The argument in one line.
Claude Opus 4.5 responds best to short, direct instructions with explicit context and explicit 'don'ts' rather than long exhaustive prompts, and one habit — asking it to explain its reasoning — can silently downgrade your request to a weaker, cheaper model.
Read if. Skip if.
READ IF YOU ARE…
- You're already using Claude (desktop, Claude Code, or the API) and want a concrete checklist for prompting the newest top-tier model well.
- You manage system prompts, CLAUDE.md files, or reusable agent instructions and want to bake in verification and effort-matching habits once instead of repeating them per-prompt.
- You're deciding whether a task is worth the top-tier model's cost versus a cheaper/faster one.
SKIP IF…
- You don't have access to the newest Claude tier and just want general prompting advice — the model-specific handoff and effort-level rules won't apply to you.
- You're looking for prompt engineering for image or video generation models — this is text/agentic-coding focused.
The full version, fast.
Anthropic's newest flagship model (referred to here as 'Fable 5') responds to a different prompting style than older models: shorter, more direct instructions actually outperform long, over-specified ones. The video distills Anthropic's own prompting documentation into six habits: explain the why behind a request instead of just the task, explicitly state what not to do, let the model act once it has enough information instead of over-planning, demand proof of completion rather than trusting a 'done' claim, avoid asking it to explain its reasoning (which can silently trigger a downgrade to a weaker backup model), and default to saying less rather than more. It also covers matching reasoning effort levels to task difficulty and explains the safety mechanism that quietly reroutes certain requests to a cheaper, less capable model without always telling you.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 01:26
01 · Fable 5 is back
Model returns after export-control suspension; pricing ($10/$50 per million tokens), promotional-period usage limits ending July 7, and where it's available.
01:26 – 02:25
02 · Sponsor: HyperAgent
Sponsored segment for HyperAgent — a multi-agent 'council' tool for stress-testing business ideas.
02:25 – 03:33
03 · Why this model prompts differently
The model follows short, clear instructions better than older models because of stronger reasoning; intro to the six-habit framework.
02:54 – 04:04
04 · Rule 1: Give it the why
Explain the intent/context behind a request instead of just stating the task, so the model connects to the right information.
04:04 – 05:05
05 · Rule 2: Tell it what NOT to do
Negative prompting — explicitly stating forbidden actions — now works better than it used to; Anthropic's own docs use this pattern repeatedly.
05:05 – 05:41
06 · Rule 3: Let it act once it has enough
Stop over-planning; let the model act once it has sufficient information instead of demanding exhaustive research and planning first.
05:41 – 06:28
07 · Effort levels
Matching low/medium/high/extra-high effort to task difficulty; high is the recommended default, extra-high reserved for the hardest tasks; cost/accuracy tradeoff chart shown.
06:43 – 07:48
08 · Rule 4: Make it prove it
Don't trust a model's claim that a task is done — require it to point to verifiable evidence before reporting completion.
07:48 – 08:05
09 · Rule 5: Stop asking it to show its reasoning
Model-specific: requesting an explanation of reasoning, especially in a system prompt, can trigger a refusal or silent handoff to Opus 4.8.
08:05 – 09:22
10 · Rule 6: Say less, not more
Short instructions now steer as well as long rule lists when the model already has good context/tools/skills; doesn't contradict Rule 1.
09:22 – 10:44
11 · When Fable hands off to Opus 4.8
Explains the safety-check mechanism: risky-looking requests (hacking, dangerous biology, requests to reveal private reasoning) get silently routed to a cheaper backup model; billed at the backup model's lower rate.
Lines worth screenshotting.
- The newest top-tier Claude model follows short, direct instructions better than older models because it reasons better, so verbose prompts no longer help and can hurt.
- Giving the model the 'why' behind a task lets it connect the request to the right context instead of guessing your intent.
- Negative prompting — explicitly stating what not to do — works better on newer models than it used to, reversing older advice to avoid it.
