是按你现在真实样子来运作的地方，并提出具体的修改建议，缩小这个差距。

Compare what my files say I am against what my recent behavior and work actually reveal flag every place the system is optimizing for who I said I was instead of who I am now and propose the specific edits that close the gap.

在某种程度上，这其实就是清理上下文信息。

Now to some extent, this is just context hygiene.

丹尼尔在这里做的，其实是借着新模型发布的机会，做一次彻底的上下文清理，确保你为模型准备的所有资料，比如那个 agents.md 文件，都能真实反映你本人以及你打算交给这个新模型的项目。

And what Daniel is really doing here is using the moment of a new model release to engage in that sort of context hygiene and make sure all of the information that you've surrounded your model with, things like your agents dot m d file, are actually reflective of the person and projects that you're bringing to that new model.

丹尼尔还特别推崇用“搞大项目”的方式来测试新模型。

Daniel's also a fan of testing new models by going really, really big.

他专门整理了一类提示词，叫作“整体生活与工作优化”。

He has a whole category of prompts that he calls overall life and work optimization, I.

第三类。

E.

大范围规划提示，主要帮你理清优先级、项目安排，并制定一个计划，以便在接下来的关键几年里实现收益最大化。

Massive scope prompts that basically help you sort out your priorities, your projects, and come up with a plan for maximizing the next crucial few years.

他提到的一个例子叫“宏观视角”，提示语类似这样：把我所有在线的项目、在智能体里进行的活动，以及通过网络搜索了解到的所有信息都梳理一遍，然后深入分析。

One he calls big picture is a prompt like Take a look at all my various projects in writing online, all the activities we've been doing in the harness, and everything you know about from web search and analyze it deeply.

接着再看看我所在的领域、人工智能行业、社会现状，以及未来的整体趋势。

Then look at what's going on in my field, in AI, and in society, and in the future in general.

你能不能告诉我，什么样的方案能帮我解决日本那边说的 VikiGuy 这个问题？

And tell me what solves the Japanese concept of VikiGuy for me.

我该专注做什么事情，既能让我有成就感，又能赚到钱？

What should I be working on that gives me fulfillment but is also lucrative?

是自己单干，找人合作，还是去公司上班？

Doing it for myself, with collaborators, or working for a corporation?

如果有具体的建议，尽管提出来；如果你需要更多背景信息，也可以先问我一些问题，再给出答案。

Make concrete recommendations if you have them and feel free to interview me for more context before creating the output.

我觉得这种提问方式未必适合所有人，但不管怎样，用这种问题来测试模型的反应，哪怕只是为了摸摸它的风格底细，都是挺好的方法。

Now, I don't think that this is necessarily going to be everyone's cup of tea, but I do think that this sort of prompt is a great way, even if it's just for the sake of taking the vibe temperature of a model, to understand how it engages with these sorts of questions.

丹尼尔还有个建议，最近越来越多人提到，就是重新回到‘循环’这个思路。

Another suggestion from Daniel, which is starting to come up more and more frequently, is to once again think in terms of loops.

马特·舒默在他的 Fable 5 贴士里也提过类似的观点，他叫它‘循环到达标为止’，特别适合创意类任务。

Matt Schumer actually included this in his Fable 5 tips as well, with a recommendation he called loop it until it hits the bar, especially for creative tasks.

不过要做到这一点，你得先回过头去看马特之前说的：给 Fable 明确一个‘完成’的标准。

Now for this, you have to go back to Matt's previous recommendation to give Fable a real bar for done.

马特说，如果你让Fable做高质量的东西，它会停在自己认为‘足够好’的水平，而这个水平通常比你的要求低。

Matt writes, if you tell Fable to make something high quality, it stops at its own idea of good enough, which is usually lower than yours.

所以我从不使用形容词。

So I don't use adjectives.

我会给它一个可以自我评估的明确标准，而且我会把这个标准定得很高。

I give it a bar it can check itself against, and I make that bar hard.

有时候我会自己写出测试要求。

Sometimes I write the test myself.

比如，让一个陌生人根本看不出我们的渲染图和真实照片的区别。

Something concrete, like a stranger can't tell our render from the real photo.

