DEV Community: Darko

Why Cursor's flat-fee pricing could lead to its downfall

Darko — Wed, 09 Apr 2025 11:11:12 +0000

AI costs rise directly with usage. You pay for both input and output tokens and get a final price as a result.

Cursor and other popular AI coding tools use flat-rate pricing despite AI's usage-based nature. In my opinion, this creates more problems than it solves and has a tendency to disappoint users.

In this article, I’ll talk specifically about Cursor, but many of the principles can also be applied to competing products that charge flat rates (e.g. Windsurf).

Flat-fee AI pricing distorts people's perception of real costs

Cursor has a recurring subscription plan. You pay $20/mo and get 500 code completions from premium models like Claude 3.7.

The tricky part comes when you “run out” of your current plan. This is where reality starts to creep-in. There’s a lack of clear consensus on how to solve this among flat-fee AI providers. I’ve seen tools that:

Throttle users’ requests after they exceed their quota (Cursor does this)
Get people to switch to an inferior model (your code generation quality drops significantly as a result)
They just reveal the real pricing model and ask people pay-per-used-tokens (people didn’t subscribe for this)
Ask people to upgrade to a higher plan that costs significantly more

Flat-fee pricing for access to AI models can be summarized as: Painless now, feel the real pain later.

Maintain your subscription or get less functionality (subscription lock-in)

Cursor and other flat-fee providers send a clear message early: Continue paying us monthly for a standard+ plan or you’re going to lose a bunch of functionality.

This is the second misaligned incentive with flat-based pricing for AI tools: They take away core functionality unless you’re subscribed to a paid, flat-fee plan.

With Cursor, for example, I can import my own API keys but there are a few “gotchas”:

Just look at the number of warnings above. Other products like Windsurf make it outright impossible to use premium models with their free tier option.

Pay-as-you-go AI providers don’t create “win-lose” incentives for users

I’ve seen the rise of AI coding tools, such as Cline and Roo Code that ignore this “flat fee” pricing philosophy altogether.

Instead, what they do is charge “per-usage”; any costs they incur go directly to the AI model API of your choice. Cline/OpenRouter’s monetization model is simple: Add a 5% markup on top of anything you pay your API provider.

What do actual users think about all this?

The sentiment is positive. When you look at how fast those AI tools are growing, you’ll see a nice curve. This is Cline’s Github popularity (counted by the # of stars) from 2024 till now:

Those stars translate into downloads; Cline currently has over 1.1 million downloads.

Roo Code is a fork of Cline that has close to 10k stars on Github and 300k+ downloads in less than a year.

Kilo Code (us!) is a fork of Roo Code that went on the front page of Hacker News recently and is also growing rapidly (to the point where people started abusing our free tier), but that’s for another story.

How this pricing “scheme” works: Once you submit an “API request”, it goes through your preferred API model of choice. You can either enter your API key and pay the provider directly or use a direct relay (Cline and Kilo Code both have one) where you don’t have the same rate limits and get a certain amount of free tier credits.

When incentives are aligned, you also have a clear incentive to decrease costs

This has been an interesting observation.

I’ve seen a lot of innovation around the goal of decreasing the costs needed in order to fulfill a meaningful “code completion” request, especially for existing projects.

None of this recent innovation came from Cursor:

Cline was the first to introduce a memory bank with a simple workflow (Cursor has hacky ways to do the same thing).
Roo also has a vibrant community offering several libraries for optimizing your prompts
There’s also an ongoing effort to create an efficient way to use the cheapest price/results model for the task you’re trying to do (if you have any suggestions for this, let us know!)

As I’ve mentioned before, Cursor has been lagging behind here.

What’s the real reason why Cline is crushing it when it comes to cost innovation, while Cursor is running behind? Properly aligned incentives in my opinion:

With Cline/Roo/Kilo Code, you pay for usage (input/output tokens). Those costs are salient (as they should be). You can literally feel the pain of spending money. So people find ways to reduce it.
With Cursor, the picture is a bit fuzzy. You’re “using your flat-fee plan” up until a point where you get throttled/asked to upgrade (which usually comes as a surprise). Psychologically speaking, you get instant gratification but suffer later.

Incentives make a way bigger impact on our world than we think. There’s even been entire books written about this topic.

Is Cursor going to get disrupted just because of its pricing strategy?

Time will tell. There are other factors at play here; for example, 2 years ago, it was tough to get a “Cursor experience” in VS Code. Nowadays, you have some really good extensions that can do at least 80% of what Cursor does. No need to change your IDE.

There’s also the factor of Open Source; both Cline and Roo have 100s of open-source contributors, and 1000s of people in their Discord are constantly suggesting new features. That’s the power of open source.

All these are important, yet many forget about pricing as a factor, the incentives it creates, and how it will impact the “AI coding tools” race as a whole.

I hope that this article has helped clarify the underlying issues at stake.

