Yeah basically! I thiught it was a bit ridiculous that they would charge for something so easy to make yourself. For example, I have a series of md files with iterated prompts in a folder and start off a human generated prompt to feed into [insert favorite LLM].
an example seed:
"create a prompt for an agent that will help me reduce prompt token usage and speed up results without losing necessary complexity. can you build a prompt that I use to this end?"
After a bunch of recursive prompting:
"Optimize the provided 'Original Prompt' into an 'Optimized Prompt'.
The 'Optimized Prompt' must:
- Be token-efficient.
- Be maximally clear, precise, unambiguous, with direct instructions.
- Be ideal for advanced AI model processing.
- Preserve the 'Original Prompt's' core intent and task.
- Retain 'Original Prompt's' details, nuances, analytical requirements, output formats, and complexity, without oversimplification.
Apply this optimization method:
1. From 'Original Prompt', eliminate: conversational filler, redundancy, pleasantries, self-references.
2. Use: strong, direct action verbs.
3. Be: specific, direct. Replace vague terms with precise equivalents.
4. Clearly state: task, context, constraints, output format. Explicitly define implied formats (e.g., list, JSON, steps).
5. Logically group related instructions.
6. Ensure 'Optimized Prompt' is a direct command."
It's brutishly simple, but part of the (imho self evident) process is editing the prompts as you continue to feed it back in on itself.
One of the original AI-whoah moments was that italian fashion/harry potter mashup [0]. Some time after it came a "making of" video which outlined how they used ChatGPT to prompt Midjourney to create the images [1]. To me, this established using generative systems to create prompts for other generative as an effective pattern for using such systems.
This is well written so if AI wrote some of it I can't tell, but there is an advantage in not using AI at all to write. It makes you think better, and carefully craft sentences. It is also more respectful to the audience.
Nothing to do with software is real engineering in any case. Engineering outside the software world actually refers to real codified standards (not just "standards" based on vibes or popularity or generic best practices) that are enforced by central professional regulatory bodies. These are fields where your mistakes can land you in prison or debarred from practicing in the field. No such thing exists in software, so until it does then really anyone in this field should give themselves a real critical review over how snotty they want to be with terminology.
Let's say I write a prompt and instruct ChatGPT to analyze the prompt and apply the best guidelines on prompt engineering to it to generate a new prompt that does the job better. Then I take that generated prompt, give it to DeepSeek with the same instructions, and take the output, give it to Claude, and so on through all the released LLMs until we circle back to ChatGPT. This is something like the telephone game, but will the final result be a better or worse prompt than the original?
Garbage in, garbage out - GIGO - still seems to apply to LLMs. It might be nice if LLMs would respond with 'this prompt doesn't compile, try again' while emitting an error report, like compilers do, if some minimal standard wasn't met.
I'm on hot-take fire today, so may as well keep it going. I'm of the opinion this is the new field of programming. Think of it like Game Engine scripting. The engine is made, but all the level design is going to have to be done with scripting. People that know how to design innovative and practical prompt solutions are going to be quite valuable. In fact, Prompt Engineering is all that's left. Don't even try to out-hot-take me on this one.
Even if you're 100% accurate that the advent of LLMs means that the field of software engineering has effectively devolved into prompt engineering and AI-wrangling, that is a change that we should fight with full-throated, actual-Luddite levels of defense. Your own analogy – the way a tiny, tiny core of 'real' engineers develop game engines, and then the entire field of game development just 'scripts' those engines – sends up a ton of red flags for me.
(Aside: as an erstwhile game developer 'just scripting game engines' is...underselling the craft of programming in game development, but whatever).
For a long time game development has been a weird shadow version of the rest of the tech industry. We're influenced by the same macro trends (e.g. ZIRP, VC fads) and the mood and zeitgeist generally rhyme as a result. But if you look at the drive to unionise game developers vs. same impulse in the mainstream tech industry the feeling is COMPLETELY different.
