I made the website to be a small, longer term SEO bet. I did research and found people often would search things like “GPT 4o price,” so thought consolidating prices in one place and making programmatic webpages for each model could work on Google.

I made the website in a couple hours, collecting the prices for each token manually and entering them into a csv file. The image above is just a simple Nuxt app on top of that file. That was on a Thursday and I decided I would try to get my first backlinks on Hacker News the next day.

I put a link to the site with the headline LLM API Pricing Data at about 8AM EST and checked back an hour later and was surprised to see my post racking up upvotes and surging to number 3 on overall leaderboard. I checked my traffic on PostHog and saw that in the past hour 1000 visitors had come to the site in the past hour with no indication of slowing down.

I had to think quick about how to capitalize on this traffic to get some lasting value out of it that day. I figured the surge would only last for the rest of the day at most, which did not leave time for me to offer an actual free or paid service around the data. So I figured I could at least try to capture people’s emails and created a newsletter sign up form using Tally and put it under the pricing table.

While doing that my Hacker News post was starting to spread over to Twitter (the footer of my website had my Twitter username, @aellman) and comments on the post itself were growing. On twitter I was being tagged in posts asking for an API of the data. On the post itself there were comments coming in asking me to include providers in the data, to make changes in the UI and even that some of my data was inaccurate. I responded with updates to the website in real time: fixing the data, adding Xai models and responding to comments about what UI changes I would make in the future.

By the time the next day arrived I had over 100 newsletter subscribers and 12,000 unique visitors had come to the website.

I quickly learned that getting to the top of Hacker News does more than just get you traffic from the post itself. By the next day I had acquired over 100 backlinks from websites that pull popular posts from Hacker News, leading to a quick climb to a Domain Authority of 13 on Ahrefs. Throughout the week my website would also be featured on several Podcasts and Newsletters, helping sustain my traffic at over 1000 visitors per day. You can see my to date traffic below:

Over that week I acquired 250 newsletter subscribers, updated my site to include image models and made my first free tool, an AI Coding Tracker which has about 20 free users. I plan on continuing to ride my Hacker News success and growing the site’s authority through backlink optimization and making SEO focused pages like the new token counter and pricing calculator. In the long term, I see this site becoming an API that helps lower the LLM usage bill for API apps. The success on Hacker News and the site’s growing Newsletter confirms this is a point of interest for devs. I’m not exactly sure what that will look like but hope to learn from the site’s growing audience what the biggest pain points are.

I’m not exactly sure what the lesson from this story is. I have posted to Hack News since, about the coding tracker I made, and didn’t get a single upvote. Maybe it’s that the internet is random and that if you want to go viral you need to just post a lot to make your own luck. Or that unless you have a big following that going viral is rare and you should be prepared to take advantage of the moment when it happens. In either case I am looking forward to building more sustainable traffic to Price Per Token and making it more valuable than just a table of token prices.

I now realize how expensive those requests are. About 3 days into July, I got a warning from Cursor that I would be switched to usage-based pricing (how did that happen so quickly??). I continued using the service as I normally would, unknowingly racking up a sizable bill. In July, I spent $131 total on Cursor, including subscription, so around a 700% increase from the previous month. This is mostly because I code almost solely with Claude, which accounts for around 93% of my tokens in July.

You may now be wondering how I'm so aware of the specific metrics around my vibe coding. The reason is that I was so shocked at the end of July that I decided to code my own tool that would track my usage for me and help me stay on top of my bill for both Claude Code and Cursor.

It is the first package I ever made and the first repo I've open-sourced. I coded it in Python (my language of choice, coming from a data science background). It functions just as a simple CLI tool that refreshes your live Cursor and Claude code usage every 5 minutes and displays it on your terminal.

The coolest part, though, is that it sends your data to a Django server and gives you a nice dashboard on my website, breaking down your daily usage by model so you can see where your money is going. You can see my usage so far this month in the chart below:

Seeing your token counts accumulating throughout the month will also come in handy for Claude Code when they start their new rate limiting. Soon, Anthropic will also be introducing weekly token rate limits to their max-tier ($200) subscribers, which means you may also need to start being aware of WTD token usage in order to avoid a surprise throttle.

I thought it would be fun to make some kind of model ranking. So, taking inspiration from Open Router's ranking page, I created a leaderboard dashboard of my own. It works by pooling together the MTD token usage of everyone who uses my package and gives a breakdown of the most popular models and how much they make up of the global usage total. You can check that out at the link to my tool above.

Figuring out how to retrieve the data from Cursor was a struggle. In their API documentation, there is only mention of a usage data endpoint for teams, which doesn't work for me as a solo dev. After almost giving up, I noticed that when I downloaded my usage data in CSV format. I could actually see my browser quickly redirect to an API. From other open source projects, I knew you could get your access token from a local SQLite database Cursor created on every user's machine. So I just tried hitting that endpoint with my credentials from the DB, and I got back the CSV of my usage! It's not in any documentation, so hopefully, Cursor doesn't shut that down if they read this!

