- Decision Trees: Clear branching logic with ├── and └── notation
- Sequential Steps: Numbered, ordered procedures instead of scattered explanations
- Prerequisites: Explicit dependency checks before proceeding
2. AI Agent Optimizations
- Tool Call Clarity: Exact function names and parameters
- Binary Decisions: Clear yes/no conditions instead of ambiguous language
- Error Handling: Specific failure conditions and next steps
- Verification Steps: "Recheck" instructions after each fix
3. Cognitive Load Reduction
- Reference Tables: Quick lookup for tools and purposes
- Pattern Recognition: Common issue combinations and their solutions
- Critical Reminders: Common AI mistakes section to prevent errors
4. Actionable Language
- Removed verbose explanations mixed with instructions
- Consolidated multiple documents' logic into single workflows
- Used imperative commands: "Check X", "If Y then Z"
- Added immediate verification steps
In other words, just like programming, we’re writing better instructions. In this case, we’re asking it to think out loud more clearly. It’s almost like whiteboard interview prep.
It’s quite amazing because it means programming is fully entering the natural language phase of the timeline.
If you aren’t a solid clear writer, you may not make it in the brave new world.
>If you aren’t a solid clear writer, you may not make it in the brave new world.
Have you not heard of all the AI startups that can turn a 3-word thought into very clearly written prose to be lovingly poured into the waiting mouth of your AI agent?
Great! A diviner has vibe-exposed the arcane magic word knowledge on the steps to ultimate knowledgeplasty! Come let us get together to share more trial-and-error wordsmithery, Together we will someday have ultimate power!
If the model creators themselves arent sharing this magic-word bullshitteryy then why is anyone spending time on this? It is just going to change with every model release
>GPT-5 showed significant improvement only in one benchmark domain - which is Telecom. The other ones have been somehow overlooked during model presentation - therefore we won’t bother about them either.
I work at OpenAI and you can partly blame me for our emphasis on Telecom. While we no doubt highlight the evals that make us look good, let me defend why the emphasis on Telecom isn't unprincipled cherry picking.
Telecom was made after Retail and Airline, and fixes some of their problems. In Retail and Airline, the model is graded against a ground truth reference solution. Grading against a reference solution makes grading easier, but has the downside that valid alternative solutions can receive scores of 0 by the automatic grading. This, along with some user model issues, is partly why Airline and Retail scores stopped climbing with the latest generations of models and are stuck around 60% / 80%. I'd bet you $100 that a superintelligence would probably plateau around here too, as getting 100% requires perfect guessing of which valid solution is written as the reference solution.
In Telecom, the authors (Barres et al.) made the grading less brittle by grading against outcome states, which may be achieved via multiple solutions, rather than by matching against a single specific solution. They also improved the user modeling and some other things too. So Telecom is the much better eval, with a much cleaner signal, which is partly why models can score as high as 97% instead of getting mired at 60%/80% due to brittle grading and other issues.
Even if I had never seen GPT-5's numbers, I like to think I would have said ahead of time that Telecom is much better than Airline/Retail for measuring tool use.
Incidentally, another thing to keep in mind when critically looking at OpenAI and others reporting their scores on these evals is that the evals give no partial credit - so sometimes you can have very good models that do all but one thing perfectly, which results in very poor scores. If you tried generalizing to tasks that don't trigger that quirk, you might get much better performance than the eval scores suggest (or vice versa, if your tasks trigger a quirk not present in the eval).
Appreciated the response! I noticed the same when I ran tau2 myself on gpt-5 and 4.1, where gpt-5 is really good at looking at tool results and interleaving those with thinking, while 4.1/o3 struggles to decide the proper next tool to use even with thinking. To some extent, gpt-5 is too good at figuring out the right tool to use in one go. Amazing progress.
I wish they had published what prompt was given to Claude to improve GPT-5-mini's performance, as well as a before and after comparison of a prompt that underwent this transformation.
Thanks for the feedback, appreciate it!
It makes lot of sense - I'll update the article with links to the actual prompts.
Initially I thought these would be too lengthy for the article and no one would care, but as it seems people are really interested in it. Of course I'd be happy to share the details.
I see that you've added links to a pull request that show the previous and final optimized prompts. However, the OP was asking for the prompt you gave to Claude to assist you in optimizing your prompt. Would you mind sharing that one? (That way nobody has to reverse engineer the instructions from the diff you provided.)
I feel like eventually we’ll get LLMs that will act like compilers do now. So they will take a prompt and turn it into an optimized prompt for a bigger LLM.
The only problem is I feel like having to have Claude rewrite the prompt negates some of the efficiency and latency benefits of using mini. For system prompts obviously this doesn't matter, but for actual continuous user interaction, it feels unworkable.
It definitely makes sense that improving formatting and clarity for these smaller models would really help with performance, but I'm wondering if gpt5-mini is already smart enough to handle that reformatting, and can rewrite the prompt itself, before handing it off to another instance of itself.
Really intresting. What did the original prompt look like? Perhaps the original prompt was not that good? I feel like the changes claude suggested (except a couple maybe) are already pretty well known prompt engineering practices.
Prompt changes affect output substantially (just look up arxiv), the difficult part is find an optimal structure to yield the best results. It is a bit expensive to do a lot of testing on your own, so it all boils down to feels and experience at the moment. Then you mix up tool calls, other agent calls, client functions and this gets terribly hard to evaluate.
I am still puzzled how distance between policies can have an effect on the output. And how a simple retry fixes everything.
This is very much what dspy aims to address. Learning the incantations necessary to prompt well can be replaced by an algorithmic loop and example labelled cases.
