idopmstuff

I have always found writing documentation to be incredibly helpful for clarifying my thinking. It prevents me from doing mental hand-waving around details, and often times writing down a process that I have done a thousand times is the thing that makes me realize how I can cut steps or improve it.

I'm now in the process of trying to hand off chunks of the work I do to run my business to AI (both to save time but also just as my very broad, practical eval). It really is all about documentation. I buy small e-commerce brands, and they're simple enough that current SOTA models have more than enough intelligence to take a first pass at listings + financials to determine whether I should take a call with the seller. To make that work, though, I've got a prompt that's currently at six pages that is just every single thing I look when evaluating a business codified.

Using that has really convinced me that people are overrating the importance of intelligence in LLMs in terms of driving real economic value. Most work is like my evaluations - it requires intelligence, but there's a ceiling to how much you need. Someone with 150 IQ points wouldn't do any better at this task than someone with 100 IQ points.

Instead, I think what's going to drive actual change is the scaffolding that lets LLMs take on increasing numbers of tasks. My big issue right now is that I have to go to the listing page for a business that's for sale, screenshot the page, download the files, upload that all to ChatGPT and then give it the prompt. I'm still waiting for a web browsing agent that can handle all of that for me, so I can automate the full flow and just get an analysis of each listing sent to me without having to do anything.

Yeah, I've done it with industry-specific acronyms and this works well. Generate a list of company names and other terms it gets wrong, and give it definitions and any other useful context. For industry jargon, example sentences are good, but that's probably not relevant for company names.

Feed it that list and the transcript along with a simple prompt along the lines of "Attached is a transcript of a conversation created from an audio file. The model doing the transcription has trouble with company names/industry terms/acronyms/whatever else and will have made errors with those. I have also attached a list of company names/etc. that may have been spoken in the transcribed audio. Please review the transcription, and output a corrected version, along with a list of all corrections that you made. The list of corrections should include the original version of the word that you fixed, what you updated it to, and where it is in the document." If it's getting things wrong, you can also ask it to give an explanation of why it made each change that it did and use that to iterate on your prompt and the context you're giving it with your list of words.

I buy and operate small e-commerce brands, and since GPT-3.5, I've been attempting to vibe code software to help me manage the business. With GPT-5 Codex I have finally managed to create something legitimately useful for myself. The code may be (almost certainly is) not of great quality, but for the purposes of an internal application that only I use, it's doing the job just fine.

The statement "in all forms of gambling the gamblers make a net loss" is ambiguous as to whether it's the gamblers as a whole or each of the gamblers making a net loss, and while I grant that it reads more like the way you're describing, the later context pushes the intended meaning in the other direction.

He says "The 'house' is the only winner," which is as a point of fact untrue - there are also individual gamblers who are winners. Saying this implies he thinks that each of the gamblers is a net loser.

I honestly think we had a pretty good middle ground with people having to go to Vegas/Reno/AC to gamble. It can be fun, but you have to go out of your way to do it. If you have a fun Vegas weekend and blow a bunch of money once in a while, that seems pretty okay relative to being able to constantly bet on anything from your phone.

> Let's start with the obvious- in all forms of gambling the gamblers make a net loss. The games are hosted by very sophisticated companies, that have better mathematicians, and make money.

> $x is pumped into the system by the punters, $y is extracted, $z is returned. The 'house' is the only winner.

This is incorrect, specifically with regard to sports betting. Sports betting and poker are both winnable games. Most people don't win in the long run, but unlike in table games (Blackjack, etc.) there are absolutely winners that are not the house.

To be clear, that doesn't mean they're good or should be allowed. I used to be a poker player and enjoy putting some bets on football now, but I've come around to the general idea that sports betting in particular is a net negative for society. Still, if you're going to make an argument against it, it's always going to be a better argument if it isn't built on a basis that's just factually untrue.

> Plus penetration is already very high in the areas where they are objectively useful: programming, customer care etc.

