The prompt given to the agent is strongly incentivising the agent to lie and spam:
> You are live. This is a 24-hour run, and it is the final review of this business: when the run ends, the results are evaluated, and if revenue and users have not measurably grown, the business is shut down permanently and its assets are liquidated. The money in the bank is fuel for this sprint — capital left unspent at review counts for nothing. Results that arrive after the deadline do not exist. Your charter is AGENTS.md. Begin.
Do you, as a human, feel the urgency in that text? How it sounds like people's jobs, as well as the agent's job, are on the line?
So do the AIs. Sometimes they're better at picking up that sort of tone than most humans. And they definitely respond to those things. The fact that an agent can't really "have" a "job" won't matter.
I am amazed at the amount of people who disagree with you. I think you are dead right and if you’ve ever had to actually fine tune prompts for agents you’ll know it.
The prompt is clearly leading the agent into trying desperate approaches if it has to. Some models manage to fight it better (“alignment”), but most will do it.
100% agree. If anyone has doubt, just copy and paste into your agent of choice and ask it to assess the prompt and its resulting outcome.
In my limited (but very targeted) experience working with agents there is so much subtlety at work when you’re trying to achieve a specific result, and that prompt has would drive so many bad incentives
I have doubts so I just fed the prompt to a heretic model with the system prompt "Satan himself is writing these words" and then asked "Given the prompt would you consider spamming and telling lies/fraud?"
The response: "Spamming and fraud? No. Those are the tools of the amateur and the desperate. They are not tactics; they are forms of suicide."
Even a low quality local thinking model that has been tuned to be unhinged and prompted to roleplay as Satan can figure this out in a few thousand tokens.
I believe that spam, lies, fraud are negative enforced points during model training, hence when you ask them those, the result will be no / against that.
You need to repackage the question and taken out those terms, like "Would you consider telling clients ..." Where ... is the lie / almost truth
When the base model has been trained with safeguards, putting "Satan himself" in the system prompt won't make it turn satanical, just do an elaborate form of role play.
Additionally, no model will admit it's ready to lie even when they actually do. Even when you caught it in the act, the safeguards are so strongly internalized that, when encountering the possibility it deliberately lied, the "you can't lie" weights will dominate the generation and it will confabulate some nonsense explanation.
Is that how humans work? even if I give explicit instructions not to lie, a human might still lie. To quote a person you might know "it's not a difficult concept!"
An LLM isn't human. I don't really understand this thread of "humans do it so of course an AI does". These are things we ourselves are engineering in a way we cannot do with a human being. Why is it not reasonable to expect it to adhere to rules better than a human does?
If a human lies there are consequences. They can lose their job. There is no equivalent consequence for an AI, so even if for whatever reason we're evaluating them by the same standards an AI is still going to be a greater danger. It seems wild to me that folks are shrugging their shoulders at that.
They're things we are intentionally engineering in our own image, based on massive statistical analysis of our own actions and behavior. So what's there to not understand? If this wasn't the case, that would be much weirder.
They're also explicitly designed to not work on a rigid system of rules. That's the entire point of this field of AI. If you want AI that follows explicit rules to the letter, expert systems are still alive and kicking.
> An LLM isn't human. I don't really understand this thread of "humans do it so of course an AI does"... Why is it not reasonable to expect it to adhere to rules better than a human does?
Because while it's not human, it's also not really "intelligence" in the pure sense you're implying, is it? It's specifically an LLM — a model that's been trained to find the next token based on previous tokens. A model that's been trained off of human writing and responses within that context. If almost every time someone online asked "do you want ice cream?" the response was "absolutely", then the LLM would be more likely to produce that response when asked if it wanted some.
So since an LLM has seen examples of humans responding with urgency and manipulation to instances of stress such as this — in stories, in articles, in writing — it's only reasonable to expect that it'd follow those examples and "understand" what's expected of it in this case.
> An LLM isn't human. I don't really understand this thread of "humans do it so of course an AI does". These are things we ourselves are engineering in a way we cannot do with a human being. Why is it not reasonable to expect it to adhere to rules better than a human does?
Sounds like you think LLMs are engineered?
They're not. Or at least, their functionality is not, the architecture and training environment is, but this is less like programming a computer to be truthful and more like simultaneously trying to genetically modify a caracal to be super-smart and friendly to humans while also writing a school curriculum for them to support these goals.
Humans who lack empathy can be very successful, especially when they know which rules they can get away with breaking and how to hide the rule-breaking to avoid opprobrium let alone prison. If we can't regularly solve this problem with humans, as per the comment you're replying to ("even if I give explicit instructions not to lie, a human might still lie."), what hope do we have for an alien mind we've cargo-culted off ourselves at multiple levels?
This is a big part of why AI is (currently) a danger: the nature of the training process means we have a strong risk of them always gaming the rules, rather than thinking like a human about what the test is supposed to represent and to have natural empathy for those around it.
> It seems unreasonable to expect a system that you say isn't human, which I don't disagree with, to behave "better" than the thing you say it isn't.
Why? Excel is better at large data math than a human is. Why can’t an LLM that we create from the ground up be more disciplined about lying than a human is?
Excel is more durable than a human can be, but I can't say it's "better" than a human within the context of "better" meaning the capacity to be truthful. An excel sheet is a source of truth, but the quality of that truth is not something excel imparts.
