by valianter on 8/28/24, 6:05 PM with 47 comments
SAAS brings the cool things people build in our insular computing community to the general user in an intuitive way, and that's celebrated. Not sure why these "GPT Wrapper" companies are getting so much hate.
by throwaway2016a on 8/28/24, 6:31 PM
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.
by marginalia_nu on 8/28/24, 6:39 PM
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.
by candiddevmike on 8/28/24, 6:20 PM
by wiradikusuma on 8/28/24, 6:51 PM
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.
by bognition on 8/28/24, 6:25 PM
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.
by api on 8/28/24, 6:18 PM
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.
by throwup238 on 8/28/24, 6:48 PM
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.
by gspencley on 8/28/24, 6:56 PM
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.
by coronapl on 8/28/24, 6:33 PM
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.
by fidotron on 8/28/24, 6:37 PM
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.
by bitbasher on 8/28/24, 6:37 PM
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.
by romeros on 8/28/24, 6:28 PM
by scarmig on 8/28/24, 6:34 PM
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.
by todd3834 on 8/28/24, 6:27 PM
by robswc on 8/28/24, 6:54 PM
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.
by codingdave on 8/28/24, 6:46 PM
by tnolet on 8/28/24, 6:23 PM
by kristiandupont on 8/28/24, 6:20 PM
by harel on 8/28/24, 6:30 PM
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.
by Bumblebeeeee on 8/28/24, 6:44 PM
by sidcool on 8/28/24, 6:08 PM
by remram on 8/28/24, 6:53 PM
by leros on 8/28/24, 6:33 PM
by PaulHoule on 8/28/24, 6:08 PM
by Terr_ on 8/28/24, 6:31 PM
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.
by JohnMakin on 8/28/24, 6:23 PM
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.
by impure on 8/28/24, 6:31 PM
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.
by jumploops on 8/28/24, 6:30 PM
not iPhone wrapper: Uber
by dirtybirdnj on 8/28/24, 6:38 PM
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.