That's great, but Becky at the farmer's market doesn't want to be re-inventing notepad on her computer to sell sourdough.

This smells like "Self-checkouts", "perfect bespoke solution so that everyone can check their own groceries out, no need to wait for a slow/expensive _human_ to get the job done, except you end up basically just doing the cashiers job and waiting around inconveniently whenever you want to get some beer

And yet I'd much rather self checkout than have a cashier do it for me.

At Costco, they forced me to empty out my very fat cart and scan each item individually. That's when I realized the value of cashiers.

It feels like “virtual stuck behind a bus”, then doing the company a favor for the experience.

The only time, for me, this is true, is when it's empty and I have very few items. If there's a line, I'd rather wait for the cashier. Self-checkout seems to get clogged up with people waiting for the supervising human the busier it is, so to have a human or two (one ringing one bagging) when I've got a cart and there's a line, is advantageous.

I'm also pretty anti-social and actually prefer the robotic banter of quickly checking out with a person to the anxious nightmare of just trying to buy something real quick and now waiting because the machine hates life more than me.

Oh and many of the folks doing the bagging at some of the stores are disabled, and I dunno - I hope we're taking care of people in those jobs.

For ninety nine out of a hundred scenarios, I've never had to wait for a supervising human at all. Maybe it's where I shop, but generally speaking, I've almost never had a problem with people needing to see my checked out items.

It's a marginal improvement perhaps, but nothing to write home about. And for other things (e.g. booking travel) the self-service experience can be worse.

Not sure about that, I prefer to know all aspects of booking travel that I would not trust to another agent, human or otherwise. Therefore, I believe your statement applies to some people and not others.

And often you are still supervised by sometimes rude security guards who demand the receipts. So you have the worst of all worlds: You pay the supermarket, do the work and are treated like dirt.

On a tangent: what about self scanning? Here in Sweden that is the norm. For the last ten years I've brought my own bags, packed everything like I want it as I grab it - with a quick blip from the scanner - and then checkout is no line, no cashier, no unpacking and packing things again. Just scanning my card and going through a brief screen to check that the summary looks right.

Instead of a guard there's a scanner by the exit gate where you scan your ticket. In my case just a small stub since the actual ticket is sent digitally.

I think it works so smoothly, much better than before this system was introduced. My recent experience buying groceries in Belgium was very similar.

Here in sweden it is not "the norm" by any definition of "norm". Also it's incredibly slower and not "one blip away".

And the scanner to exit is terrible for fire safety and accessibility. Also I've seen it break with the result that you had to trigger the alarm to leave.

It does not work smoothly at all, but perhaps you live in a different sweden than I.

It works pretty well in the UK. The nice thing is, I can pack my bags as I shop, so I don't feel stressed when I get to the checkout.

I use this method of shopping whenever available.

In sweden you have to unpack and re-pack at the checkout. And if you have a backpack you can't place stuff inside it. No. You have to unpack, place everything on a scale and then place everything in your backpack again.

And for items like bread or so you need to navigate a number of menus to find it.

And of course you are asked if you are a member, if you'd like to become a member, if you want to do a donation, and a number of other useless questions.

That's just self checkout. Some places, like Ica, have self scanning which works so much better. You just scan the bar code as you pick things up - which is just in one motion as you put things in your bag. No need to unpack as you pay.

Yes you do need to be a member to use the system, but that's a very small price to pay for the convenience and speed of avoiding all lines and stress of the regular checkout.

Yeah there's like 15 places in sweden which have this, and it's "the norm"?

Sweden is a big country so I could definitely be wrong about how things look across it. However, in the greater Gothenburg area every Ica I've visited has it - that includes everything from big ones in the city and shopping malls to small ones in the suburbs 30km outside the city (where I live).

In the greater Göteborg area, no Coop, Willys, Lidl, has this. And some ICA have it. Casually, the one closer to your home, it seems.

Going from "a supermarket near my home" to "every single supermarket in sweden" is kinda of a big leap.

Try ICA in Olskroken or Godhemsgatan for example…

I Most supermarkets have self checkout, but even that isn't omnipresent (and it's usually slower).

British supermarkets mostly require the use of their app to self-scan. You need to either use your phone or use your app to start the scanner. Once you have done that it works OK. A bit error prone.

The other problem is that, like self checkout, it often requires human intervention (buying alcohol or high value items, for examples) still requires human intervention. This can require a long wait at the end. I once got so sick of waiting that I left my shopping and walked away and went somewhere else.

>That's great, but Becky at the farmer's market doesn't want to be re-inventing notepad on her computer to sell sourdough.

That's exactly why she asks an LLM to do it for her. A program like this would be <1k LOC, well within even the current LLM one-shot-working-app domain.

Right. Jevons Paradox.

The lower barrier to entry might mean the average quality of software goes WAY down. But this could be great if it means that there are 10x as many little apps that make one person happy by doing only as much as they need.

The above is also consistent with the quantity of excellent software going way up as well (just slower than the below-average stuff).

What the android/apple appstores have taught me compared to the old symbian appstore is that we don't need more apps, we need decent ones. The current appstores have more.

One of the advantages of FDroid is that although tiny compared to Play Store, it is much better curated.

Discovery definitely becomes more important in this world.

Fortunately AI helps with that too.

Why do you think it will not soon be biased as search is?

I think AI gives us the opportunity to write our own ranking algorithms. Of course most people will pick an off-the-shelf editorial perspective but I expect these to be much more configurable than the current state with a tiny number of algorithmic gatekeepers.

The cost of AI is training it. If you can't do it yourself (and you can't), you don't decide what the weights are.

You can steer an AI without training it. It’s called Prompt Engineering.

It doesn’t give you full control, but I believe it gives you enough control to do what I proposed for most consumers.

Prompt engineering is just the new google fu, but if google wants to ban something it won't be there and if they want to bump something it will be bumped.

IMO the problem with the current software landscape isn't the number of applications, it's the quality and availability of integrations. Everything is in it's own little kingdom, so the communication overhead is extreme and grows exponentially. And by overhead I mean human cognition - I need to remember what's in my notes, and then put that in my calendar, and then email it to myself or something, and on and on.

We have these really amazing computers with very high-quality software but still, so many processes are manual because we have too many hops from tool to tool. And each hop is a marginal increase in complexity, and an increase in risk of things going wrong.

I don't think AI will solve this necessarily, and it could make it worse. But, if we did want to improve it, AI can act as a sort of spongy API. It can make decisions in the face of uncertainty.

I imagine a world where I want to make a doctor's appointment and I don't open any apps at all, or pick up the phone. I tell my AI assistant, and it works it out, and I have all my records in one place. Maybe the AI assistant works over a phone line, like some sort of prehistoric amalgamation of old technology and new. Maybe the AI assistant doesn't even talk to a receptionist, but another AI assistant. Maybe it gives me spongy human information, but speaks to other's of it's kind in more defined interfaces and dialects. Maybe, in the future, there are no applications at all. Applications are, after all, a human construction. A device intended to replace labor, but a gross approximation of it.

>. They fall on their face trying to work in sprawling codebases.

>However, they are excellent at writing small bespoke programs

For programmers I predict a future of a million micro services.

Sprawling has always been an undesirable and unnecessary byproduct of growing code bases, but there's been no easy solution to that. I wonder if LLMs would perform better on a mono repo of many small services than one sprawling monolith.

Micro service systems are just huge sprawling code bases with more glue code. Calling something over a network instead of via a local function call is still calling something.

For years one of my favorite experiences (amusing given your username) was being on calls for incidents where they get the dev on for X thing and an exec goes "I thought we got rid of that" and a bunch of people sheepishly explain it wasn't really retired... it was repurposed as an API. I especially loved it when the "retirement" of the broken thing was the execs big achievement. (The comedic nuance often being the thing could have been retired for real, but they demanded a timeline that necessitated the "fake" retirement)

It still happens but it's not a favorite experience anymore. It's just a source of loathing for MBA culture.

The pure play MBA is the capitalist West's equivalent of the Soviet apparatchik:

https://en.wikipedia.org/wiki/Apparatchik

Been saying that for years. Private equity is perhaps analogous to the Politburo.

Another way to look at it is - why force the agent to grapple with the whole code base when they can rapidly standup many single purpose services instead?