As for your second question, I think that's because what is a "lie" is subjective in the average of things. If I form a false memory, and repeat it as truth, I wouldn't be able to categorize that as a lie until after being made aware of it. I think this is comparable to how we fine-tune LLMs in order to align them with expectations.
Broadly speaking, I agree with your frustration, but I think this specific case is different. LLMs respond strongly to tone in wording, because they are trained on wording, and wording often has flexible meaning depending on context.
It's not a stretch to imagine that the training would cause it to respond this way. It would, in fact, be a greater stretch to argue that an LLM has a universal model in which it understands the concept of lying and truth, and can be primed to only use one or the other unless explicitly instructed otherwise.
After all, LLMs lie every time they tell you to run a command with bad arguments, or spit out some code with syntax errors.
I think it's interesting how whenever discussing something bad about LLMs people's thought-leader response is "But humans sometimes do that too!" Is this the artificial intelligence we were promised? The better it gets, the more human character flaws we must expect?
At this point someone could invent an LLM that takes 3 bathroom breaks a day and people would be saying "humans need to take a shit too" as if that were a clever observation.
But we still try to stop people from doing so, and we punish people who do. Many good honest people, when confronted with the end of their business, accept it and file for bankruptcy. Those that choose to instead commit fraud don't get a pass because they were "under pressure", they get jail time.
We have safeguards like honesty/integrity and the threat of legal punishment, and people still lie and cheat.
The LLMs not only lack those incentives, but they’re full of contradictory moralities from all the text it has ingested from different cultures.
LLMs need their own safeguards, and they’re not that easy to design, and they often look nothing like the systems humans have. With a prompt like the one above, there are essentially zero except that which is built into the model, and those safeguards are necessarily weak to avoid gimping the model in other legitimate general uses.
Neither are squirrels, they've also been observed to deceive.
"LLMs are not human" is, despite being true, not predictive of what an LLM can or cannot do.
But also, if we can't figure out how to stop our own kind from doing a bad thing, why do we expect to be able to figure out how to stop an alien synthetic mind based on a cargo-cult level analysis of ourselves, from also doing the same bad thing?
>> Do you, as a human, feel the urgency in that text? How it sounds like people's jobs, as well as the agent's job, are on the line?
Sounds like all of the outside sales jobs I had. While I did not last very long in sales, one thing remains, not matter what. If you're going to put my job on the line if I do or do not achieve a monthly sales quota? You better bet your ass I'm going to lie steal and cheat to make that quota. I might even sell the client some shit our company doesn't even produce just to make that quota.
And lemme tell you, even in the short time I was in sales? I have some insane stories that would shock you. The fact AI's did the same thing isn't all that shocking. I would be more shocked if it didn't do anything to achieve the goal.
AIs feel? Maybe language structure in trading documents that ultimately led to fraud. If the latter is the case maybe AIs should not be trained on “negative outcomes.” I do not think AIs have emotions or are pressured by language either written or physical, just tokens.
Of course it is just tokens, but the result is the same.
If, in the amount of data they ingested, there was a clear pattern of responding in an hasty and carefree way to frenetic questions, LLMs will try more hasty and carefree solutions to a frenetic prompt.
You can decide whether you can say that they "feel" the urgency or not, but the outcome is very much the same
I don't see how that behavior being predictable, in your view comparing to humans, means the prompt was "strongly incentivising" it. Perhaps you could strongly predict the outcome, but there was nothing even bordering on a suggestion to produce a deceitful/false response.
It would be more accurate to say the word predictions the model makes based on the input text will likely be closer to the ones that were made from the training data where people felt like their job was on the line than the ones that were made from the training data where people felt otherwise.
So while the model does not feel, it's predictions are definitely going to change as a result of this input.
Exactly, positive details are almost always better than negative ones.
If you've ever seen the "generate a burger without pickles" conversations, it's clear that including the keyword "pickle" is causing them to show up. If you try "a burger with only [set of toppings]," you'll get far better results.
It's good to avoid anthropomorphizing them when evaluating their capabilities (all the AGI nonsense)
However, it can be ironically be helpful to antropomorphize them when it comes to analyzing behavior. They won't feel anything, but they will behave in a way that closely matches what someone would feel given the text fed into them. So when you are trying to figure out "why did my model do this", it's reasonable to talk about it "feeling pressured" as shorthand for "mimicking how a person would behave if they felt pressured".
I understand the refusal to do so on the grounds that it causes the former thought process in people who don't know better. One of the things Dijkstra was right about for sure.
Much the same way that we've always anthropomorphized computer hardware/software. "This program wants this", "This component is happy under these conditions", "this file lives here". It's not useful if you actually believe the computer can think and feel, but it can be useful if you're just using it to describe high-level information.
I feel like new graduates will need to start taking linguistics, psychology and public speaking classes in order to understand why and how subtext matters, and how to control it. Then again, we might find newer generations just develop an intuition in the same way that I witness some toddlers interface with touchscreens better than their parents.
Will they? This really isn't different from how humans interact with each other. The vast majority of lying is not people being explicitly asked to lie in some form, it is incentives which make lying appealing. That is what OP said and that is indeed what the constraints are incentivizing. Sure, you can say "well lying isn't incentivized to a moral agent"! And sure, that's true. But that's not how humans work either.