另一些时候，我自己都不知道该怎么衡量想要的东西，这时候我就把这个问题交给Fable二号去解决。

Other times, I don't even know how to measure the thing I want, so I hand that problem to Fable two.

马特接着说，一旦设定了标准，我就让Fable不断循环地去对齐这个标准，然后放手让它去做。

Now Matt continues, once there's a bar, I put Fable on a loop against it and let it go.

它会不断构建、自我检查、找出最大的差距、弥补它，然后再来一遍。

It builds, checks itself, finds the biggest gap, closes it, and goes again.

马特说，他经常用斜杠循环命令，尤其是在创意工作上，因为总有一些具体的标准可以不断衡量，直到真正达到目标。

Matt says that he uses the slash loop command for this constantly, especially on creative work, where there's always something concrete to keep measuring against until it's actually there.

重要的是，马特提到，循环的核心意义在于，Fable 永远不能自己决定什么时候算完成。

Importantly, Matt writes, the whole point of the loop is that Fable never gets to decide finished.

总还会有下一个需要弥补的缺口。

There's always a next gap.

它只有在我说停，或者真的再也找不到任何可以改进的地方时才会停止——而如果你设置得当，这种情况其实很少见。

It stops when I say it's done or when it genuinely can't find anything left to fix, which is rare if you've set this up right.

将来我们可能会专门做一期节目，深入聊聊这些循环策略，不过我想先分享一下最近看到的一篇 Claude Devs 文章里的几点心得，他们把循环分成了几类，我觉得对理解这个概念的核心很有帮助。

Now, at some point, we'll do a show entirely about some of these loop strategies, but I did want to share just a couple notes from a Claude Devs post recently about getting started with loops where they break loops apart into a set of different categories that I think are useful in helping understand the concept at core.

第一种是回合制循环。

Turn based loops are the first type.

这种循环由用户指令触发，当 Claude 认为任务已完成，或者需要更多上下文时，就会暂停。

They are triggered by a user prompt, and they stop when Claude judges that it has completed the task or needs additional context.

回合制循环最适合那些不常重复、时间较短的任务。

Turn based loops are best for shorter tasks that are not part of a regular processor schedule.

你每次发的指令，都会启动一个由你主导每一步的手动循环。

Every prompt you send they write starts a manual loop with you directing each turn.

Claude 会先收集信息，执行操作，检查成果，必要时重复，最后给出回应。

Claude gathers context, takes actions, checks its work, repeats if needed, and responds.

比如他们说：让 Claude 帮你创建一个点赞按钮。

For example, they say: Ask Claude to create a Like button.

它会读你的代码，做出修改，运行测试，然后交回一个它认为能用的版本。

It reads your code, makes the edit, runs the tests, and hands back something it believes works.

接着你需要手动检查结果，再发出下一条指令。

You then manually check the work and write the next prompt.

你可以把手动检查的步骤写成一个技能，来提升验证环节的效率。

You can improve the verification step by encoding your manual steps as a Skill.

这样 Claude 就能更完整地自行检查整个工作流程。

Md so Claude can check more of its own work end to end.

这跟目标导向的循环不同，目标导向的循环也是由手动指令触发，但停止条件是目标达成，或者达到最大轮次。

Contrast that with a goal based loop, which is triggered by a manual prompt in real time, where the stop criteria is the goal being achieved or the maximum number of turns being reached.

他们说，这种方式最适合有明确验收标准的任务，尤其在单次操作无法完成时特别有用。

This, they say, is best used for tasks that have verifiable exit criteria, and is useful when a single turn is not enough.

他们指出，代理在能够反复迭代时表现得更好。

Agents do better, they write, when they can iterate.

当你明确定义了成功标准后，他们说，Claude 就不必自己判断什么算‘足够好’而提前结束循环。

And when you define the success criteria, they say, Claude doesn't have to make a determination on what is good enough and end the loop early.

每次 Claude 试图停止时，都会有一个评估模型检查你的条件，如果未达标就让它继续工作，直到达成目标或达到你设定的轮次上限。

Each time Claude tries to stop, an evaluator model checks your condition and sends it back to work until the goal is met or a number of turns you define is reached.

基于时间的循环由特定的时间间隔触发，可以在用户取消或任务完成时停止。

A time based loop is triggered by a specific time interval that can be stopped when the user cancels it or when the work completes.

