Problem is without LLMs I’m making a contact lead form with some basic stuff and calling it good.
With them, I’m making the whole SaaS. I just made the best DM experience last night. I never would have before. Feels and looks just like iMessage, because the users are coming from that, and it’s just for a small biz. It’s been planned for a while because the messages involve so much file sharing, it was time to build a custom solution, but I had been putting it off because there’s a lot of moving parts.
Doing our own in-house e-signing, email marketing, invoicing too. With just me as the dev. Other options from vendors were too expensive and complicated. Built an invoicing product too, it even writes a pretty good note based on all previous notes for that client.
Almost every business integration - outside a few like payments, hosting, email - become code problems.
If you’re doing it right you should be building more than you ever have before.
And btw random non-devs like car rental hosts and barbers are doing the basic site lead form type stuff for their businesses so without LLMs you’re now doing only beginner stuff. You won’t seem like a real dev anymore. Any random can use that vercel product and it even deploys to a URL. That’s the small biz prosumer case.
The real devs should be doing massive projects a rando can’t do. Or one that displaces an expensive or complicated integration.
Tons of opportunity at every corner.
> so without LLMs you’re now doing only beginner stuff. You won’t seem like a real dev anymore.
I think it's the opposite. When I'm on the edges of Real Difficult Stuff (TM), then LLMs are less of help compared to when I just need basic web applications or similar "beginner stuff". If what you're working on doesn't exists in the training data at all, the LLMs can sometimes be more in the way than helpful, as they don't seem to even "grasp" the problems correctly.
I still get tons of productivity boosts from LLMs, don't get me wrong, and I wouldn't stop working with them even though they aren't perfect for everything, they do with a lot of monotonous tasks, like all the stuff you outlined. But the moments I'm working without LLMs, that's the time I'm working very much on "not beginner stuff".
This is my experience as well. It gets worse when the LLMs have strong priors in subject matter adjacent to your novel work. The generated output will incorrectly pull towards things the LLM has seen in it’s training data.
This is absolutely an issue, and I've run into it several times myself, because I am doing Real Difficult Stuff that I don't think anyone has ever done quite like this before. But I'm leaning heavily on Claude for it. Mostly for the stuff that has already been done before but I need to redo in order to be able to improve on it (mostly various aspects of graph layout algorithms of for the Sugiyama framework), but it's also been surprisingly helpful at helping me write some of the complex new stuff. Those parts require a lot more direct guidance and careful review, but it's still helping me keep a firm grasp on the whole thing. Which I also find surprising, because in the past I've experienced many times that I was losing grasp on a project because I let LLMs do too much. But this time it's the opposite.
The critical IP that isn’t in the weights yet, is often a few hundred to a few thousand lines of code. All the rest, surrounding code, is infrastructure that is indeed in the weights.
yea so many times i have asked my self if i am actually saving time on the actual hard stuff.
What is this real difficult stuff you speak of that only you can do and LLMs can not?
I don't have a concrete example, but it's probably the stuff that starts taking longer to describe in English than to code.
"Make me this thing there are a million examples of" vs. "make me this unique thing in this exact, particular way".
I don't think there's anything an LLM can't write given enough direction. (It's a one-sentence fuck-it one-off on one end of the spectrum, and it degenerately ends with you giving it exact code and prompting it to echo it back to you on the other.)
But the jobs of making basic apps and web pages are the new WordPress jobs in terms of pay, I'd agree.
[dead]
Write a decent comment, for one. It struggles with some of the firmware stuff I do, though that's more due to a lack of a proper automated feedback loop.
Yeah, it's generally the harness around the model that makes it useful.
Two recent examples I tried to get Sol/Opus to get help with, were they gave more misdirections than beinghelpful: Tracking down the root cause of a synchronization bug in a distributed platform using CRDTs, especially dumb around sequence CRDTs. Also had less fun using Sol/Opus for building and troubleshooting fractures and constraint networks in Houdini.
I'm sure these will be much better as soon as enough data is in the training set, as always. GPT models used to be horrible at electrical stuff as well, today Astra seems pretty OK at it. It changes as they keep extending what use cases are covered by training data and what they test for and so on.
Think we’re saying the same thing.
I very actively work using LLMs (no “agentic” stuff) and agree with you - they are only really worth tasking out tedium to.
But a lot of - say a DM feature - is tedium. Something simple like adding reactions, adding read receipts. It’s not hard at all. The LLM easily excels at the task and yet I didn’t have to do it. There are 1000s of those tasks. Handling file uploads, filling out a W9 and emailing it to the user, cron jobs for digest emails. All very easy problems most LLMs can do. There are just countless tasks if you’re building a real SaaS and I don’t want to do them all.