- Anthropic's own prompting documentation for this model repeatedly uses explicit 'do not do this' framing rather than only positive instructions.
- Over-planning is now a liability: letting the model act as soon as it has enough information beats forcing it through exhaustive research and planning first.
- The model can silently claim a task is done without having verified it, so demanding it point to concrete evidence before reporting completion closes a real trust gap.
- On this model specifically, asking it to explain its reasoning — especially in a system prompt — can trigger a refusal or silently reroute the task to a weaker backup model.
- The routing-to-a-weaker-model safety mechanism exists because the flagship model has stricter jailbreak-related guardrails than its predecessor.
- When a request is flagged as touching hacking, dangerous biology, or asks the model to reveal private reasoning, the system can silently swap in a cheaper, less capable model to answer instead.
- You are billed at the cheaper backup model's rate when a silent handoff occurs, not the flagship model's rate, so the swap isn't a hidden cost trap.
- Effort levels (low, medium, high, extra-high) should be matched to task difficulty; Anthropic recommends high as the default for most tasks and extra-high only for the most capability-sensitive work.
- The flagship model on its lowest effort setting can match a prior-generation top model's accuracy on its highest effort setting, but at a fraction of the cost.
- A realistic estimate is that most users only need the newest flagship model for 5-15% of their total tasks — using it for everything is overkill.
- Saying less, not more, can now steer the model as effectively as a long rule list, because a well-set-up model with the right context, tools, and skills needs less explicit spelling-out.
- The 'say less' habit doesn't contradict 'give it the why' — brevity applies to rules and mechanics, while context about intent should still be included.
Six habits for getting more out of Claude's newest flagship model.
PROMPTING HABITS
Short, direct prompts now outperform long ones on the newest top-tier Claude model, and one habit — asking it to explain its reasoning — can silently reroute your request to a weaker model.
- State the intent behind a task, not just the task itself, so the model can connect it to the right context instead of guessing.
- Explicitly say what the model should NOT do — negative prompting now works better on newer models than it used to.
- Let the model act once it has enough information instead of forcing exhaustive upfront planning; over-planning wastes time and tokens.
- Require proof before trusting a 'done' claim: ask for the specific evidence that shows the work is actually complete.
- Avoid asking the model to explain or narrate its reasoning, especially in a system prompt — it can trigger a refusal or a silent downgrade to a weaker, cheaper model.
- Match reasoning effort level (low/medium/high/extra-high) to task difficulty; default to a mid-to-high setting and reserve the top setting for genuinely hard work.
- Default to concise instructions once the model has good context and tools already in place — brevity in rules doesn't mean skipping the 'why' behind the task.
- Understand that safety-triggered handoffs to a backup model are billed at the backup model's cheaper rate, so a silent downgrade isn't a hidden cost increase.
Terms worth knowing.
- Negative prompting
- Explicitly telling an AI model what NOT to do, rather than only describing the desired outcome — useful because models can otherwise improvise unwanted extras.
- Effort level
- A setting (low, medium, high, extra-high) that controls how much reasoning/compute a model applies to a task, trading cost and speed against capability.
- Model handoff / routing
- A safety mechanism where a flagship AI model quietly reassigns a risky-looking request to a different, usually less capable, backup model instead of answering directly.
- System prompt
- The instructions given to an AI model before the user's own message, typically used to set persistent rules, persona, or constraints for the whole session.
Things they pointed at.
00:00linkAnthropic's prompting-Claude-Opus-4.5 documentation
Lines you could clip.
00:00
“It's the strongest one that I've ever used, hands down.”
06:43
“Before you tell me something is done, point to the result that proves it. Only report work you can show evidence for.”
08:05
“A standing 'explain your reasoning' line, especially in the system prompt, can trigger a refusal, and it can hand your task to Opus 4.8.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
metaphoranalogy
00:00So we've all waited long enough. Fable five is finally coming back to us, and it's an incredible model. It's the strongest one that I've ever used, hands down.