Rewriting prompts don't come with no costs. The cost here is that different prompts work for different contexts and is not generalisable. The rewritten prompt here will not work well for other cases like medical or social advice.
I think this rewriting of prompts technique is the reason "reasoning" models perform well - they know exactly how to rewrite the prompts for a context.
FWIW I don't trust these benchmarks fully because a huge bump like this is not expected - I would expect OpenAI to optimise enough to let such gaps open.
Many of the "look at what I did programming LLMs" blog posts on Hacker News have been developed and put out in academic papers and groups. The posts which gain traction here seem to be perennially behind the state of the art.
I wonder if it would be possible to improve even further on the benchmark by simply showing Claude the current hardest problems and asking it to improve the prompt without including any specifics related to the problems
> Removed verbose explanations mixed with instructions
Is Claude rewriting generic instructions once, or is it rewriting the core task statement each time? If so, I'm not sure how you prevent information leakage: Claude might easily be "solving" some of the tasks and inserting subtle hints on the approach. I think this result is very interesting if it holds after rewriting only the generic instructions, even if the performance boost is lower.
Doesn't saying "check -> action" suggest you're taking _away_ the agentic capabilities, and optimizing for the benchmark, meaning it's no longer a good benchmark for agentic capabilities?
That's like being able to see the test before taking it
Can you point me to any resources on DSPy that don't make it look like magic though? It used to be all the hype for a while and then everyone moved on from it.
1. Structure & Flow
2. AI Agent Optimizations 3. Cognitive Load Reduction 4. Actionable LanguageSoon enough Im sure we’ll start to see programming languages that are geared towards interacting with llms
https://en.wikipedia.org/wiki/Lojban
It’s quite amazing because it means programming is fully entering the natural language phase of the timeline.
If you aren’t a solid clear writer, you may not make it in the brave new world.
Have you not heard of all the AI startups that can turn a 3-word thought into very clearly written prose to be lovingly poured into the waiting mouth of your AI agent?
In fact, according to theory, we're writing executable proofs.
We already have people praying to the machine gods, so I guess your future is next week?
If the model creators themselves arent sharing this magic-word bullshitteryy then why is anyone spending time on this? It is just going to change with every model release
I work at OpenAI and you can partly blame me for our emphasis on Telecom. While we no doubt highlight the evals that make us look good, let me defend why the emphasis on Telecom isn't unprincipled cherry picking.
Telecom was made after Retail and Airline, and fixes some of their problems. In Retail and Airline, the model is graded against a ground truth reference solution. Grading against a reference solution makes grading easier, but has the downside that valid alternative solutions can receive scores of 0 by the automatic grading. This, along with some user model issues, is partly why Airline and Retail scores stopped climbing with the latest generations of models and are stuck around 60% / 80%. I'd bet you $100 that a superintelligence would probably plateau around here too, as getting 100% requires perfect guessing of which valid solution is written as the reference solution.
In Telecom, the authors (Barres et al.) made the grading less brittle by grading against outcome states, which may be achieved via multiple solutions, rather than by matching against a single specific solution. They also improved the user modeling and some other things too. So Telecom is the much better eval, with a much cleaner signal, which is partly why models can score as high as 97% instead of getting mired at 60%/80% due to brittle grading and other issues.
Even if I had never seen GPT-5's numbers, I like to think I would have said ahead of time that Telecom is much better than Airline/Retail for measuring tool use.
Incidentally, another thing to keep in mind when critically looking at OpenAI and others reporting their scores on these evals is that the evals give no partial credit - so sometimes you can have very good models that do all but one thing perfectly, which results in very poor scores. If you tried generalizing to tasks that don't trigger that quirk, you might get much better performance than the eval scores suggest (or vice versa, if your tasks trigger a quirk not present in the eval).
Here's the tau2-bench paper if anyone wants to read more: https://arxiv.org/abs/2506.07982
Should be available now, although it might take a while for CDN to propagate.
https://github.com/mieciu/tau2-bench/pull/1/files
It definitely makes sense that improving formatting and clarity for these smaller models would really help with performance, but I'm wondering if gpt5-mini is already smart enough to handle that reformatting, and can rewrite the prompt itself, before handing it off to another instance of itself.
Overall an awesome article!
In this (telecom) benchmark you can review agent policies and manuals here: 1) https://github.com/sierra-research/tau2-bench/blob/main/data... 2) https://github.com/sierra-research/tau2-bench/blob/main/data...
Of course these are just parts of the prompt, you can inspect benchamark code to see how these are rendered to actual LLM calls.
In case someone is not familiar with framework methodology I've wrote a separate article covering that (with some of my thoughts) -> https://quesma.com/blog/tau2-from-llm-benchmark-to-blueprint...
Prompt changes affect output substantially (just look up arxiv), the difficult part is find an optimal structure to yield the best results. It is a bit expensive to do a lot of testing on your own, so it all boils down to feels and experience at the moment. Then you mix up tool calls, other agent calls, client functions and this gets terribly hard to evaluate.
I am still puzzled how distance between policies can have an effect on the output. And how a simple retry fixes everything.
I think this rewriting of prompts technique is the reason "reasoning" models perform well - they know exactly how to rewrite the prompts for a context.
FWIW I don't trust these benchmarks fully because a huge bump like this is not expected - I would expect OpenAI to optimise enough to let such gaps open.
Definitely interesting, thank you!
Is Claude rewriting generic instructions once, or is it rewriting the core task statement each time? If so, I'm not sure how you prevent information leakage: Claude might easily be "solving" some of the tasks and inserting subtle hints on the approach. I think this result is very interesting if it holds after rewriting only the generic instructions, even if the performance boost is lower.
That's like being able to see the test before taking it
Into the trash it goes.