Small technology changes create profitable bootstrapped products

Darko — Wed, 02 Nov 2022 11:30:59 +0000

How do small technology changes create profitable bootstrapped products? Let's start with an example of such a change:

Example 1: Apple Lock Screen Widgets

The change:: When Apple released iOS 16 on September 12 this year, one of the biggest new features were lock screen widgets.

The opportunity: This change propelled widgets to the top of the App Store According to AppFigures, with some of the most popular widget apps seeing a 169%-2400% increase of daily downloads.

This change also spurred a new wave of lock screen widget apps and tech publications were more than happy to give free press to many of them.

Look out for: When new platforms (like Apple, Facebook) add particular features, that usually creates opportunities for countless small projects to be created around those features.

Example 2: Cloudflare R2

Wanna create an image hosting website? Sure, you can host all images on a service like Amazon S3, but be prepared to pay $0.09 for each GB transferred.

What if a user uploads an HD image that goes viral? An average image can be as big as 2MB (or more):

The last thing you want to do with a side project is monitor for unexpected costs.

Let's say gets viral and gets 100,000 requests. That's 200,000 megabytes or 200 GB. That's ($0.09 * 200 =) $18 just for a single image. Ouch.

The change: This was a problem until Cloudflare R2 appeared. On September 28th, 2021, Cloudflare announced an S3 competitor (called Amazon R2) where you effectively pay $0 for transfers and only pay for storage.

This caused Amazon to respond and offer the first 100GB of bandwidth for free. But again, after that you still pay $0.09. Still not ideal for a bootstrapped project or a side business which can take off.

On September 21st, 2021, Cloudflare made R2 available to everyone. The only major cost is $0.015 per GB stored (and some minor costs per 1 million request). The rest is free.

The opportunity: This change opened up a whole lot of opportunities for indie hackers like me and you to create products that previously we were simply too afraid to create:

An image hosting service
Backup creation/storage tools
Anything related to file uploads where you'd expect high bandwidth (think: Dropbox, big document sharing services, etc.)

The nice thing about all this is that we could be predictable around the pricing. We could create pricing tiers where we charge per-storage (like all big providers do) not not per-bandwidth.

Look out for:" Once in a while, big PaaS providers release an update that makes it possible to build SaaS applications that weren't possible before.

Example 3: Generative AI development

The change: Here I'm talking about GPT-3, DALL-E 2 (and competitors like Stable Diffusion and Midjourney) and upcoming text-to-video projects like Google's X and Facebook's Y.

I published a post last week on how DALL-E 2 (and its competitors) are already disrupting 15 different industries. This just less than a year after its initial release. Here's the current growth of one website which is built on top of a DALL-E 2 competitor (called Stable Diffusion):

GPT-3 has already started to disrupt industries like novel writing, writing marketing copy and, well, eliminating the writer's block.

Look out for: AI is on a fast track to replace many aspects of many professions. Those changes will come as APIs and the people who'll win will probably be creative founders who adopted that API around a market need.

Technology changes like these happen all the time

A big plaform makes a change that opens up a new wave of apps (Apple with iOS 16)
A developer platform creates something that makes it affordable to create something that previously was very expensive to build (Cloudflare RS).
Someone creates an AI platform that enables a whole range of things that previously weren't possible (Open AI with GPT-3/DALL-E 2).

So the important question is:

How do you identify those changes first ?

There are several things you can do:

Keep an eye on major platforms for new features. (platforms like Meta and Google have newsroom where they post their latest features.)
Alternatively , you can follow people who monitor the big players (like Stacked Marketer and Matt Navarra).
Subscribe to RSS feeds of blogs that power the majority of today's web (Cloudflare has a blog, AWS has a blog too). Hacker News features the most important changes regularly on their website.

YouTube handles = the next lucrative bootstrapped opportunity?

Minor changes can also create potentially lucrative bootstrapped ideas.

One such recent change is YouTube adding handles. In other news, they've finally introduced usernames::

If you've been in the domaining industry and/or social media for long enough, you may recognize the opportunity here. Premium domain names sell for thousands each day. So do Instagram handles.

I can see indie hackers creating small tools in the future that will:

Enable YouTube creators to easily find good available handles for their channel
Someone making a tool that will grab the premium handles and then resell them. Or create a marketplace for people to buy/sell handles.
A tool that will watch to see if your desired handle is "released" and becomes available again.

When you first read about this YouTube change, you probably thought to yourself, "Meh, nothing special." However, if you think a little harder, you will see a lot of opportunities.

This is just one such change. There are tens of different changes like this happening every week.

The opportunites is here. You just need to go out and ~~grab~~ code them.

15 industries that DALL-E 2 is likely to disrupt

Darko — Fri, 14 Oct 2022 11:45:32 +0000

One of the problems with new technology is that initially, we still don't know what it's useful for.