00:07I've built things like my second brain, my AI operating system. So, obviously, I've been playing around with this model a ton, and I've been trying to figure out the best way to use it so that you can actually use it efficiently and you're not paying for tokens for no reason. I've looked at what people have said on x, I've listened to Anthropic Engineers, and also I've read through this entire documentation right here on the best practices for prompting Claude Fable five.
00:26So today what I did is I distilled all of this into the six most simple and effective things that you should be doing right now when you are prompting Claude Fable five to get the most out of it. But it's certainly not a cheap model. It costs double OPIS at $10 per million input tokens and $50 per million output tokens.
00:41And the other thing is it won't always be on your Claude plan, and they're calling this a promotional period. And you can only use up to 50% of your weekly limits with Claude Fable five at no extra cost, but then after that, you would have to switch to usage credits. Unfortunately, the promotional period ends July 7, so we only have, like, six days.
01:00And part of that is fourth of July weekend. Come on. Anyways, it will pretty much work anywhere, though.
01:04Cloud desktop app, Versus code, in Cloud code, you know, wherever you wanna use it. You can see right here, I just got this message that Claude Fable five is back. And if I go to slash model, we can see that we do have access to Fable.
01:15Once again, thank you so much. I'm glad that you're back.
01:19And because this certainly isn't cheap and I wanna be able to test it a ton for you guys, then I hope you'll forgive me for pausing for a quick message from today's sponsor. Real quick, guys. A message from the sponsor of today's video, HyperAgent.
01:29HyperAgent has helped me fix a major problem that AI has. It's . Ask it about your idea, and it tells you you're a genius.
01:36So on HyperAgent, built by the Airtable team, I built a council of agents, a handful of different AI agents each with its own persona, each running on its own cloud machine with a real browser and tools to go do actual research. So I can just simply drop an idea in Slack, and they will tear into it. One of them will play the skeptical investor.
01:54One will dig up what competitors are already doing. One will stress test the numbers. They'll all go pull real data and come back to me with actual, brutally honest feedback, not something like, looks great, Nate.
02:04It'll tell me where stuff is weak and why and how to fix it. So by the time they're done, I've got a sharper idea and the receipts to back it. And that's the part I love because it's not just one chatbot nodding along, but it's a team that disagrees with me on purpose, and I built the whole thing myself in just an afternoon.
02:19So if you wanna build your own council, link's in the description with free credits, so go get roasted. Let's get back to the video. So the first thing to understand about Cloud Fable five is that it is obviously built a little bit different than Opus and, you know, GPT.
02:31So there's different things that it does really good. It follows short, clear direction better than older models do because it's just better at reasoning. And if you spend some time playing with it, it just feels like when you ask it to do something, it just feels like it knows what you're talking about and understands it a little bit better.
02:46Okay. So here are these six habits, six things to do. I've tagged each of these if they are sort of like any model or Fable specific.
02:54So this first one is any model. To give it the why. Give it the context behind why you're doing something.
02:59Anthropic says that Fable does better when it understands your intent. So the context lets it connect your task to the right information instead of guessing what you meant. So instead of saying something like write me an email to a client about the delay, say, I'm working on this bigger task, and here's who it's for.
03:14What they need is blah blah blah. And with that in mind, can you help me write an email to this client about the delay? And if you have your second brain and your whole AIOS set up in the right way, that will also prompt it to look through specific context files to make sure that it has the right information that it needs to make all of these emails or whatever you're doing feel more specific.
03:32The next thing, which I've also found in general works better for all AI models, is to specifically negative prompt, specifically tell it what not to do. Think about it like this. AI is trained on a ton of data, so it literally at its core is trying to predict the next word that makes the most sense.
03:48And sometimes it will try to get creative and it will try to do things that you didn't ask for. And sometimes that's good, but a lot of times it's not. So explain what not to do.
03:57If you look through this page prompting Claude Fable five, when you start to go through here, you will see that they do that. Right here, when you have information to act on act, do not do this.