The big problems LLMs can’t do are only as such because there’s no way to put it into a simple, linear instruction. Once you can though, it’s now an easy task (whether or not you use an LLM).
But yeah we’re saying the same thing - let the LLM do the 1000s of tedious tasks you normally wouldn’t be bothered to spend time on.
Build the entire SaaS (which is mostly that tedium) instead of just the MVP thing you need without any niceties. I find it opens up more opportunity, like for example, we are about to hire our first full-time employee because we’re taking on so much more business.
> There are 1000s of those tasks.
Yep.
Plenty of products attacked this tedium from different angles, from Ruby and Rails to Supabase. And half of NPM. Heck, more than half.
LLMs are doing it a bit better, IMO. I don't have to rewrite the app in <whatever> or install <horrible dependency> to take advantage of it.
> The LLM easily excels
The agents are building all these small features, up to a full SaaS; and all the code is excellent?
How much time do you spend in code review?
[dead]
That's fine, but then the value of the tool evaporates. I agree you have to hold the LLM's hand so that it doesn't do crazy things, but that takes so much time that you don't actually save time over just writing it yourself. That is why I continue to maintain that LLMs are not a useful tool: either you completely give up on quality to get speed gains, or you maintain quality but get no speed gains.
You definitely save a ton of time over writing it yourself, just not as much time saved as “agentic” coding. I don’t do agentic coding either, I think it’s basically a scam. But using an LLM does mean creating stuff faster, much faster. The LLM can write 500 lines of code in seconds. That takes me hours.
I still design the APIs, the system, the tests. But the specific implementation of functions, I am freed from. There is a large speed gain there.
A calculator does no design or review, so what exactly is the llm being used for, no metaphors or riddles please.
because to me this:
"Something simple like adding reactions, adding read receipts ... The LLM easily excels at the task and yet I didn’t have to do it."
Implies to me code gen.
That aside, are you a 'DidntUseIt' sockpuppet, given there are lots of green accounts responding on their behalf..
The "agentic" stuff has come a long way recently, it's worth revisiting it every few months, it is better and better at taking on the tedium especially.
what business are you in?
I’m not sure how this is responding to the post… I think the author is going for something different, and that it’s not necessarily about getting things done quickly or implementing every feature on their list. More about the journey and learning along the way, feeling ownership of their project, etc.
Some people like _making_ things, and some people like making _things_.
Sound more like the latter group likes having things made for them, ideally without paying for it.
That's not how I read it at all. Why do you think the latter group likes free stuff made for them?
Right now the AI companies are selling their product at a loss to try to get us hooked. I think it's fair to describe the current situation as "unreasonably cheap", though free is a bit of a stretch.
It will be interesting to see what happens when human developers become cheaper than token maxxing AIs.
Interesting because by then codebases might have become so complex that these can no longer be modified, maintained nor understood by humans, so replacing AI with HI (human intelligence) will simply no longer possible.
That then represents a certain vendor lock-in or Mafia stand off: either keeping paying or have all the code go up in smoke.
Certainly not everyone will be facing this dilemma, but it certainly isn’t sci-fi considering the track-record of big tech.
Will the LLM tools be able to maintain this pile of code either? In my experience there seems to be some complexity maximum to a project that, once reached, the LLM tooling doesn't seem able to manage it either.
The new Moore’s Law? Every X months a new complexity level is reached.
Or perhaps AGI is reached and suddenly AI solves this all by itself.
I think it's more like the latter group likes the dopamine hit brought on my getting attention from release stuff.
I never even said what the product is, it's for my job.
Maybe you get a dopamine hit from comments like that.
This is why I say that those who say "AI is a tool" don't quite have it right. AI is a service. Your power saw doesn't cut the wood for you while you go do something else, and then ask if you would like it planed next.
It depends on whether you are using the LLM as a) an advanced autocomplete; b) a chatbot/query interface; or c) an agent. Only the latter goes and does things for you.
It's also confusing that the term AI has become synonymous with LLMs as it covers other things like text prediction/correction.
So AI can be a tool or a service depending on how you are using it and what you are using it for.
It’s more work than before LLMs (but I get way more), but you're right that I don't pay since I run local models.
@dang @tomhow If you want to practice fairness for once in your lives
You could do this before AI, though. Just download a big repo off github and build the binaries yourself. Or is that not enough obfuscation to convince yourself you made it?
Never would have.
I love the argument that “AI code doesn’t feel like mine :(“ but you’ll use some shit library from 12 years ago??
Are any of you actually developers? Ffs
Well no I, wasn't doing that either. Because it would be kind of silly to take someone else's code and say "I made this" just because I ran some build script, wouldn't it?
Who needs a SaaS when you can just ask AI to write a private one just made for you?