这种方式非常适合处理重复性工作。

This is going to be useful for things like recurring work.

另一方面，主动循环是由事件或计划触发的，不需要人为实时启动。

Proactive loops, on the other hand, are those that are triggered by an event or schedule, and don't require a human in real time to set them going.

停止条件依然是达成特定目标，系统会自动持续运行，直到你手动关闭它。

The stop criteria is once again when a specific goal is met, with the routine running itself until you turn it off.

接下来几个月，我们将重点探索如何把这种基于循环的交互方式，从编程领域拓展到创意和知识型工作中去。

Now, a lot of what we'll be exploring over the next few months is how to bring this sort of loop based interaction outside of the realm of coding and into more creative and knowledge work.

虽然我们对循环本身的讨论就到此为止了，但之所以在这一集的结尾提到它们，是因为这些建议的核心思想其实有两个层面。

Even though we're not getting deeper than that when it comes to loops, the reason it's worth mentioning them at the end of this particular episode is that the big theme and takeaway of all of these recommendations is two part.

最简单来说，它们都在提醒我们：每当我们迎来一个新模型，尤其是当这个新模型在智能上实现巨大飞跃时，都必须花力气重新测试所有用法，弄清楚过去对旧模型的提示和交互方式，哪些已经无法充分发挥新模型的潜力，甚至反而拖了后腿。

At a simplest level, they are all reminders that every time we get a new model, especially when those new models represent a big jump in intelligence, we need to do the hard work of going through and trying everything, and figuring out where the ways that we prompted and interacted with the old models either no longer get the most out of the new models or actively harm them.

但更重要的是第二点——从长远来看可能更关键：我们要找到那些真正释放这种智能跃升独特能力的新方法。

But secondly, and perhaps in the long run even more importantly, we need to figure out the new techniques which actually unlock the differentiated capabilities of that jump in intelligence.

很多时候，这些新方法不是简单的提示技巧，而是像这些循环一样，全新的交互模式。

And many times those things aren't just going to be prompting tips, but totally new interaction patterns like these loops represent.

如果非要从这些内容里提炼一个核心启示，那就是：把你的野心提得更高一点，几乎要默认最新模型无所不能，只有大胆尝试那些最疯狂、最具挑战性的想法，才能真正摸清它的边界在哪里——毕竟，边界迟早会自己显现出来。

If there is one takeaway from all of these pieces, it is to ratchet up one's ambition, and almost to assume no limits on what the newest model can do so that by trying the biggest, most challenging things you can think of, you can actually figure out what those limits are once they inevitably reveal themselves.

这项工作的第一层，总是从个人开始，但我认为，组织层面也存在类似的逻辑。

The first layer of this work is always going to happen on an individual level, but I think that there is an analogy for organizations here as well.

尤其是在商业或工作场景中，我们很容易不假思索地继续用AI做那些老早就用它做的事。

It is so easy, especially in a business or work context, to default to using AI for the things that we're already using AI for.

我们已经找到价值的地方，平时做的那些工作，往往只是把AI当成更快、更便宜、或者稍微好一点的工具。

The places where we've already found value, the type of work that we already do, and to view AI as just a way to do it faster or cheaper or maybe a little bit better.

但归根结底，还是做同样的事。

But ultimately to do that same work.

真正的突破，在于重新定义我们和工作之间的关系。

The unlock, of course, is a new relationship with work.

甚至发现一些以前根本不可能出现的新类型工作。

And even unlocking new categories of work that weren't possible before.

要搞清楚怎么做这件事，远没有听起来那么简单，但也特别令人兴奋。

Figuring out how to do that is a lot less simple, but it's also really exciting.

希望今天分享的这些建议，能为你应对各种层次的工作提供实用的工具。

Hopefully some of the tips you've heard today give you tools you can use across that full spectrum of work.

希望即便五分钟之后新模型上线，这些建议依然管用。

Hopefully, they're still useful about five minutes from now when we get the next model leap.

好了，今天的AI每日简报就到这里。

For now, that's gonna do it for today's AI Daily Brief.

感谢你们收听或观看，我们下次再见，保重。

Appreciate you listening or watching as always, and until next time, peace.