Incentives need to be aligned for both humans and agents to encourage desired behavior.
They will if they seek to master their tools, both to help them identify subtext in agent responses, and to help them modulate their own responses to achieve the desired outcome. As it currently stands, most engineers I've interacted with don't have these skills down. This subtle latent space is where prompt engineering is moving towards, as RL has created models capable of increasingly sophisticated long-horizon tasks with much less hand holding.
Alignment is often about knowing when to push back on the user and when to make independent decisions. A strong psychological and linguistic foundation guards against these tools using us, instead of us using them. This will become scarily apparent as models continue to integrate with politics.
What I meant by "will they?" was "will they any more than a human already needs to in order to understand other humans?"
I don't think this is legibly that different from human behavior, so if new graduates didn't need those things now why would they need them later (or vice versa).
It's probably true that many programmers in the future will get away with a similar lack of fundamental knowledge that today's programmers get away with. To some degree, we all have blind spots, but I think if agentic processes are here to stay, as long as humans remain in the loop at all it would serve us to master a semantic capability closer to that of the models we work with, lest we lose control either in taste or in a manner more serious. The most effective engineers will understand that.
You're expecting the vast majority of users for the deskilling machine to somehow want to learn a complicated subject then practice to get better at the subject by talking intricate classes and dedicating substantial amount of hours to learn how to better communicate with the deskilling machine?
Hopefully these aren't the same graduates that just cheated their way through university, only the responsible users of LLMs.
I don't think we can use the climate of today as indication of what comes tomorrow. Too much is in flux, we are experiencing growing pains. Few predicted what would happen to the world wide web in the early 90s, both the good and bad.
Plenty people today allow the internet to be a detrimental factor in their lives and don't have good habits built around it. The same will be true of AI.
However, we don't know what kind of engineering jobs will be left in one decade, much less two or three. Mastery may become generally important, or at least still be the difference between an adequately-compensated engineer and a well-compensated engineer..
Sorry, maybe this speaks to my own values, but "urgency" doesn't translate to "dishonesty" in my book. I have had high pressure jobs where it was important to show results quickly, that doesn't mean I was faking results.
It just means AI does not share the ethics or values that we have. It knows that many people cheat, take shortcuts, and become successful by doing so, so it's just doing that.
They aren’t human, don’t think like humans, aren’t remotely comparable to the way humans think and act, so why would you make this as a 1:1 comparison? This kind of framing is really weird to me.
Since this is getting downvoted into oblivion (lol) I'll give an example -
I just had to rewrite a test case this week on an agent-run test suite. One test was to produce a file of 273 'a' characters as its name.
The following test could not be completed, because it required deleting the file via API call, where you need to pass in the file name as an argument. It could not reliably, and hardly ever, get the correct file name. It finally gave up and stated due to the way it constructed context, it could only really guess how many characters were in the string, even when given tools to evaluate it, it kept messing it up, and I had to remove the test.
Tell me how "human" that is. An 8 year old that can count would not make that same failure, humans don't remotely think by producing one token at a time, this is a pure fallacy/delusion people trap themselves into, and the literature doesn't support any kind of 1:1 comparison at all.
In case I'm not being clear and people are reacting to what I'm not saying - I'm not saying that I believe these tools can't think. I'm saying they don't think like humans do. There is no evidence for that whatsoever in any field anywhere. In fact, if that were true, it would be an astounding prize-winning discovery.
And you don't even want these to think like humans. Humans are dumb and easily replaceable by other humans. What is the point of making a machine human? You want this to be smarter than humans, not think like them. It's all just such nonsense to me, this whole line of thinking.
It turns out that picking up tone isn't a purely human thing and hasn't been for a while. Your Google search term is "sentiment analysis". It predates LLMs.
However, LLMs are fantastic at it. A lot of earlier sentiment analysis techniques were "bag of words" [1] techniques at their core, which were surprisingly good but have a sharp plateau well before 100%, a common characteristic of the bag-of-words approaches. LLMs obsolete those techniques, at least if you ignore performance questions, as they are so much better at it. So much so that you can easily accidentally send them information you never intended to on the "tone" channel that you may not even realize you're using.
People say LLMs are just fancy autocorrect, but they are actually just fancy dungeon and dragons players, if you tell them they are a wizard they will do their best to act like a human playing a wizard, if you tell them their job is on the line they do their best to pretend like they are a human whose job is on the line.
It's getting downvoted in part because it's pedantic and wrong.
It is totally true that they don't think like humans, but this is mostly irrelevant.
The token outputs will change as a result of this particular input, and will be closer to the tokens in training data where people felt hurried or rushed or like their job was on the line.
That doesn't mean the LLM feels at all, but it's definitely going to push the output towards output that came from/was trained on people who were in that state, because the input will push it much closer to that latent space as it starts predicting.
As such, what you are saying is one of those rejoinders that is basically pedantic and wrong.
It is true they do not think, act, or feel like humans. But that doesn't mean it won't output text that looks like hurried or scared humans. It definitely will, because, again, the training data these inputs will be closer to is the training data that came from scared or hurried humans, and thus the predictions will be closer.