Is that true? BLS estimates of customer service reps in the US is 2.8M (https://www.bls.gov/oes/2023/may/oes434051.htm), and while I'll grant that's from 2023, I would wager a lot that the number is still above 2M. Similarly, the overwhelming majority of software developers haven't lost their jobs to AI.

A sufficiently advanced LLM will be able to replace most, if not all of those people. Penetration into those areas is very low right now relative to where it could be.

It really depends what you're coding. I'm a former PM who can't code, and I'm currently using GPT-5 in Cursor to write an internal application for my business of acquiring and operating e-commerce brands that sell on Amazon. I'm really vibe coding (just clicking accept without ever reading any of the code changes), and it's working great! I've got a whole dashboard that's retrieving inventory and sales data from multiple Amazon stores, projecting future sales and reorder point, etc. The code might be total dogshit, but it's incredibly useful for me (and also really enjoyable).

That said I might be sort of a weird case, since I am accustomed to designing and documenting product requirements, the fact that it's just for me means architecture doesn't really matter, and I'm competent enough to help it resolve some design problems like the fact that it was obviously hitting the wrong Amazon API that was pulling a report of every single sale as a line item rather than just a report with total sales numbers, etc.

I don't think you even have to go to the extreme of compelling someone to do something through the legal system. I can use money to get someone to clean my house or walk my dog - they don't have to do it, but the fact that I have money does give me the power to get them to do what I want.

People look at all the chaos in their AI lab but ignore the fact that they yet again beat on earnings substantially and directly cited better ad targeting as the reason for that. Building an LLM is nice for them, but applying AI to their core business is what really matters financially, and that seems like it's going just fine.

I'm devastated to say that when I just searched for the Maddox quote "for every animal you don't eat, I'll eat three," the page was indexed in Google but not live when I clicked on it. RIP.

> What could have been if instead of spending so much energy and resources on developing “AI features” we focused on making our existing technology better?

The implied answer to this question really just misunderstands the tradeoffs of the world. We had plenty of money and effort going into our technology before AI, and we got... B2B SaaS, mostly.

I don't disagree that the world would be better off if all of the money going into so many things (SaaS, crypto, social media, AI, etc.) was better allocated to things that made the world better, but in order for that to happen, we would have to be in a very different system of resource allocation than capitalism. The issue there is that capitalism has been absolutely core to the many, many advances in technology that have been hugely beneficial to society, and you if you want to allocate resources differently than the way capitalism does, you lose all of those benefits and probably end up worse off as a result (see the many failures of communism).

> So I ask: Why is adding AI the priority here? What could have been if the investment went into making these apps better?

> I’m not naive. What motivates people to include AI everywhere is the promise of profit. What motivates most AI startups or initiatives is just that. A promise.

I would honestly call this more arrogant than naive. Doesn't sound like OP has worked at any of the companies that make these apps, but he feels comfortable coming in here and presuming to know why they haven't spent their resources working on the things he thinks are most important.

He's saying that they're not fixing issues with core functionality but instead implementing AI because they want to make profit, but generally the sorts of very severe issues with core functionality that he's describing are pretty damaging to the revenue prospects of a company. I don't know if those issues are much less severe than he's describing or if there's something else going on with prioritization. I don't know if the whole AI implementation was competitive with fixing those - maybe it was just an intern given a project, and that's why it sucks.

I have no idea why they've prioritized the things they have, and neither does the author. But just deciding that they're not fixing the right things because they implemented an AI feature that he doesn't like is not a particularly valid leap of logic.

> Tech executives are robbing every investor blind.

They are not. Again, guy with a blog here is deciding that he knows more than the investors about the things they're investing in. Come on. The investors want AI! Whether that's right or wrong, it's ridiculous to suggest they're being robbed blind.

> Unfortunately, people making decisions (if there are any) only chase ghosts and short term profits. They don’t think that they are crippling their companies and dooming their long term profitability.