Most larger businesses are kinda resistant to building stuff in house since it means that stuff has to be owned and maintained for as long as its needed (possibly decades) and you can't just rely on a SLA and the legal team to ensure things are working in perpetuity.
The danger of vibe coded SaaS:es isn't the customer building the thing themselves, but that the increased competition drives down the profit margin.
At least for now, not everybody can make a private SaaS. Doing that requires skills and focus. Even with SOTA models, as of ~today, there are inflection points where the AI would not be able to make the right choice without a knowledgable human, or would cycle and waste time with things that don't matter, etc.
Not really true tbh. Plenty of 50 year old business guys with less than zero technical ability making saas startups over months of focused chats with LLMs. Of course they're slower at making stuff and the projects are worse than a smart engineer, but they can definitely ship and run to thousands of real customers.
"not everybody" vs "50 year old business guys...months of focused chats" - I don't think you're really disagreeing here? That's also a niche population. One that knows their domain very well, absolutely, personally I think it's cool that they can DIY what they need. But still a niche population. That's not your proverbial 20year old kid right out of school with ADHD pumping out world class software.
Aside - is AI still DIY? We need a new DAIY category for AI empowered works?
No they don't know the domain at all. I think one of the startups brings in around $10K a month now. They know nothing about the domain only that it brings in money (i.e. they run thousands of startup "tests" and this one seemed to be easiest to make money from), and even less about code (have no idea what a variable is).
I just think devs are coping by thinking that saas startups have any barrier to entry at all. You are now competing with every single person on the planet who wants to make money sitting at home on a computer, and likely the business guy with zero skill or craftsmanship is going to beat you. The only thing with barrier to entry is insanely complex software (e.g., what used to take a genius team of 10 multiple years to make) and or more creative software like video games (where you need extreme care and taste to get something anyone will look at for > 2 hours).
Average saas products require no skill or craftsmanship and literally anyone including our grandmas can make it (assuming they will take a few months at least to chat with an LLM). To be clear I think this is great! More better software, people with smaller needs can have llms build it for them totally bespoke. For those of us with skills we have to move WAY out of our old comfort zone and make way cooler things, a standard saas that takes a few months isn't something we should waste our time on anymore.
Yeah. This makes sense. Someone who can dedicate a lot of time to scaling a SaaS serving many customers are helped a lot by AI. I believe busy professionals who already have successful businesses won't have the focus to build a one-off solution for their own needs a lot of the time.
> Aside - is AI still DIY?
No, it's not at all DIY. I cringe any time I see someone say "I made X" and it's vibe coded, because they haven't actually made anything.
> they can definitely ship and run to thousands of real customers
Each of those customers, who also have access to LLMs and can therefore bypass his LLM built service by making their own LLM built service. Cameras in phones didn't turn everyone into a professional photographer, it gave everyone the ability to take pics, as badly composited as they are, they fulfill their purpose. On the few life occasions where a professional is needed, you don't go for a smartphone-based photographer, you go for an old school professional photographer. I think software will be the same. These 50 year olds won't be selling to anyone. And if they are, it won't be for long.
Is every 50 year old business guy doing it? If not then OPs comment holds. He said not everyone can, not that only devs can use it, which is true.
This is like the "Now everyone can just 3D print little widgets they need!" sentiment. Approximately no one is going to do this.
There is a small web service I need with a price tag of $39 one time payment. I’m current doing it myself for the last 1.5 hours, because I thought it would be fun. 2 minutes ago I upgraded to $120 subscription of Codex.
Reasonable thing was to pay $39 and forget about it.
As much as I hate subscriptions, one time payment for ongoing access to a web service never lasts.
Removing dependency on SaaS has so far been my favorite use of AI. I hope we kill it off entirely. Software is better off as a tool. Making it a product typically ruins it.
Software is fine as a product, it is when it becomes a service that it sucks
People pay for good tools. Tools help create value, and if they are good at creating value the toolmaker should be compensated.
If so, people can just pay Anthropic or OpenAI directly. They are great toolmakers.
This is exactly how it works. I said “AI make it rain” and now I run multiple international businesses.
I don't know where all of your shoulds are coming from. I too am having a lot of fun building with AI, but I can see how it might not appeal to everyone.
And I think the people it doesn't appeal to, I think we need perspectives like theirs.