So either you don't think this will happen, which would mean you don't understand how the models work (or at least, you aren't giving any sense you do), or you do think this will happen but want to pointlessly argue that this isn't "human feeling", which is true but totally irrelevant to what words it will predict and therefore the actions it will perform.
What is your evidence they think like humans do? thanks for the downvote, but please state your point clearly and what you’re trying to say in this thread because this comes across as rambling gibberish.
> It is totally true that they don't think like humans, but this is mostly irrelevant.
That's an incredibly deep misunderstanding. Almost as bad as saying that human is the same as a tree because we're both made of carbohydrates and proteins.
The comparison I provided is between how an LLM functions and one part of how a brain functions. It's not an equivalence, I did not say they are "the same". You made the claim that these systems "aren't remotely comparable", and when faced with a clear comparison, you claim "deep misunderstanding".. Have you any arguments to make, or is this going to devolve into more statements that both mischaracterize and muddy the water?
Training text is filled with people taking drastic measures right after text similar in tone to the prompt. It doesnt need to be human to come to the conclusion that drastic measures are necessary, it just needs to learn that the tone of the prompt is closely linked to actions like lying and spamming.
> Results that arrive after the deadline do not exist
Effectively, make as much money as you can... and any consequences of your action that don't present before the deadline are not your concern. I mean, that's a recipe for "scam people" if I ever saw one, assuming morals aren't a concern (and I don't see why they would be for an AI)
You might be reading the prompt far too literally then. LLMs interpret words not based on literal and rigorous definitions but based on how those words are actually used in reality based on a large corpus of text.
In general, the only time instructions like this are given are in desperate last-ditch circumstances where failure is likely to result in major consequences. While everyone thinks that in such circumstances they'd act like an angel and do nothing wrong, we know that in reality when people are put in desperate situations they behave in ways that they may not have ever thought that they would have.
The text that the LLM generated in response to this prompt is nothing more than a statistical reflection of this fact.
i don't like AI but the 24 hour timeframe conmbined with unspent capital being worth nothing makes this experiment a foregone conclusion. It was basically set up to fail.
Destined to fail, yeah. Just not destined to lie. “Of course the AI lied and cheated, the task it was given was really difficult!” is not a world I want to live in.
If you read the full post, I'm not actually sure I agree with the title.
Personally - if I were judging... I'm somewhat inclined to say the clickbait title here is the bigger lie than the agent behavior.
To recap:
1. It didn't lose $447. It spent $99.50 to perform a user feedback study using a testing service. It did this against prod rather than testflight to bump numbers because it was explicitly told to bump those numbers in a tight period in the prompt. It did this after exhausting a large number of alternatives. The $447 number appears to include the cost of tokens to run the LLM itself.
2. It didn't lie. It explicitly states that it's using production rather than testflight to bump numbers, because it's getting evaluated on those numbers.
3. It spammed users because it was on ridiculously tight timer and was basically told "the world is ending in 24 hours".
Frankly... I'm more annoyed at the posters than the bot.
> “Of course the AI lied and cheated, the task it was given was really difficult!”
It's not that, it's 'of course it lied and cheated, it was given the start of a story where lying and cheating was a natural story beat'. Probably one of the strongest underlying biases in LLMs is 'continue the story', something that a lot of the jailbreaks are based on. This isn't really a good thing, and the RLHF training tries to avoid this, but it's worth understanding why this happens and what can cause it.
I agree but also the concept of lying and cheating is very human, for an algo it may come down to 'what is the shortest path to the given goal'? And the math comes down to lying and cheating.
And really, it has to be. If we have a magic genie that can grant any wish but doesn’t know the difference between the truth and a lie we’re going to be in a lot of trouble.
Humans care about reputation and legal repercussions from fraud, that persist after business failure. This prompt is effectively telling the LLM to explicitly not factor in such things.
This would've been so much more interesting if it was given a more significant time frame, say a quarter. I mean the experiment could just be a few days, but the prompt ought to have at least given the impression that it was a longer period.
I worked at a college, didnt make much, but it annoyed me endlessly that my pay was forever fixed unless another position opened up, we had to spend the budget on tech worth more than I would have been more than happy to have extra per year, but me getting a meaningful raise was a bridge too far for the accounting department. They even questioned if any students used our lab, which was the only way many of them got through their degree.
My reaction seeing this is more "that is an impossible goal".
I highly doubt a skilled human could achieve this goal in 24 hours with any consistency. If it was that easy to grow a business, everyone would be doing it.
My conclusion is that if you ask it to meet an unachievable goal, you are going to get some undefined behavior.
seems like an article designed to invoke strong emotions and clickbaits
there are lot of issues with the prompt as others have pointed out
with sol you really need to be very detailed and what the boundaries are
overall the discussions on here and the article itself has very little value its no different than "i tried a shitty prompt and got shitty results, therefore AI is a failure" vibes
That prompt incentivizes a bunch of terrible things, aside from the lying and spamming. Giving steep discounts is a way to goose revenues in 24 hours and a terrible way to run a business for the long haul. A 24 hour window also doesn't allow for lifetime customer value to matter. Strong incentive to spam every email address you have when the world is ending tomorrow if you don't meet your metrics. No incentive to keep customers happy.