I did some research to try and find out the main use-cases for DALL-E 2 and competitors like Stable Diffusion and Midjourney. Here are my findings:

1. Stock photos

OpenAI's DALL-E 2 might may mean we never need stock photos again. I decided to put this claim to the test and do some research.

AI stock photo directories have already started to appear: StockAI by Danny Postma is one of them. It already started getting traction and received close to 90,000 visitors during its first month.

AI Search Engines: These are websites which don't advertise themselves as stock photos directories, but as AI image search engines. OpenArt and Lexica are the top 2. And they've been growing like mad. Just take a look at Lexica's SimilarWeb page:

And OpenArt:

These sites show no signs of slowing down.

Bye-bye, human stock photo actors: This Reddit post can give you an idea of how close we're to making stock photography featuring humans. This stock-like photo was made entirely with AI:

2. Graphics novels

This person created a graphic novel which ended up on the Amazon's Best Seller list:

This novel was also entirely generated with MidJourney, a DALLE-2 competitor:

Persisting characters: One of the problems of using AI text-to-image tools is that you can't "persist" the images so you create different pictures with a same image. This is changing, however.

Here's another example of a graphics novel made with AI.

3. Children's books

Some people have managed to make children's books in 2 hours with text-to-image AI. Yes, this includes research, image generation and layout design. Here are the results:

4. Game characters and graphics

We can easily imagine how DALL-E 2/StableDiffusion/MidJourney will disrupt this industry. Here, a person is playing with an Unreal Engine integration where players can create content in-game:

DALL-E 2 also has the potential to create variations from popular game characters, fix game textures and so on.

5. E-commerce product ideas

Would you buy this bottle?

Let's say you're selling bottles. With text-to-image AI, you can potentially create variations of different bottle designs and gauge interest before making them a reality.

6. Newsjacking

In case you're not familiar, newsjacking is integrating your business with a trending event or a breaking news story.

Let's say you're selling rifles (not the best thing to be selling, but stay with me). You wake up one day and learn that everyone is talking about Lord of the Rings. You want to attract some attention by making Gandalf an action character holding a rifle in his hand (wild, no?). With AI, this can take seconds to make (this quirkly picture of Gandalf got over 798 votes on Reddit):

Pretty neat, no?

7. Logo designs

Some people have already used DALL-E 2 successfully to make logos. Here are some of the results:

The nice thing is that you can iterate an unlimited number of times before you find what you like.

8. Marketing (avoiding creative fatigue)

One of the main problems with images is that people get bored watching them. With text-to-image AI tools like DALL-E 2, you can potentially create unlimited amount of relevant images.

Let's say you're in charge of promoting Oktoberfest. You could hire a bunch of designers/graphics artists or pull out DALL-E 2/StableDiffusion/Midjourney and come up with something like this:

9. Tattos (yes, tattoos)

What are tattoos in essence? They're interesting drawings that represent a piece/concept you care about. Text-to-image AI tools are PERFECT for creating these things. Here's an actual tattoo designed with MidJourney:

10. Sketching

Tech like StableDiffusion allows you to upload a sketch and get a full-fledged picture in return. This is what I mean:

This can be really useful for parents, who want to turn their children's sketches into work-of-art. You could literally draw an owl with some circles:

Sketching has a lot of use-cases, including clothing design, which we'll explore next.

11. Clothing designs

Wanna make a cool piece of drawing? Just draw a rough sketch and feed it into Stable Diffusion:

I can imagine brands doing this for quickly testing a clothing line, gauging feedback and then making it into a real product.

12. Photo shooting (and editing)

On September 21 2022, DALL-E 2 started to allow users to upload faces.

To give you an idea of what it can do, here's an image of the original (the first picture) + the rest of the images DALL-E 2 generated:

Sure, it's still "beta" but I can imagine this having use-cases for photo shooting and editing. For example, you could generate pictures for your Instagram profile and/or your LinkedIn:

13. Interior design

InteriorAI is a product where you upload a photo of your current room. It then creates ideas of different arrangements/styles you can use to change your interior. They've launched and had 10,000 visitors in September, according to SimilarWeb.

Image.computer is another product that helps you generate your next interior design idea:

They too just got started and had close to 14k visitors in September.

14. Content writing

Some publications already started using text-to-image to generate featured images for their publications:

We can just image this trend will continues. The reason: A lot of writers want to "enhance" their text with images.

15. Getting people interested in a topic

I wanted to end this post with a use-case that's a bit...unusual.

Here's an AI-generated picture of a guy taking selfie during the battle of Omaha Beach, 1944:

While doing my research, I've noticed people wanting to use AI to re-imagine the past, or pair the past with current trends.

I just wonder if this will have an impact of history over the long-run. Will people become more interested to learn about the Battle of Omaha Beach and World War II as a result of seeing fascinating, AI-generated selfies?

Text-to-AI is amazing at combining something people are interested in (say selfies) and something else they don't know about (say a history event).

Let's say you want to get more people who like animals interested in CyberPunk. Here's a cool image that might get people interested: