The "hate" is for apps that only pre-package the GPT with a pre-written prompt and don't do much else. And the reason is, I think, because simply making a prompt template is usually not a significant enough value-add except for the most untechnical users. Most users can solve their problems just chatting with the GPT directly.

To get to the first part of your question: no. Every SaaS I've worked on for instance has included months if not years of creating intellectual property (often from scratch) so that the user's are receiving significant value over what they could do themselves.

Jealousy aside, I think they are poor business models because there is little to no barrier to entry.

The critique is against GPT wrappers where the entire product is just a translation layer that makes API requests to e.g. OpenAI is that this isn't much of a moat: Not only can competitors trivially do the same, your LLM API provider can pull the rug from under you by rolling your business model into the next version of their chat product. This latter scenario isn't hypothetical, but has happened several times.

If you build something more substantial that also incorporates GPT calls (e.g. copilot), that's a different thing, and not typically what people refer to when they talk about GPT wrappers.

Aren't all programs just wrappers around kernel calls?

Really, it's all just wrappers around the strong nuclear force.

Funny to think the strong nuclear force is just one dude’s underfunded and barely maintained side project that might collapse some day

The important bits of most applications aren't in syscalls; the important bits of a SaaS app might well be a program it wraps. Or not, of course; sometimes an easy UI is the important bit.

Luckily, we’ve overloaded the work kernel, so yes—we can all agree that every program is a kernel calls. Every useful program is just a fancy wrapper around DGEMM kernels.

Ah yes, but in what order should you make those system calls?

Tapping with a hammer: $2. Knowing where to tap: $9,998.00

Not the kernel.. (insert ackchyually meme here)

I'm writing a book (https://opinionatedlaunch.com/) on building and running subscription-based mobile apps (so, SaaS) from a technical implementation perspective (as opposed to a marketing or "startup" book). The book's case study is an app that generates PDF certificates for webinars. I chose that idea because it's straightforward, so people can focus on the book's lessons (e.g., "You want Google login like in the app? Check page 15").

Generating PDFs is very easy. You can either call some other SaaS' API (free plan available) or roll your own using Puppeteer (this is what I use). BUT to make it a full-fledged SaaS, you need to do more work than just that I-can-build-over-the-weekend thing:

  * Signups and logins (email/password login, forgot password, Google login, Apple login if you use Google login on iPhone).
  * Onboarding screens for new users.
  * Show stats to make people think it's worth paying for your app.
  * Subscription: handling trials, integration with App/Play Store, and downgrades/upgrades.
  * Automatic sub-domain and custom domain tenancy (e.g. acme-inc.myapp.com).
  * Email: SPF, DKIM, DMARC, email templating, copy, compatibility with different email clients.
  * Push notification.
  * Android, iOS and Web versions of your app.
  * Integration with Zapier so your app won't be a silo.
  * Even things you probably don't think important: Store page with nice screenshots and good copywriting, a nice app icon that doesn't look like it's designed by a programmer.
  * Oh God, the website for your app, the dreaded marketing copy, a pricing page, knowledge base, and blog.
  * Chat bubble in the bottom right of your website and email routing for customer support.
  * Social media presence *shudder*. If you think it's as simple as creating account in X and setting reminders to post content, think again.

To answer your question: "Isn't all SaaS just wrappers?" I don't know if it's "just" wrappers, but it's a VERY thick wrapper :D

You used to be able to slap a UI on a DB and call it a day. When Salesforce as launched having a unified DB for all sales information with a UI for controls and reporting was revolutionary!

Today the world is very different. You need to do more than provide a system of record, you have to build a product that makes it easier for customers to accomplish some task.

There are still UIs on top of a DB. I've consulted for several fast-growing startups that import data from some antiquated system, report on it in different ways, and their UIs are just massive data tables. In many spaces, that's innovative. Keep in mind, there are still industries that haven't really been tech-enabled and are 20+ years behind.

Most SaaS is a UI in front of a SQL database with a few services built around it. Sometimes these services are quite minimal. Sometimes it really is just a UI for a database with not much else.

The fact that it earns so much shows that the thing people pay for, for the most part, is usability and convenience... as if Apple being the most valuable company in the world for a while didn't already illustrate that.

I agree, but I would actually say that this is a good thing. If a SaaS can effectively map a real-world problem or pain point and solve it in a simple way, the goal has been achieved. Sometimes, there’s no need to overengineer a product when it can leverage existing technologies in a smart way.

The problem is that all the profit goes to the people who own the UI and the hosting, while the stuff solving the deeper problems is often completely unsupported.

I'm not saying the people taking things the last mile to the customer and making a good UI/UX should not make money, just that the people who built the stuff they're building on should also make money.

Software has turned into an industry where that largely is not the case.