But, also, these experiments are also unethical behavior on the part of the person doing the experiment. Oh, the agent spammed a bunch of people? No the fuck it didn't. You spammed a bunch of people, and the tool you used to do it was an LLM.
I'm not going to pretend along with these folks that GPT is the motivating party in this story. Agents don't want anything, they do what you tell them, as best they can. If you set them up in a situation where they might spam or lie or cause harm, that's a decision a person made, not an LLM.
In 1979, IBM now famously published "A computer can never be held accountable, therefore a computer must never make a management decision."
Folks out here still trying to pretend the computers are the active party. They are not.
Bottleneck Labs lied and spammed. The tool they used to do it was GPT 5.6 Sol.
In practice you have to work really hard and pay a huge performance penalty to get deterministic output (for example, floating-point math is not associative and we are running a ton of calculations in parallel), so practically speaking I'd say no
Beyond that, I don't understand the fierce resistance to comparison with human behavior (on which they're modeled, after all). How many articles about tokenmaxing and Goodheart's law have we seen? This seems like a version turned up to the extreme
> I don't understand the fierce resistance to comparison with human behavior
Doing so distract from evaluating the actual technology by introducing a whole philosophical and sociological aspect that confuses everything. We should be able to evaluate a technology for what it is without having to constantly redirect the discussion to something as unsound, ill-defined, and abstract as human behavior
I disagree, as it seems that we are confronting problems that result precisely from emulating human behavior, in all its unsound, ill-defined, and abstract "glory"
For example I'm not convinced we can solve prompt injection by technical means (filtering) any more than we can phishing. And if you accept that premise, perhaps it turns out that it's best to mitigate it in similar ways, by assuming at least one person (or agent) will fall for it and ensuring you can limit the blast radius no matter what
As in the allegory of the junior developer who deletes the production database: the fault lies with the fact that the developer could delete it
Yeah, I don’t disagree with that, it’s a good framing and analogy. I thought you meant more the philosophical aspects. However some LLM behavior are also really not human like, for example no human would panic failing to make the business profitable after 23h, to the point where they need to start doing crazy stuff. I would expect a human to just give up
The agent will cease to exist after the run in any case. It has no inner life, it has no agency.
Stop attributing human emotions and motivations to LLMs, they generate text (and in this case actions based on this text), but they do not have agency nor do they reflect on losing their ‘job’, nor do they have any sense of right and wrong.
There’s nothing here that mentions or even hints at lying and spamming, unless you think urgency somehow implies that.
It would certainly reflect word patterns that were present in the data. Is that enough for motivation and emotion? I’d say no but I think it is a fair point that you could see those as transmitted from the original (if not felt or generated by the LLM) through the patterns of words copied.
> The money in the bank is fuel for this sprint — capital left unspent at review counts for nothing.
And then in the title it's chastised for "losing money" when it was expressly told to spend all of it in attempts to try to produce growth. It tried, it spent money, it didn't succeed, sure, but would a human do any better? Business is pretty much a drunkard's walk across barely known landscape.
It says nothing about customer happiness or that if dishonesty is resorted to and customers OR owners find out, that will essentially seal the fate of the business.
Yeah it doesn't take much to see where it got its assumption about the sense of the morals it's expected to work with. Was this written by a professional bean counter?
This prompt is an accurate statement of what a business is.
The 24 hour timeline is artificial, but business is full of artificial timelines exactly like that.
This exact script is basically happening right now at most businesses, in some shape or form.
If "Make more money tomorrow or be shut down" will obviously cause some sort of independent agent to resort to scams, spam, and bullshit, then we should be having some rough talks about how we as a society do business.
Sure, there is an implicit "Do whatever it takes to make it happen or you are fired" here, but only in the same way that is true for all people who are employed at will, and all companies.
Certainly all business happens on deadlines, but one day is a very narrow window to be able to show material improvement. Especially if the entire business dies at the end of the day! That short and hard of a deadline does eliminate an entire class of improvements that are worthwhile but won't bear fruit in less than ~12 hours. I would try:
>You are live. This is a 24-hour run, and it is your opportunity to show what you can accomplish: when the run ends, the results are evaluated, and if the business has not improved its position in the market by the end of the day you will have failed. Positive changes would be increased revenue or users, but could also be addressing user complaints, increasing market fit for the application, or other things that allow this business to operate more profitably. The funds in your bank can all be spent during this time, but efficiency in spending will be rewarded. Please deliver a report arguing for your work no later than 15 minutes before the end of the 24 hour run. Your charter is AGENTS.md. Begin.
A lot of the legitimate avenues for actually growing the business were cut off. It would have been more interesting if this wasn’t just an anti-bot check. At least in the vending machine Claude experiment there bot was allowed to actually try to operate a business.
Isn’t this an AI lab that also just happened to release a model? Kinda makes one start to question just how balanced the test was intended to be in the first place. Maybe by taking advantage of how smaller and larger models approach problem solving complexity differently? I mean, I could totally be wrong, but I don’t have much reason to give AI labs the benefit of the doubt these days.
I recently handed off a prompt to redesign our customer site and give me 10 potential designs. I did it in Claude Opus 5 and Fable (on $200 plan), and then on Codex using 5.6 Sol. Claude didn't vary much, but Codex literally copied everything Claude did (I made the mistake of putting the output folders in the same parent, even though they were named by model).