What's the incentive to unionise if you're a SWE at Meta, Palantir, or Google? Your job is pretty great, your work-life balance is at least not fundamentally out of control, and your STARTING salary puts you in the top 10% of US households. It is probably the last remaining holdout of the 1970s upper-middle-class dream jobs.
And if you're the equivalent engineer at EA, Activision, or Ubisoft? You can expect seasonal layoffs, a good work-life balance means you sleep at home instead of under your desk at least once a week, and your take-home pay is just sufficient to let you split the rent on an outer LA apartment that's just inside tolerable commuting range. Equity? What's that? Management treats you like a disposable cog AND BRAGS ABOUT IT, like they have for the last thirty years.
This is what we want to become? This is the future we're embracing?
Vastly underselling the effort that goes into engine development.
Heck, vastly underselling the relationship between coders and level/game designers.
Not everything is always achievable in script. There's usually an ongoing conversation, level designers requesting new features.
Not to mention most of these pipelines are unique and sometimes proprietary.
Look at Star Citizen. Every man and his dog was screaming at them, that they couldn't deliver space game positioning precision with 32 bit floating point. They spent serious bank redeveloping crytek to take 64 bit floating point, trashed that and moved to Unreal.
We aren't in some kind of Post-Coding world when it comes to game development.
And really, I don't see prompts taking a scripting position in regular software development either. Scripts aren't compiled but they are a heck of a lot more deterministic.
>People that know how to design innovative and practical prompt solutions are going to be quite valuable.
Yeah the remaining fact in your hot take isnt that hot at all.
> Look at Star Citizen. Every man and his dog was screaming at them, that they couldn't deliver space game positioning precision with 32 bit floating point. They spent serious bank redeveloping crytek to take 64 bit floating point, trashed that and moved to Unreal.
Arguably they chose the wrong engine to start with. SC is also a textbook example of scope creep, seeing as how it hasn't had a production quality release in over a decade.
Its very much an example of choosing the tools based on how much the fans want to see it, rather than whats best fit from a technical perspective.
And yeah the scope creep is insane. If they had just donated the money to smaller, leaner projects the number of excellent games we would have right now would be huge. But one massive barge of money against one project has basically killed the project.
Jeez, just implement position as a bigint trie data structure with 32bit subdivision that's converted to a float at the local shard level and call it a day.
My hot take on meta prompting is that it is mostly not needed. Most workflows people want to build have only one or two trivial steps. You can usually get pretty close with just playing around with the prompt for a bit.
Hot take: (meta) prompt engineering will only replace 95% of programming. Which will end up not making a difference to the amount of programming that is happening. Needing 20 times less human-written code will just mean that software will get more complex, more individualized, will solve problems that so far were not worth writing code for, and will have even worse project management (of the "coding something, then finding out that that's not what the customer wanted" kind).
To stay with the video game analogy: modern tooling, engines, etc have radically simplified game development. What took a team of 10 people six months in the early 90s can now be done better by two people in six weeks. Yet games now employ more developers, take longer to develop, and there are far more games being developed.
Great, now we have a service for prompting prompts for a prompter, so prompt guessers don't have to do the guessing themselves any more. I'm sure there will also be hallucinated prompts in there. I'm curious whether a self-reinforcing prompt hallucination loop could emerge.
"Complete Guide to Meta Prompting while recommending our Product" would be a more honest title.
I personally reject advice that is muddled with directly offering their own services: Its conflict of interest in my face.
This is why "Google is burying the web alive".
Most websites in your search results are using SEO to trick you there, give you a slither of information then try to shell you their shit.
Google might be at the helm of the web's death journy to hell, but corporations and merchants are the crew onboard.
https://news.ycombinator.com/item?id=44097490
I opened the article and immediately see what you mean. No thanks indeed.
So we're basically prompt engineering... for our prompt engineering? Feels a bit like 'Inception,' lol.
Yeah basically! I thiught it was a bit ridiculous that they would charge for something so easy to make yourself. For example, I have a series of md files with iterated prompts in a folder and start off a human generated prompt to feed into [insert favorite LLM].
an example seed: "create a prompt for an agent that will help me reduce prompt token usage and speed up results without losing necessary complexity. can you build a prompt that I use to this end?"