Agreed. It's also true for most social platforms.

It's because they add almost zero value over just calling the APIs yourself or using an open source electron app that does. That doesn't mean they can't provide value to the average Joe, but for the technically inclined like HN's (or even Reddit's) audience they just look like cynical cash grabs.

If you take a look at products like the Cursor IDE, they're also "just" wrappers around OpenAI and Anthropic API calls but they've put a lot of work integrating them into VSCode in a way that a) adds a lot of value and b) can't be replicated by a competent developer in a weekend. There's nothing wrong with selling a product you made in a weekend, but if it's that simple chances are there's an open source equivalent that's just a github search away.

SaaS stands for "software as a service." When I think of SaaS I think of subscription-based software. It is usually web-based (meaning web applications that run in a browser and talk to a remote server), but I can't think of any reason that it necessarily has to be, and one could probably call newer versions of Photoshop / Adobe Creative Cloud "SaaS" since that is subscription based.

This distinguishes SaaS from traditional software that was sold as a good. In the good-based model, you purchased a copy of a particular version of the software and could install it and use it independently, on your own machine, no remote servers required for the rest of your life without requiring any "services" from the creator of that that software for the normal every day operation of that software.

As someone else already pointed out, all software (and most hardware) is an abstraction over top of something more primitive.

But to say that all SaaS is a "wrapper" around another service seems to be making the case that SaaS as such cannot provide original functionality that is not exposed by some preexisting lower-level service and that's not true. I currently work for a SaaS company and while we do provide integration with 3rd party services, our core functionality is entirely developed in-house. Our databases which store customer data are analogous to what storage devices on our customers' own devices would be doing if it were good-based software.

If you think about it, all software is already an abstraction of hardware by definition.

The real debate might be whether AI is the core technology driving a SaaS product or just an add-on to keep up with current trends. For example, SaaS products used to classify marathon runners’ images likely rely heavily on computer vision AI models—without them, the task would be extremely tedious. On the other hand, AI summarization in an email client is more of a nice-to-have feature than a necessity.

> If you think about it, all software is already an abstraction of hardware by definition.

Not really—interaction with a lot of hardware is defined by software. Just looking at a cpu you have no clue how someone would use it or what they would use it for or even what kind of capabilities it has for human interaction.

Isn't software itself an abstraction layer to communicate with the underlying physical components? In theory, it's possible to build a machine that solves a problem without any software at all. Isn't this what an automaton represents—hardware directly engineered to perform specific tasks without the need for software?

> Isn't hardware itself an abstraction layer to communicate with the underlying physical components?

Sure, but you're really straining the word "abstraction" here. By this definition all metaphysical concepts are just abstractions around material reality. Which is fine, albeit not that useful a definition of abstraction.

I agree. My point was that all software components rely on some form of abstraction. A SaaS, for example, is essentially a wrapper around a database, just as the database itself is a wrapper around OS APIs, and so on.

There has been a strand of SaaS companies that at least begin life as very trivial wrappers. Web based versions of CLI tools being the classic, anything from whois to ffmpeg.

Among these there always were low effort participants, but detractors (myself included) often ignore just how resistant to friction modern consumers (including devs) are. You really can build businesses by reducing the hoops someone has to jump through because to the end user that is valuable.

I can't quite put my finger on exactly why I dislike "gpt wrappers".

I think it's because they more or less boil down to a cash grab with little chance of a future. You provide users with a feature they presumably pay for while piggybacking off a large vendor that will, in due time, raise prices and/or make your business irrelevant with their own features.

It's no better than an NFT. You promise X, until the rug is pulled.

Does it solve my problem? Yes? Shut up and take my money! ~ Sincerely, a customer.

Everything is a wrapper.

The questions with GPT-wrappers are 1) how much would it would cost you to create an equivalent or better wrapper and 2) how much would it cost OpenAI etc. to replicate the same wrapper, either explicitly or through improved model capabilities.

The answers to those questions (especially the latter) make LLM-wrappers questionable to invest in or depend on. Even if it's great, there's no particular reason to think it will offer much over the base API in 6 months.

Sure, you can filter out some of the criticism as just haters. However, some of it might be deserved. It really depends on what the product is doing. Since Chat GPT is a very useful consumer facing product, any kind of chat bot built on top might not add a ton of value. Some companies are using an LLM such as GPT in very creative ways where you couldn’t get 80% of the value by just using the chat interface. You probably don’t see much hate in those cases.

> I just don't understand why people seem so upset about these new companies being GPT wrappers.

A lot are just close to vaporware. It also takes advantage of people more than most business models.

If your business is just adding a system prompt and trying to keep users from figuring out you could just do this with ChatGPT, I can't say I'm a fan. Not a hater... but not a fan.