A large code base with modules with clear interfaces accomplishes most of this and is more efficient. Often vastly more efficient. Compare the cost of a function call or an in process message queue with a network API call over http.

Cloud providers make more money the more they can get people to use inefficient designs with more moving parts to rack up more charges. Bonus if it also locks you into managed services. Double bonus if those are proprietary. Complexity benefits cloud hosts.

Microservices, like all patterns, sometimes make sense. Like all patterns they often get overused.

Yep all true. Im not really arguing a point so much as thinking out loud - I guess the core idea I'm chewing on is whether LLM assisted development is enough of a paradigm shift that things that were formerly bad ideas become good ideas - but perhaps this particular line of thought (microservices) is the wrong one.

Micro-services are part of a system, so to create anything meaningful with them your context window has to include knowledge about the rest of the system anyway. If your system is large and complex, that means a huge context window even if your code itself is small.

Yeah i mean either they can or they cant handle large code contexts. I guess the theory is that a more organized and concise expression of your code base (not necessarily the raw code) is easier for agents to digest.

The issue as I have observed it is that software is written to cover as many use cases for as many users as possible. Obviously executives want to sell their product to as many hands as possible.

But end users often only need a tiny fraction of what the software is fully capable of, leaving a situation where you need you need to purchase 100% to use 5%.

LLMs can break down this wall by offering the user the ability to just make a bespoke 5% utility. I personally have had enormous success doing this.

Yeah if code-gen gets to a point similar to image-gen, where a user can prompt a full micro utility, then we will see the dual erosion of more DIY and also lower barrier of entry to the industry. Still most software value today is not about the utility but the IP, the network, and convenience. I suppose convenience is the main pillar at stake here.

Do you have any proof whatsoever that non-programmers are using LLMs to write small bespoke apps for them successfully? Programming ecosystems are generally very unfriendly to this type of use case, with plenty of setup required to get something usable out even if you have the code available. Especially if you'd like to run the results on your phone.

Sure, an LLM might be able to guide you through the steps, and even help when you stumble, but you still have to follow a hundred little steps exactly with no intuition whether things will work out at the end. I very much doubt that this is something many people will even think to ask for, and then follow-through with.

Especially since the code quality of LLMs is nowhere near what you make it out to be. You'll still need to bear with them and go through a lot of trial and error to get anything running, especially if you have no knowledge of the terms of art, nor any clue of what might be going wrong if it goes wrong. If you've ever seen consumer bug reports, you might have some idea of the quality of feedback the LLM may be getting back if something is not working perfectly the first time. It's very likely to be closer to "the button you added is not working" than to "when I click the button, the application crashes" or "when I click the button, the application freezes for a few seconds and then pops up an error saying that input validation failed" or "the button is not showing up on the screen" [...]

> Do you have any proof whatsoever that non-programmers are using LLMs to write small bespoke apps for them successfully?

I’m radiologist, I’ve been paying for software that sped up my reporting like 200 usd per month. I’ve remade all the functionality I need in one evening with cursor and added some things that I’ve found missing from the original software.

My company is a non-tech company that is in manufacturing. I'm not a programmer and niether is anyone else here.

So far I have created 7 programs that are now used daily in production. One of them replaces a $1k/yr/usr CAD package, and another we used to bring in a contractor to write. The others a miscellaneous apps for automating/simplifying our in houses processes. None of the programs are more than 6k LOC.

> This is a huge blind-spot I keep seeing people miss. LLM "programmers" are way way more useful to regular people than they are to career programmers.

Kinda. They're good, and I like them, but I think of them like a power tool: just because you can buy an angle grinder or a plasma cutter from a discount supermarket, doesn't mean the tool is safe in the hands of an untrained amateur.

Someone using it to help fix up a spreadsheet? Probably fine. But you should at least be able to read the code to the level you don't get this deliberate bad example to illustrate the point:

  #!/usr/bin/python3
  total_sales = 0.0
  def add_sale(total_sales, amount):
    total_sales = total_sales + amount
    print("total_sales: " + str(total_sales))

Still useful, still sufficiently advanced technology that for most people it is (Clarketech) magic, but also still has some rough edges.