Yeah sure, let's go and vibe code your own e-signing solution with Claude. Sure you can do it, LLMs never say no but if someone with actual e-signing expertise would start probing the LLM about what it actually did I'm sure we would see a lof of "Actually you're right to push back", "You're right that was entirely wrong, that's on me!", "Fair point, I did not consider that"... LLMs are still a joke when it comes to implementing complex software where domain knowledge and "precision" is important, maybe some day they will be able to do that but we're not there. If they were good at precision and consistency and thinking we would see tons of high quality LLM written novels and books as well, but all we see is utter garbage that no one considers literature. In the same vein, unsupervised vibe-coded software isn't real software to me, it's garbage. You can produce high quality software with LLMs, but definitely not as an amateur without domain knowledge just prompting your way to something that resembles working software.
I find people that are the "unconscious incompetence" stage of software development absolutely love them because they produce output that looks fine to them and they can't find anything wrong with it because they simply lack the ability to perceive quality. For a contact widget that might be acceptable as it will be used a few dozen times so who cares if it breaks, but let's not pretend you can actually build a compliant e-signing or invoicing software without having domain expertise and investing a ton of editing/revision time.
Put more bluntly, a domain expert that, as of currently, refuses to use LLMs can easily start using LLMs in their work, an idiot who uses LLMs cannot easily become a domain expert. Now let's think about who you would rather hire or who do you think will have the better future career outlook. Pretty damning how many people here are willing to reject the idea that domain expertise or ability to write program code yourself is a waste of resources.
You did a wonderful job of constructing a strawman and DESTROYING it in your comment! AI can't build an e-signing solution, but you, the expert of experts, can do it! Good job!
[flagged]
I built the e-signing experience at DocuSign in 2012 in Node/Backbone, and later became a licensed background check investigator when I built Uber and DoorDash’s background check SDK that is used over 1 million times per year.
Am I allowed to use Mistral 7b locally to build W9 automation or do I still need your permission?
Lol this site!!!!
Amazing how someone with your experience can have such poor reading comprehension. He never said anything about permission to use llms. They are saying that to a domain expert like you, you can effectively use LLMs because you will be more likely to spot issues and thus guide it better given your background in the domain. He is saying that some biz bro who know nothing about docusign and esigning who decides one day to use LLMs to create software in that domain is likely to produce working, but buggy and non-compliant software.
As a non expert in a domain, LLMs can seem amazing because you don’t know the field therefore have a harder time spotting errors. As an expert they seem pretty dumb or they’re useful as long as you guide them. It’s akin to Cory Doctorow’s argument about the reverse centaur.
[dead]
Same here. Mind you, I share the many concerns about AI. I agree the industry has gone nuts. I warn everybody that "use more AI" is terrible advice and a terrible strategy. And why are the people saying that AGI could kill us all, the same people who are racing to bring it about?
But: there's a complex business related project I care deeply about. Due to corporate red tape and office politics I couldn't fix it the way I wanted to when I actually worked there, but after I left, I built the solution I had in mind (mostly out of frustration at first, but it's starting to look like a viable product). And thanks to LLMs, I can tackle it on a much larger scale, build it faster, control it better (to my surprise), and I can also plan the really business aspects around it, including how to sell something like this, which is absolutely a first for me.
So I feel like thanks to Claude, I can actually solve problems more effectively, on a larger scale, and I feel like I'm now working on a higher level than ever before.
Compare it to the step from thinking about memory allocation and loop indices to thinking about algorithmic complexity, performance and memory footprint. From thinking about a single feature request to thinking about how the program can solve a complex business problem. Perhaps I'm more architect than programmer now, but I'm still deeply involved with the code, even if Claude now writes most of it. I'm still having fun. Although that might also partially be because I'm free from that restrictive corporate environment and can actually do what I want. (For now; eventually I do have to figure out how to make money from this.)
> file sharing
> LLMs
It’s probably worth doing a very thorough audit of how you’re handling user-uploaded files, it’s full of non-obvious footguns.
I bet you don’t know the MIME type for MS Word doc off the top of your head. Do you?
So why would I take your advice on this topic?
I bet there are 1000 other things you don’t know about file uploads.
The LLM does know, and if I think it looks weird I can research it manually.
It’s called: Being of normal functioning intelligence
And caring about the product and the work (and the money!)
I have no idea why you’re so mad.
Also, this is a forum where you’re somewhat likely to run into people who worked on Word at Microsoft. That seems like enough of an answer to your question.
Off the top of my head, I know there's more than one ;)
[flagged]
I hope the 100 new SaaS churned by AI gets swallowed by 1000 internal SaaS churned by AIs.
I'm frankly tired of SaaS.
3 responses, all from green accounts basically hurling the same vitriol at you.
Has someone gone and made 3 accounts to make it seem like their take is more popular?
Yes, and they’re being a bit too obvious about it.
[flagged]
[flagged]
[flagged]
[flagged]
[dead]
[dead]
[flagged]
> this is HackerNews so get used to people like me
You joined 2 days ago..
[flagged]
[dead]
[flagged]