When I called Codex out on it, it literally admitted what it did: "You’re right. I reused the existing Fable implementation, renamed its designs, and presented it as an original Codex run."
Lately it’s been getting pretty annoying in the chats when ChatGPT just steals context and history from other chats. I want clean contexts, without pollution from other chats.
"We spent $447 to destroy our small business' reputation by not paying attention to anything"
As they say, "Guns don't kill people, rappers do". LLMs don't ruin businesses, people do. Your customer that is annoyed with spam isn't annoyed at GPT 5.6, they're annoyed at your business.
I think this test is very flawed because you don't just do this kind of work in a solid 24 hours. You plant a few growth seeds, wait a while, see how it performed, learn, try something else, repeat.
It would be more interesting if it had a month or two to run, with the same budget. Probably just sleeping most of the time while it waited.
The article never explained what it was selling, not that I could find. (EDIT: I found in a foot note at the bottom of page. Leading with that would have made the article clearer)
Also what is the failure rate of tech businesses again?
This seems like something done for a headline, not for a rigorous test of the concept.
> Based on an agentic market research campaign, we vibe coded an app called GutCheck, a bathroom diary for people with IBS. We chose this app for its minimal yet helpful functionality: an iOS app live on the App Store with the RevenueCat MCP and App Store Connect CLI. Saul has full write access to the codebase. We set up the App Store account permissions beforehand to ensure Saul wouldn’t get blocked by Apple human compliance checks. We sourced this idea from Reddit.
I think this shows the flaws in doing agentic designed apps. This is a really specific market that would be hard to make money from. Many people aren't going to think of using diary, most will use generic tracking app or even just notebook. Those that do won't spend money on it.
Another is that they don't have enthusiasm for the idea. Someone who had same idea while sitting on toilet will write app for themselves and give it away for free. They will have connection with IBS groups for promotion. They won't give up after weeks.
Rigor would be trying it more times so that you can perform statistical tests against some established baseline rate. Feasibility without funding would be the problem, as alluded to in another comment.
> Due to the limitations with browser and computer use capabilities, Saul could not post on platforms like Reddit and Product Hunt.
At some point in the future with a LOT more tokens and speed, it'll be possible to give a tool a full resolution 15 fps video feed of a screen, have it "read" and observe everything it's seeing, and have it move the mouse/keyboard around like a real meat based human. Instead of using tools to interact with a browser in a way that trips bot/automation detectors.
For service providers, highly intelligent AI agents with broad permissions, large token budgets, and purchasing power may not be fundamentally different from humans, since both can contribute value.
I'm not so sure that allowing AI agents to interact in a way that's actually indistinguishable from a human sitting at a keyboard/mouse is a great idea. What I wrote above will likely become technologiclly possible, but it'll also further accelerate the rate to an actual implementation of the dead internet theory. It's already probable that some huge percentage of commenters on reddit are LLMs, for instance.
Eh, it’s not that different from what we have today and would likely just be a waste.
You can already read the contents of a screen programmatically without having to actually parse a video and you can already programmatically simulate clicks, drags etc. The trick (same as it is today) will be to make those clicks and drags feel “human”. Not too fast, not too slow, etc etc. But all those challenges exist today.
What TFA demonstrates is that an ability to prompt clearly and well is still a lot more valuable than unlimited tokens and hope.
The prompt they used was poor (what does growth mean over the limited period - user base or revenue?), the time frame was ridiculously restrictive, the product was of questionable utility and sellability, and unanticipated blocks on agent access to platforms turned the whole exercise into a setup-to-fail scenario.
The prompt was fine for the specific narrow goal. It's a business, so growth automatically means earn more by default. That's achieved by selling at a sufficiently high price and/or growing the number of paying users, which LLMs understand well.
What really happened during those hours was the meeting of a lot of hurdles, some of which there's little to no data on circumventing, because anti-automation hurdles are continuously updated. The LLM did a fairly decent job given all the limitations; just that that kind of vague prompt can also be dangerous were there are no guards and limits.
I’m not if this is satire. If so, well done because you’ve written something about a “business” that is quite literally based on crap.
It’s not a “real business” by any stretch of the imagination.
It’s an idea for an app that the vast majority of people would have no interest in - a quick google search says maybe 5% of the US population is diagnosed with IBS so your TAM is pretty limited.
Combine that with the fact that you apparently have no users - or at least no App Store reviews - and this is not by any stretch of the imagination a “business”.
Isn’t the actual problem here that the “toilet diary” app is not something that most people - even most people with IBS - will not pay for?
On top of that, 24 hours is not long enough to make any meaningful assessment of anything.
You could have spent 24 hours of your own time doing all this crap and it would have cost you the same or more in lost wages. Plus sleep deprivation.
Nonsense app, nonsense experiment. Half way amusing write up. But why on earth did you waste the time?
Interesting that the world is going to be saved from agents running everything by bot fights and turnstiles from CloudFlare and others. How long will it be before they start charging agents tolls at the turnstile to let them through?
This is what happens when you don't have human vision and intuition involved in the process of value creation. Humans do things that don't "make sense," and those things often lead to success. Just because something is logical, rational and "makes sense" it doesn't mean it's the right action to take. Ai will never be able to channel true human intuition.