If you have more functionality with actions, events and context, that's fine.

The problem is not with some product being a "wrapper" around something else. The problem is when the product adds nothing else of value to that underlying resource. GPT wrappers may add window dressing, but ultimately are just sending input to an LLM and regurgitating the output.

Most SaaS is not about tech. It's about marketing, sales, customer success, operations.

I don't think I have seen this hate personally, has that been on here?

1. What hate?

2. Saying all SaaS are wrappers is a wrong assumption. A service company provides a service of sorts via the medium of software. Take a CRM, it's a system to aggregate contacts, leads, deals etc. and their interaction in some sales pipeline. There might be 3rd party services integrated but the service itself is not a wrapper. You could be cheeky and say you are wrapping a SQL database but then every software written is a wrapper. So?

3. This is the real kicker: It's a question of dependency. If your entire business is completely dependent on another company, you've already failed. In the early days of Facebook's ads api, I had an ad-tech startup that managed ads in bulk on Facebook and later Instagram etc. Facebook was our "partner". Our "friend". They took us to dinner and sent us flowers. We had money, success, the lot. Until one day, our partner becomes our number 1 competitor. And thus an entire industry imploded.

Oh the Facebook thing. We had an ads account manager who really did take us to dinner. Would call, ask if we needed anything etc. It was great.

One day she ghosted us. I mean that, she ceased all communicated overnight. And in a month or something, the whole thing collapsed on everyone in the field. That was fun.

I also meant it in the most literal sense - they really treated us like partners, gifts, dinners, etc. And we (and I'm sure other like us) technically helped them build the ads api, innovate on it, bring them insane amounts of ad spend from our clients etc.

How it is implemented is completely irrelevant to users, something developers tend to forget. The merit and validation comes from customers (or lack thereof)

I feel it's mainly envy. Building SaaS is relatively easier than a database company, and in many cases SaaS earns more than many others.

SaaS is EXPLICITLY a software provided to you as a service. You could just run the software. This is no great secret.

Isn't every retail store just a wrapper around existing logistics infrastructure?

> Isn't all SaaS just wrappers?

Hahahahahah*gasp*ahahahah! Oh, I almost wish that were true!

My current workplace? There's stuff like a custom flowchart-language, with pseudo-programs that differ for each customer and have to match their implicit remote business rules. It talks to dozens of different APIs of other products of which a given customer might use a few, and each have their own unique insanities, like breaking HTTP specs, or falling over with just a little traffic so that we need our own caching layers.

Then there are some revenue-generating features that are basically custom-glue logic for synchronization and hand-offs between those client-systems, because not all clients are able/willing to develop a nice one-stop centralized management system of their own as a precondition of doing business with us and using our primary features.

> Not sure why these "GPT Wrapper" companies are getting so much hate.

Because they are that kind of lame shovelware and thin wrappers designed to extract money from the confused.

> I just don't understand why people seem so upset about these new companies being GPT wrappers.

I mean there's a big difference between varying degrees of what a "wrapper" is - the issue I personally have are applications/SaaS that make absolutely wild claims about what it can do by integrating AI and the application backend is a stupidly thin wrapper around GPT that just sends the prompt and returns the result.

If you're doing a lot of other stuff, fine, but in this space there's no real definition of what "integrating AI" means other than just being exactly that, a wrapper around GPT. Which really isn't innovative or groundbreaking.

Regarding other SaaS being "wrappers" - yea, typically these "wrappers" are around proprietary code the company has also developed, not just blatantly re-selling another SaaS's services.

A business doesn't need to be innovative or groundbreaking to make money.

So which GPT wrappers are profitable?

That's a very dismissive view point. I guess technically all software is just a wrapper around some other software or libraries that's a wrapper around the language that's just a wrapper around assembly instructions.

As for 'GPT wrappers', I haven't seen many mentions of this. It might be more common in non-technical circles where people downplay the role of 'AI' in the marketing. I don't see it as a negative if the product does something else in addition to AI. I have noticed many products only doing AI and then you could use the AI wrapper term negatively because they're so easy to recreate using existing free AI products.

iPhone wrapper: flashlight app

not iPhone wrapper: Uber

SaaS is just hype around MRR.

Monthly Recurring Revenue is a morally bankrupt business model that attempts to take things that do not belong in this pattern and force them on consumers.

It offloads the risk from the company onto the customer. They get money, you have to remember to cancel.

They can raise rates any time, and you are at their mercy.

They are rent seeking parasites.

SaaS can be good, but I have seen it used for evil far more than good.

The proliferation of this pattern means that less and less people build software you can buy and own. Less and less software can run on your compter.

What are you going to do when the internet goes down? Then next clownstrike event?

SaaS is a liability. The cloud is not your friend.