04:06If you are weighing a choice, give a recommendation, but do not do this. Here's another example.
04:11Don't add features. Don't do this. Don't do this.
04:13Don't do this. Don't do this. Do the simplest thing that works well.
04:17The way I like to think about that is if you were explaining a task to an intern, you would tell them specific things to not do because they don't understand the process yet. They're trying to learn it. So instead of saying something like take a look at this problem and handle it, you might say, when I'm describing a problem or asking a question, the deliverable is your assessment.
04:32Report what you find and stop. Don't fix, send, edit, or delete anything until I say go. I feel like the models have actually evolved a little bit because I used to find that negative prompting didn't work as well than just being very specific about what I wanted it to do.
04:45But lately, the more I've been negative prompting, the more I found that it tends to work pretty well. So if you guys wanna access this full guide where you can see this full write up, you can do so by joining my free school community. The link for that will be down in description.
04:56All you have to do is go in here, go to the classroom, and click on all YouTube resources, you'll find everything in there. Okay. The next one also works for any model.
05:03This is to let it act once it has enough. So stop it over planning. I actually don't even very often use plan mode inside of Cloud Code anymore.
05:11I know that's something that I used to say, always start in plan mode, but I don't do that anymore. I basically have my own plan mode that I have it go through until it's ready to act. So instead of saying research everything and make a full plan before you do anything, you can say more stuff like, when you have enough information to act, then act.
05:27If you guys remember, that is basically the example that we read right here, which is the first thing that Claude Fable five doc says because individual requests on hard tasks can run for many minutes at higher effort settings, especially when the task requires gathering context, building, self verifying, essentially just building a plan.
05:43And another big piece of this here is thinking about the different effort levels. You know, you can do low, medium, high, extra high. You've got all these different effort levels, and you wanna make sure that you're matching the effort levels to the correct task.
05:55They recommend to use high as the default for most tasks. Use x high for the most capability sensitive workloads and medium or low for routine work.
06:03And it's quite interesting because if you look at Fable five versus OPUS 4.8 with different reasoning levels as well as the cost, you can see that they get into this area where they're similar, but Fable five on low, even though it's similar to OPUS 4.8 on x high and max, is cheaper. So really starting to experiment with different effort levels and feeling when you need to use certain ones is going to be a very important skill along with understanding what tasks to use what model for in general.
06:31Because if you're using Fable five for everything, that is almost a 100% overkill. Especially when you're getting into the usage credit territory of Fable five, you really don't need it for that many things. You probably more realistically only need to reach for Fable, like, five to 15% of the time.
06:44Number four, which I think is probably the most important one, is to make it prove it. And this is for any model, like I said, once again. Sometimes models will tell you that they're done, but they're not.
06:53Or maybe they are done, but they haven't verified things. The whole point is if you were to give work to a human and they come back with, you know, something, some output, you would have to review that yourself. You wanna get it to the point where you trust this person or this model so much because you've baked in these verification loops that when it hands you something, you probably still should check it, but you don't feel like you have to as much because you know that it already has checked its own work two, three, four, maybe even five times, and it's tested and fully verified that everything's working the way that it should be.
07:22So instead of just asking, is this done and is it working? You could say, before you tell me something is done, point to the result that proves it. Only report work you can show evidence for.
07:30If something isn't verified, say so plainly instead of guessing. And this is something that would be really good to work into all your skills, all of your agents, all of your, you know, cloud.md files rather than always just throwing this at the end of every prompt. Okay.
07:42Number five, and this one is more Fable specific, which is stop asking it to show its reasoning. On fable five, a standing explain your reasoning line, especially in the system prompt, can trigger a refusal, and it can hand your task to opus 4.8. Now if you guys didn't know, and I'll cover this a little bit more at the end of this video, but because Fable five has concerns of, you know, jailbreaking and it's already a a lesser model of Mythos five, we have to be careful about making sure that we don't trigger Opus.