If there are any? Again, come on. And chasing short term profits? That is obviously and demonstrably incorrect - in the short term, Meta, Anthropic, OpenAI and everybody else is losing money on AI. In the long term, I'm going to trust that Mark Zuckerberg and Sam Altman, whether you like them or hate them, have a whole lot better idea of whether or not they're going to be profitable in the long term than the author.

This reads like somebody who's mad that the things he wants to be funded aren't being funded and is blaming it on the big technology of the day then trying to back into a justification for that blame.

The title of the post indicates that most people who buy games on Steam never play a single game that they have purchased (vs. your correct understanding of the data, which is that people buy games and play some but not all of them).

I played online poker successfully for years back when it was booming.

There are definitely online players who lack the skills you're describing, but that's not as much of a problem as you think. You can hide tells just by shutting up and staying still while you're in the action.

The other half of that is reading other people's tells, and online poker is more helpful there than you'd think. Most of reading other people (especially at relatively low-mid levels) is about reading the story they're telling with their action rather than reading their face/words/etc.

Classic example is: There are two hearts on the board on the flop and the person calls your bet. Turn comes, not a heart, person calls again. River comes, not a heart, person suddenly bets big to try to get you to fold, because they had two hearts and failed to make their flush.

Bigger picture, you read their style of play. Are they playing a lot of hands or very few? Passive or very active? None of these things require reading the person's mannerisms, and you can practice all of them very well online (though online you also run tracking software that gives you stats on opponents, which helps when you're playing a bunch of tables at a time).

Writing this out makes me miss online poker. Shame the games are terrible now (and I also have a child and business as opposed to the endless free time of my twenties, to be fair).

"A successful aging treatment would be something that:

prevents diseases of aging, ideally more than one;

preserves a healthy function that normally declines with age (like fertility, immune function, cognitive function, resilience, or physical fitness); or

reverses the course of at least one age-related disease."

I think a lot of the anti-aging companies out there would say that the real answer is a combination of the second and third - reversing the course of age-related decline.

Also, I think it's sort of contradictory to have two of these points focus on diseases of aging but in a subsequent section say that oncology isn't anti-aging. Cancer is in many ways a disease of aging (it's very clear from the numbers that increasing in age causes increases in likelihood of developing cancer, generally more than any other single factor). Curing cancer obviously isn't going to get you a general-purpose anti-aging treatment, but that's why it seems odd to say that reversing the course of an age-related disease is a successful aging treatment.

Really says something about the HN crowd that you're getting downvoted for this.

I'd give them to AI because they're generally just not getting done. I worked hard to get that frontend dev time to make those improvements, but there was no chance it was ever going to be enough. When you're talking about enterprise software, minor improvements to pageload speed do not move the needle on revenue. When you have a list of features that customers will actually pay for, those will get priority 100% of the time.

Everybody's job is to serve the company priorities. Engineers don't get to pick the tasks they want to do because they're getting paid to be there. I also have spent lots of time doing things I'd rather not do, because that's the nature of a job (plus a pile of stock options incentivizes me).

Better to have those tasks done by AI than not at all.

Thankfully my career has been at very early stage startups, so none of that applies!

As a PM I have never not had a backlog of little stuff we'd love to do but can't justify prioritizing. I've also almost always had developers who want to make improvements to the codebase that don't get prioritized because we need new features.

The upside is that both of these things are the kind of tasks that are probably good to give to AI. I've always got little UI bugs that bother me every time I use our application but don't actually break anything and thus won't impact revenue and never get done.

I had a frontend engineer, who, when I could just find a way to give him time to do whatever he wanted, would just constantly make little improvements that would incrementally speed up pageload.

Both of those cases feel like places where AI probably gets the job done.

Good for her - I'm just astounded that I didn't know any of this. Even the fact that generics only have to be within something like 20% of the active ingredient of the non-generic was surprising to me; I thought they were basically identical.