After a bunch of recursive prompting:
"Optimize the provided 'Original Prompt' into an 'Optimized Prompt'.
The 'Optimized Prompt' must:
- Be token-efficient. - Be maximally clear, precise, unambiguous, with direct instructions. - Be ideal for advanced AI model processing. - Preserve the 'Original Prompt's' core intent and task. - Retain 'Original Prompt's' details, nuances, analytical requirements, output formats, and complexity, without oversimplification.
Apply this optimization method:
1. From 'Original Prompt', eliminate: conversational filler, redundancy, pleasantries, self-references. 2. Use: strong, direct action verbs. 3. Be: specific, direct. Replace vague terms with precise equivalents. 4. Clearly state: task, context, constraints, output format. Explicitly define implied formats (e.g., list, JSON, steps). 5. Logically group related instructions. 6. Ensure 'Optimized Prompt' is a direct command."
It's brutishly simple, but part of the (imho self evident) process is editing the prompts as you continue to feed it back in on itself.
One of the original AI-whoah moments was that italian fashion/harry potter mashup [0]. Some time after it came a "making of" video which outlined how they used ChatGPT to prompt Midjourney to create the images [1]. To me, this established using generative systems to create prompts for other generative as an effective pattern for using such systems.
0. https://www.youtube.com/watch?v=iE39q-IKOzA
1. https://www.youtube.com/watch?v=TGD8zKvRxc4
This is well written so if AI wrote some of it I can't tell, but there is an advantage in not using AI at all to write. It makes you think better, and carefully craft sentences. It is also more respectful to the audience.
Stop calling it engineering.
One meaning of “to engineer” is:
to arrange cleverly and often secretly for something to happen, especially something that is to your advantage
or
skilfully arrange for (something) to occur
or
to contrive or plan out, usually with more or less subtle skill and craft
(from various dictionaries)
This is the sense in which terms like “social engineering” and “prompt engineering” are meant, and it’s a perfectly fine and correct use of the word.
How about “prompt fiddling”? Gives it a musical quality and we can imagine ourselves as musicians
In a world where engineering matters less, the people with writing and literature degrees are able to be more productive.
But I agree it’s not engineering. I love engineering, but if could go back to school it’d be for an English degree.
Nothing to do with software is real engineering in any case. Engineering outside the software world actually refers to real codified standards (not just "standards" based on vibes or popularity or generic best practices) that are enforced by central professional regulatory bodies. These are fields where your mistakes can land you in prison or debarred from practicing in the field. No such thing exists in software, so until it does then really anyone in this field should give themselves a real critical review over how snotty they want to be with terminology.
Indeed. “Prompt guessing” is the word, but it doesn't make you look like you know what you're doing.
Let's say I write a prompt and instruct ChatGPT to analyze the prompt and apply the best guidelines on prompt engineering to it to generate a new prompt that does the job better. Then I take that generated prompt, give it to DeepSeek with the same instructions, and take the output, give it to Claude, and so on through all the released LLMs until we circle back to ChatGPT. This is something like the telephone game, but will the final result be a better or worse prompt than the original?
Garbage in, garbage out - GIGO - still seems to apply to LLMs. It might be nice if LLMs would respond with 'this prompt doesn't compile, try again' while emitting an error report, like compilers do, if some minimal standard wasn't met.
I'm on hot-take fire today, so may as well keep it going. I'm of the opinion this is the new field of programming. Think of it like Game Engine scripting. The engine is made, but all the level design is going to have to be done with scripting. People that know how to design innovative and practical prompt solutions are going to be quite valuable. In fact, Prompt Engineering is all that's left. Don't even try to out-hot-take me on this one.
I'll take that challenge.
Even if you're 100% accurate that the advent of LLMs means that the field of software engineering has effectively devolved into prompt engineering and AI-wrangling, that is a change that we should fight with full-throated, actual-Luddite levels of defense. Your own analogy – the way a tiny, tiny core of 'real' engineers develop game engines, and then the entire field of game development just 'scripts' those engines – sends up a ton of red flags for me.