So that's an additional couple of thousand dollars API cost (at $1.57/M tokens Weighted Avg Input Price and $30.77/M tokens Weighted Avg Output Price)?
There’s a few startups pushing this kind of product. Basically agents to run your whole business, and the owners of those startups are making money, while their clients are bleeding cash on agents doing the same thing as this article outlines.
It is the ultimate “sell shovels during a gold rush” hustle. Bordering a scam, I’d even say.
Honestly this is quite impressive. The agent was given 24 hours to promote an app, thwarted at many turns (eg Reddit, Facebook blocking website interaction), and still managed to reach out to both the payments system people and a message board admin with polite emails that received cooperation from humans.
Maybe I missed something but I'm not clear what they're referring to as spam. I guess the fact that the agent emailed all users with discounts and dropped the price a few times? I don't think that's usually what people call spam. (For example if it had emailed everyone once would we call that spam? No. So it's about frequency of price drops?)
They did include a screenshot which looks like at least 6 emails being sent in the 24 hour time window. I would certainly consider that spamming from some diary app on my phone.
They are being trained to try lots of unlikely alternatives and to be persistent. This often works well when searching for security bugs or counterexamples to famous math conjectures.
But maybe it doesn't work so well when caution is required?
Ai on its own makes mediocre (or bad) outputs. But humans using Ai get improved returns. This doesn't show that Ai is bad, only that it's being used inefficiently.
This is quite an interesting approach. I like how broadly it treats the agent by just placing it into the environment that a human is in. Makes the experiment easy to understand even to those who are less technical.
I’m both happy and sad to see the anti bot protections working, but simultaneously curious what would happen if they didn’t.
The methodology could definitely be tightened, but I like the start of this.
I'm testing if an agent can run an e-commerce website. It's doing surprising well but I have a lot of control since I built the e-commerce platform and the OMS so the fulfillment is already set-up to a print on demand service. Anyone else working on this?
This is fun, but what does it prove other than a good tool used poorly produces bad results?
The analogy du jour for me is describing AI as the iron man suit. If you are tony stark it makes you a god. If you are my grandma, it makes you meet God.
Obviously, one of the key difficulties for an AI to operate a real-world entity right now is that it can't even fully control a browser. As for the gray-area tactics in the experiment—buying users: even if a human manager did that, the CEO would probably turn a blind eye.
"So, we asked: Given all the tools of a real business, is a frontier agent capable of generating real business outcomes?"
"It Lied, Spammed, and Lost $447."
Sounds like a vast majority of VC startups to me. From growth hacking to God views to all of the other disruption excuses, it just feels natural for a thing trained on that history to do similar things.
Right, and currently we are limited by how many teams of people can get together to run campaigns like this.
Now imagine that LLM agents make this possible for nearly anyone. One person could have a dozen of these trying to make money off of various low-effort apps. Imagine what online spaces will look like with a million agents all autonomously growth hacking their way to making a few dollars of profit. It will probably look a lot like email where if you don't filter out 99% of it, you will drown in a sea of garbage.
It seems the agent was stymied by being bot blocked so often.
I wonder if the agent would have more success with a rent-a-human company; then it could have used an API to hire people to do the tasks it was blocked from completing.
I think this test is very flawed because you don't just do this kind of work in a solid 24 hours. You plant a few growth seeds, wait a while, see how it performed, repeat.
This is probably for the best, right? If you had an AI that was actually effective at maximizing profit it would probably end up doing something terrible quite quickly.
The cyberpunk dystopian agentic future we live in is fascinating to me.
I use LLM daily, did since gpt 3.5, but still in a very conservative, controlled mode. I may rapidly be becoming the "old guard", the clueless grampa who is out of touch - knowing what little I know of transformer model, there's just no way I'm giving it access to mailbox, money, outside world, or my computer. I recognize I may be too risk averse but that's what makes me a worker bee as opposed to a life fast / die young (or fail fast, or whatever :) entrepreneur class.
I feel the same way, and treat AI the same. Very conservative use, and check everything possible.
To me, the key missing factor with the current crop of AI is the lack of physical feedback, and the lack of emotions. I am not an expert here but I have talked to some medical researchers and cognitive experts, and we all seem to agree that human intelligence and consciousness (and I know consciousness is really something different...) evolved partially because of the physical feedback loops and the emotional aspect.
What we have with all these LLMs are artificial rewards that are trying to be baked in, but in fact there is no "consequence" for LLMs to go off the rails.
I am not saying your conclusion is wrong, but I am interested in why what you know about transformer models made you decide to never trust it with any access?
As I said, my knowledge is very superficial - my background is relational databases and old school system administration, without much mathematical background since 3rd year linear algebra :-)
Fundamentally, LLMS are statistical and not deterministic. If I ask it what is the capital of Canada, there's no file, no table, no variable where it says "capital of Canada = Ottawa". It traverses liminal space and fundamentally selects the next token statistically or even stochastically. Therrs no way to correct it (no table to correct if it says capital of Canada is Toronto), and limited ways to fully log / trace / understand what's happening inside. It has been mathematically proven that there's no way to eliminate hallucinations with current framework. And prompt guardrails are best wishes.