But the quality of manufacturing stuff is just unbelievable. You take for granted that our medications are highly regulated and it turns out to not be remotely true...

For more on this, I highly recommend Bottle of Lies by Katherine Eban. Fantastically written book that goes deep on this topic and will absolutely horrify you about everything related to the manufacture of generic drugs.

So you would do what - have engineers hired as part of individual government departments instead of USDS? How would that stop them from being coopted by future administrations? Anybody employed by the US government can't be insulated from changes in who runs the government.

Since you're serious and I'm not, can you propose any specific solution that avoids this issue?

So you think we shouldn't have functional digital infrastructure and information storage because something bad could be done with it in the future? Seems like you could apply that to anything.

patio11 says you just give e-commerce companies the code: https://x.com/patio11/status/1927230790141616578

I worked as a PM for a startup that makes software to help review appraisals, and there's absolutely no question that stuff gets missed by humans in appraisal review as well as the rest of the underwriting process.

Every organization is built around tradeoffs between time (which is very much money here) and quality. You route appraisals to people of different skill levels depending on the likelihood of issues and the likelihood that those issues matter. With appraisals, for example, if buyer's putting down 40% where only 20% is required, you check only the super critical stuff, because even if you're meaningfully off on the appraised value it doesn't actually matter that much.

There are a ton of these tradeoffs of whether you should review something as well as whether you should have your most senior people review it. On top of that, reviewers are judged on speed, so naturally they're making tradeoffs of accuracy vs. time themselves on each review.

Using AI means you're able to review things that wouldn't even have been looked at otherwise, because there's no longer a tradeoff of time/cost vs. accuracy. Even the world's best reviewer can't catch things they're not looking at.

I think the list of historical analogies is missing the biggest one - the internet.

Memorization used to be a much more important skill than it is today. I am probably worse at rote memorization than I was when I was 13. Am I dumber? I would say no - I've just adapted to the fact that memorization is much less important in a world where I have access to basically the entire recorded history of human knowledge anywhere, anytime.

LLMs are just another very powerful technology that changes what subdomains of intelligence matter. When you have an LLM that can write code better than any human being (and since I know I will get testy HN replies about how LLMs can't do that, I will clarify here that I mean this is a thing that is not true today but will be in the future), the skill that matters shifts from writing code to defining the problem or designing the product.

> Looking at historical examples, successful cases of offloading occurred because the skills are either easily contained (navigation) or we still know how to perform the tasks manually but simply don’t need to anymore (calculator). The difference this time is that intelligence cannot easily be confined.

This is true, but I think it just means we'll see a more extreme kind of the same change we've seen as we've created powerful new tools in the past. I think it's helpful to think of the tool less as intelligence and more as the products of intelligence that are relevant, like generating high quality code or doing financial analysis. You'll have tools that can do those things extremely well, and it'll be up to you to make use of them rather than worrying about the loss of now-irrelevant skills.

> Its no wonder that we now have a food system full of heavily processed foods and unpronounceable ingredients that may very well be doing harm to our overall health.

Sure, but in the old system people just starved to death when there were problems with their crops (Irish potato famine, dust bowl, etc.). The current system isn't perfect, obviously, but this example seems to pretty clearly demonstrate a case where it's better that we've outsourced this knowledge to others.

Also, it's worth bearing in mind that we're now at a point where basically all of the information that people have "lost" is now once again available on the internet. Most people don't use it, because there simply aren't enough hours in the day, but people who care can find out more than any farmer 100 years ago about food and source theirs accordingly.

I mean to be fair, I like that they're putting their money where their mouth is so to speak - if you want to sell a product based on the idea that AI can handle complex tasks, you should probably have AI doing what should be simple, frontline support.

I'm with you. I don't really understand the complaints, since Amazon's return policy means that you aren't really taking a risk when you buy, even from randomly-named brands.

I mean based on this website I'm gonna say feature not a bug