For Claude Code tracking, I leveraged another open-source project called ccusage and embedded it into my package. I simply run that tool like so:

# Build the command

cmd = ["ccusage", command, "--json"]

# Add date filters if provided

if kwargs.get('since'):

cmd.extend(["--since", kwargs['since']])

if kwargs.get('until'):

cmd.extend(["--until", kwargs['until']])

# Add other options

if kwargs.get('breakdown'):

cmd.append("--breakdown")

if kwargs.get('timezone'):

cmd.extend(["--timezone", kwargs['timezone']])

print(f"🔍 Running: {' '.join(cmd)}")

# Run the command

result = subprocess.run(

cmd,

capture_output=True,

text=True,

check=True,

timeout=30

)

I decided to publish it as a Node package so it would be easier for others to install, even if they don’t have Python set up. To do this, I compiled my Python code into a standalone binary and then wrapped that binary inside a Node.js package. Pretty neat!

It was a cool but also somewhat anxiety-provoking experience pushing a repo to GitHub and making it public. What if I accidentally push confidential keys or my code isn't good? Eventually I got over it and just pushed it, hoping that it would help other people avoid the surprise I had at the end of July. If anyone here reading it decides to use it, I'd love to hear from you.

Cursor's new pricing has driven me to start using Claude Code more heavily, as you can see above. I already have multiple more uses from it this month than last (I'm writing this on August 8th). I still find Cursor more reliable, though, so hopefully with my new tool, I am able to keep those vibe code costs in check and their free auto model continues to get better!

To make more I either need to grow my traffic or push affiliate products through my blogs. I started testing the affiliate approach by putting an ad up on my Tattoo website but haven’t gotten any conversions yet. My hypothesis is that my blogs are too impersonal and not trustworthy.

A promising route to revenue could be putting more effort into the blogs and making them into more of a real experience. The problem, however, is that I don’t want to be a blogger! I want to build software! It became obvious to me that the best way to monetize what I have done since February would be to package up my tools and sell them to people that already have revenue making websites. I pair my strengths of software development and my understanding of Pinterest with their established websites. My target customers right now are bloggers and small ecommerce stores.

With my new direction I got to work and spent about a week putting a reasonable frontend on top of the tools I had built that were making all my pins. I packaged it all up and put it on postpaddle.com. It worked decently well on my first deploy but definitely had some bugs. My philosophy on bugs, though, is if there is no one complaining about them, why fix them?! So I knew I had to get users to tell me to fix the bugs (and also tell me what they really wanted).

So I took to Reddit and Facebook groups.

Those 2 posts netted me five paying users. Perfect! Just enough to find all the bugs and iterate rapidly based on their feedback. Once I make these 5 people happy, I will know it’s time to scale. For now I am keeping my user base small and providing excellent support while continuing to write on here and X.

I will continue to provide updates on my Pinterest accounts as those are still my proof of concept but will also have posts like these just talking about my solopreneur journey.

Hope you enjoyed!

Alex

My goals with the 3 growing accounts are to have the growth continue (obviously) and monetize them. I have started this by putting in an affiliate link to an AI tattoo generator in You Make Tattoos. It gets clicked on everyday but haven’t had conversions yet. I will be switching up the ad and reporting on conversion rate there as it improves. For my interior account I put in a link to another website I built, Imagine House AI. It is a tool where users can see their own rooms in different interior styles. That ad is getting less clicks than the tattoo ad and so far all my users (there are about 75) are there are from other efforts (mainly posting on Reddit). My hairstyle blog doesn’t have any links yet but plan on putting links to physical products.

Going forward I will be launching more accounts, since I feel I have a good handle on how to grow them. I will be operating differently, however, with monetization at top of mind. What I mean by this is I will have a clear idea of how each blog will make money before I launch. They will either be trying to sell digital or physical products. I will also explore running programatic ads.

Without further ado here are the traffic results:

With increased traffic I can start thinking about offering free tools on my sites- like an AI home design tool on Any Era Design or an AI tattoo design for You Make Tattoos- and receive feedback in the form of clicks and sign ups from my blog visitors telling me if they are desired. The hope here would that people would start spending more time on my sites which would help with my pin’s rankings. EventualIy they could even become paid tools. I will start sharing traffic once any of these tools are live.

Find the metrics from Pinterest below:

Increasing clicks coming in will allow me to do 2 things going forward:

• Perform analysis on my pins to see what influences number of clicks. I am posting pins with different title and description strategies, different college and overlay text styles and using several image models. I will explore how these impact pin performance and adjust posting accordingly.

• Begin to turn these blogs into more functional websites and possibly stand alone businesses. Once I reach 100 clicks/day I believe I will have enough traffic to begin testing new features beyond the blogs to keep visitors engaged. Ideas include:

  • a tattoo generator for You Make Tattoos

  • an iPhone wallpaper generator for Look At This Wallpaper

  • a custom interior design explorer where users can upload pictures and experiment with different interior themes for their homes for Any Era Design

  • a hairstyle experiment tool where users can upload pictures of themselves and try different hairstyles for What If Hair.