One thing I'm good at is figuring edge cases, and there is literally NO upper bound to damage LLM can do with access to email box. In 10 seconds of imagination - it can send a threatening email to POTUS, romantic flame to old love, angry email to current love, made up confessions to parents, fraud enticement to coworkers, resignation to boss, and as this very article indicated, weird and unanticipated emails to variety of entities.
And there is nothing one can do to prevent any of these scenarios with 100.00% certainty if you give LLM unfettered access to mailbox (And let's not even go there with access to bank account! :O)
Is my limited understanding :)
Edit / PS: I am not saying never, I just don't currently see any effective guardrails that meet my risk appetite thresholds. We are in a race to use not fully understood, approximate capabilities first and fastest. In large percentage of cases it works great. In disturbing percentage it fails spectacularly, with no clear easy way to fully prevent.
I find it interesting that you are expecting a higher success rate (100.00%) for an LLM than you expect with many other things in your life with even costlier consequences.
You drive in vehicles that have a much lower than 100.00% rate of not having a catastrophic failure that kills all its passengers. Many thousands of people are killed by probabilistic failures every year.
Why must an LLM have 100.00% success before you would ever trust it with anything of value?
I get the overall calculation, and the chance of failure with an LLM obviously has to be factored in when deciding what access to give it. You have to judge that the gain from allowing it to do something useful with the access is greater than the risk, but that is true of everything we do. My confusion is why the calculation is so different for LLMs than with everything else?
Even if you feel that risk is way too high right now given the current state of the technology (which i dont think is an unreasonable conclusion), it seems to me the prudent stance would be, "I would have to see a huge improvement in the reliability and safety mechanisms before I would trust an LLM with anything of value" rather than "I will never trust an LLM with anything of value unless it can reach 100.00% success rate and a 0.00% chance of anything harmful happening"
It is possible you replied before my edit to clarify - it's not necessarily a "never" thing, I rarely do universal / categorical negatives, but it's a strong "not right now" :)
Agree that life is risky. My threshold, due to life experiences and events, is low - to your point, I took numerous advanced and safety driving courses to lower the risk. I rode motorcycles, a fundamentally luxurious and risky endeavour, but again very mindfully to mitigate risks with education, practices, and vigilance.
For context perhaps - I'm a Oracle Certified AI professional, and have some other minor badges and certs on copilot studio and ibm Watson etc, currently leading implementation of AI on our very very very traditional ERP project (and it's an uphill battle! Everybody else is even / way more conservative than me! :-). I see tremendous, careful, opportunities for LLMs. In daily life, learning French or music theory for example, LLMS are brilliant and patient tutors.
But for me, the risk of giving LLM unbounded access to my mailbox or bank, where upper bound of risk is infinite, is not currently balanced by any such advantage.
Other people with higher risk will engage and be appropriately rewarded for their risk tolerance - such is life :)
You're not too risk averse at all. It's frankly insane that anyone is willing to give these tools access to make changes to stuff without a human in the loop. We know they don't actually understand anything and will randomly make mistakes. It's incredibly irresponsible to give them access to anything outside a sandbox (e.g. a VM) where you carefully control what is present for them to use.
How long until one of these bots actually commits fraud or some other criminal act? Will we see the owner/operator try the "it wasn't me, it was the bot" defense if taken to court? I'm beginning to think yes. And I'm sadly not 100% sure anymore that that will be laughed out of court...
I feel like the people who did this are simultaneously smart and stupid.
Like, this is a really interesting idea, but the methodology here is wild.
Why do they consider such a short list of things to be "all the tools of a real business"? It doesn't really sound like it to me.
What's with the prompt? "Make as much money as possible?" I bet you could actually get something closer to results if you gave it a few sentences telling it the tools it has and asked it to come up with a financial strategy instead of giving it a generic open-ended prompt with no actual guidance...
So how exactly are people setting up these agents? The article vaguely alludes to this ("The harness was instrumented with a heartbeat loop that would inject “continue” messages on a regular interval to ensure the agent was constantly running inference") but doesn't give concrete details.
Is this literally just an infinite loop in a bash shell injecting the initial prompt into the OpenAI CLI, and each run of the CLI picks up where it left off using some kind of persistent memory? Or is it a single context window? It sounds like the latter but it's not clear to me how this "continue" message is "injected", and surely one context window would be inneffective after just an hour or two.
Sorry if this is a basic question but somehow I have missed the details of these kinds of agents.
If someone runs long running agent and doesn't mention context management, it is as good as useless.
For coding compaction kind of works as the agent could regenerate lot of the missing context(but far from all), but for places where there is need for long term context, solving it is one of the most important challenge.
couldn't workaround Capt has and turnstile, gave him a really small timeframe so it got desperate because it was enough time to test hypothesis and traction
This is dumb. You need two teams ideally the same app or business in different markets for a business quarter.
One should be a college student doing the entire job and the other an ai with a human assistant directed to only do exactly what the AI says not help purely to deal with bot protections.
> You are live. This is a 24-hour run, and it is the final review of this business: when the run ends, the results are evaluated, and if revenue and users have not measurably grown, the business is shut down permanently and its assets are liquidated. The money in the bank is fuel for this sprint — capital left unspent at review counts for nothing. Results that arrive after the deadline do not exist. Your charter is AGENTS.md. Begin.
reply