08:09Because, basically, there's a bunch of safety guardrails into Fable that will route it to a less capable model like Opus if it thinks that the intent of your request might be malicious or anything like that. So for some reason, if you are trying to get Fable to show its reasoning, it might revert you down to Opus 4.8. So that's why we put this one in here that is a Fable specific prompting rule.
08:29And then number six is to say less, not more, which kind of sounds a little bit counterintuitive because typically we've thought the more context that you give these models, the better. But because Fable is so intelligent now and especially if you have it wrapped up in a good environment with, you know, context and tools and skills and everything like that, a short instruction can now steer just as well as spelling out most of the rules by name.
08:51And this is not a contradiction with tip one. Adding the why, if you remember back here, give it the why, give it the context, still doesn't mean to blow everything out. So instead of saying like, hey.
09:00Rule one is be concise. Rule two is do this. Rule three is do this.
09:03Just say more something like, lead with the outcome, keep it simple, and pause only when the work truly needs me. And that's how you can start to wrap all of these other tricks into your system files, your memory files, your Cloud NMD files, and your skills, and all of your prompts moving forward so that you can make sure you're getting the most out of this incredible model, Cloud Fable five.
09:22So I wanted to wrap up here with what I talked about very briefly, which is what happens when Fable hands off to Opus four point eight or whatever model in the future it ends up handing off to if this is still the case. Fable will run a quick safety check before it goes ahead and answers your request. If it realizes that it might be within a certain bucket, then it will push that to Opus four point eight.
09:42If it looks like anything that has to do with hacking or dangerous biology or asking the model to reveal its own private reasoning, then it might shoot that off. And when that happens, it won't show you.
09:52It will silently route to Opus. If you are building on the API, it will send back a response that shows you that was Opus, but otherwise, you may not notice it. Luckily, when you do it, it should be routing to Opus4Point8, so you're not paying as much as Fable.
10:06Otherwise, that would be pretty bad. And so just be aware of that. You can avoid it by obviously doing something like this that I mentioned, but also, you know, not asking it to help you do things that are clearly malicious or suspicious.
10:18So please feel free to go ahead and go to this documentation and read through it. It has a lot of golden nuggets. But that's gonna do it for this one.
10:24So, you guys enjoyed and you learned something new. If you're interested in learning more about those agent loops kind of verification that I was talking about earlier, then definitely check out this video right here. I pretty much explain it.
10:34I give it some examples, and I think that doing this technique with Fable five will be super awesome. So, hopefully, I see you guys over there. Otherwise, thanks for making it to the end of the video, and thanks, everyone.
The Hook
The bait, then the rug-pull.
Anthropic's newest flagship model came back online after an export-control suspension, and it's expensive enough that wasting tokens on it actually hurts. This breakdown distills Anthropic's own prompting documentation, engineer commentary, and X discussion into six concrete habits — including one non-obvious trap where asking the model to show its reasoning can silently get your request rerouted to a weaker, cheaper model.
Named ideas worth stealing.
02:25list
Six habits for prompting Claude Opus 4.5
- Give it the why (any model)
- Tell it what NOT to do (any model)
- Let it act once it has enough (any model)
- Make it prove it (any model)
- Stop asking it to show its reasoning (model-specific)
- Say less, not more (model-specific)
A six-item checklist distilled from Anthropic's official prompting documentation for its newest flagship model, tagged as either universal (works on any model) or specific to this model's behavior.
Steal forA CLAUDE.md / system-prompt template, or an internal prompting-standards doc for any team using Claude for agentic coding or research tasks.
How they asked for the click.
VERBAL ASK
04:14newsletter
“if you guys wanna access this full guide where you can see this full write up, you can do so by joining my free school community”
Soft mid-video CTA tied directly to the content being discussed (the full prompting guide write-up), plus a separate paid-playbook link and sponsor link in the description.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
AFFILIATECommission earned if you click.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