With Pinterest heavily weighing if visitors stay on a site for more than 10 seconds in their pin ranking algorithm I hope that adding interactive experiences would lead to more impressions.

See the results so far below!

My CTR is pretty low across the board, however, ranging from .05% to .23% for each of my accounts. I am hoping this will improve as my “grid pins”, or collage style pins with overlay text like the below, start to grow in impressions. I started posting grids about 10 days in so most of the views right now are coming from regular single image pins with no text overlay. Time will tell what kinds of pins have the best CTR. Find the performance in the charts below!

You Make Tattoos is the breakout star, already racking up over 30k impressions and growing rapidly. Outbound clicks are low mostly due to the fact that the earliest pins are the ones generating most of the views and they don’t have outbound links because of now fixed bug. I am hoping traffic starts to pick up for that account once the linked pins start getting traction. At that point I can start to think about what to do with the website to keep people on it longer- a key part of Pinterest’s algorithm is the amount of time users spend on an external site after clicking a pin.

Metrics below!

What If Hair

Any Era Design

Look At This Wallpaper

You Make Tattoos

Your Color World

This stat tells me that seasonal pins have the potential to way outperform my “evergreen” pins (for coloring books those would include my superhero, nature scene and flower coloring page boards). In response I have doubled down on seasonal themes and have already made 65 Easter coloring pages with 30 more on the way.

I don’t want to miss this opportunity in my other accounts so I went a step further and made a tool that tells me all the seasonal moments in a Pinterest category and when the peak week of the year is for that moment. It works by scraping Pinterest for key search terms I input, extracting common annotations from the pins that show up, using an Open AI assistant to find which of those annotations could be seasonal and then scraping Google Trends to see the search history of these seasonal annotations. I then start posting seasonal pins 8 weeks before their peek week.

Here is a sample output for Hairstyles:

I am going to be using this tool going forward to take advantage of the seasonal moments for all the categories I am posting in as I believe that will be essential in maximizing my possible impressions and clicks.

In their paper Improving Pinterest Search Relevance Using Large Language Models they detail how they are starting to incorporate LLMs into the algorithm. In the paper they list the important attributes of each pin they use to decide their relevance to user’s searches. Unsurprisingly, attributes like title and description are included but another key factor they mention is “the titles and descriptions of (outbound) URLs.” This makes sense since Pinterest is a platform designed to make the user click on outbound links. They WANT their users to end up on pinner’s websites. That is why they also state “A key metric optimized for Pinterest Search is the "long click", which occurs when a user clicks through to a Pin’s linked webpage and spends over ten seconds there.” So it is important for Pinterest to understand the contents of the websites that pins are linking to.

In order to obtain details about the websites that pins link to, Pinterest uses a tool called Pinterestbot to scrape the webpages in a similar way google uses Googlebot for their search. My websites up until now were not “Pinterestbot friendly.” To change this I did the following:

• Updated every page on my websites to have meta tags in my headers. These are HTML elements that describe what is on a webpage to a bot like Pinterestbot and include things like a description of what is on the page and a title of the page.

• Created a robots.txt. This is a file that gives explicit permission to robots to scrape a website and exists as a standalone page. You can go to https://anyeradesign.com/robots.txt to see one of mine.

• Added sitemaps. This is another page that is added for robots. It details where are all the pages on a website are and includes information about when they were last updated and how many images are each page. You can see one of mine at https://anyeradesign.com/sitemap.xml

• Added alt tags to all my images. These are HTML elements that provide simple descriptions of images on webpages. Pinterest is essentially a visual search engine so it is important that it can see the images on linked sites and verify they are similar to the image of the linked pin.

With all of these updates it will be very easy for Pinterestbot to see all the pages on my website and get a clear understanding of their contents. This should help my pins rank because the algorithm will have more information about their outbound link destinations and be able to confirm their relevancy.

My name is Alex and this is going to be my home for sharing what I learn about Pinterest and AI + marketing as I try to grow my own Pinterest accounts. I plan to make this a transparent journey and will be sharing updates on impressions, clicks, saves and clicks from my accounts.

As of now, I am running 5 accounts in different niches across Pinterest and each are driving traffic to respective blogs I made. If you are looking at them today, all except one are extremely basic websites that are just displaying the content I post on Pinterest. That is because the point of this project, for now, is not to grow these sites into their own businesses. It is to discover how to reliably grow on Pinterest using AI. All my content is AI generated (in the future I may try using stock photos but still with AI generated text). I hope others can use what I share here to grow their own businesses.

Without further ado let me introduce you to my Pinterest accounts and their performance about 2 weeks since launch:

First Any Era Design, posting about Interior Design. Performance:

Next What If Hair, posting about Hairstyles.

Next Look At This Wallpaper

Next You Make Tattoos, post about Tattoo designs

And finally, Your Color World. This is supporting a website I made last year that lets you generate coloring pages with AI. It posts coloring pages. You’ll notice a dip in impressions recently, this is because I didn’t post in the month of January. I started posting again last week and expect the metrics to recover.

Looking forward to updating you on how these accounts grow and what I learn along the way!