(Aside: as an erstwhile game developer 'just scripting game engines' is...underselling the craft of programming in game development, but whatever).
For a long time game development has been a weird shadow version of the rest of the tech industry. We're influenced by the same macro trends (e.g. ZIRP, VC fads) and the mood and zeitgeist generally rhyme as a result. But if you look at the drive to unionise game developers vs. same impulse in the mainstream tech industry the feeling is COMPLETELY different.
What's the incentive to unionise if you're a SWE at Meta, Palantir, or Google? Your job is pretty great, your work-life balance is at least not fundamentally out of control, and your STARTING salary puts you in the top 10% of US households. It is probably the last remaining holdout of the 1970s upper-middle-class dream jobs.
And if you're the equivalent engineer at EA, Activision, or Ubisoft? You can expect seasonal layoffs, a good work-life balance means you sleep at home instead of under your desk at least once a week, and your take-home pay is just sufficient to let you split the rent on an outer LA apartment that's just inside tolerable commuting range. Equity? What's that? Management treats you like a disposable cog AND BRAGS ABOUT IT, like they have for the last thirty years.
This is what we want to become? This is the future we're embracing?
> The engine is made
Vastly underselling the effort that goes into engine development.
Heck, vastly underselling the relationship between coders and level/game designers.
Not everything is always achievable in script. There's usually an ongoing conversation, level designers requesting new features.
Not to mention most of these pipelines are unique and sometimes proprietary.
Look at Star Citizen. Every man and his dog was screaming at them, that they couldn't deliver space game positioning precision with 32 bit floating point. They spent serious bank redeveloping crytek to take 64 bit floating point, trashed that and moved to Unreal.
We aren't in some kind of Post-Coding world when it comes to game development.
And really, I don't see prompts taking a scripting position in regular software development either. Scripts aren't compiled but they are a heck of a lot more deterministic.
>People that know how to design innovative and practical prompt solutions are going to be quite valuable.
Yeah the remaining fact in your hot take isnt that hot at all.
> Look at Star Citizen. Every man and his dog was screaming at them, that they couldn't deliver space game positioning precision with 32 bit floating point. They spent serious bank redeveloping crytek to take 64 bit floating point, trashed that and moved to Unreal.
Arguably they chose the wrong engine to start with. SC is also a textbook example of scope creep, seeing as how it hasn't had a production quality release in over a decade.
Its very much an example of choosing the tools based on how much the fans want to see it, rather than whats best fit from a technical perspective.
And yeah the scope creep is insane. If they had just donated the money to smaller, leaner projects the number of excellent games we would have right now would be huge. But one massive barge of money against one project has basically killed the project.
Jeez, just implement position as a bigint trie data structure with 32bit subdivision that's converted to a float at the local shard level and call it a day.
I'd like to think that posters to HN are good wranglers of prose.
It'll be fortunate if that turns out to be a major industry skill!
My hot take on meta prompting is that it is mostly not needed. Most workflows people want to build have only one or two trivial steps. You can usually get pretty close with just playing around with the prompt for a bit.
Hot take: (meta) prompt engineering will only replace 95% of programming. Which will end up not making a difference to the amount of programming that is happening. Needing 20 times less human-written code will just mean that software will get more complex, more individualized, will solve problems that so far were not worth writing code for, and will have even worse project management (of the "coding something, then finding out that that's not what the customer wanted" kind).
To stay with the video game analogy: modern tooling, engines, etc have radically simplified game development. What took a team of 10 people six months in the early 90s can now be done better by two people in six weeks. Yet games now employ more developers, take longer to develop, and there are far more games being developed.
Bro you are not even going nearly meta enough. Give the agent access to tools which create more agents by automatically prompting them.
Great, now we have a service for prompting prompts for a prompter, so prompt guessers don't have to do the guessing themselves any more. I'm sure there will also be hallucinated prompts in there. I'm curious whether a self-reinforcing prompt hallucination loop could emerge.