1 is an OCR and document extraction service [0]. We started with three customers asking for the same services and found none that were really useful (and supported Chinese characters) on the market at that time. Lately the product pivoted to based on (fine-tuned) LLM/VLMs and focus on adding various features that LLM out of the box are missing (fine tune based on specific customers data, prompt tune for particular type of elements e.g. Checkboxes, split 100s pages of PDF into dozens of documents with a few pages)

We're at around 55k MRR, the price model is per page, and we sign annual contracts with most clients (with some discounts)

2nd is an open-source CIAM [1]; Around 35k MRR.

We knew nothing about marketing when we started, so we partnered with local GCP/Azure as an ISV to get our first paying customers, which drove us to the more "Corporate" segment of the market.

A huge challenge is obviously how to market the product, but customer support for developers is tough as well -- you have to be developers to provide support for other developers, and sometimes it feels like you're troubleshooting for another dev team.

For example, one time we had a client email us saying they were getting incorrect results from our API suddenly, after many back-and-forth emails, we finally asked if we could do troubleshooting with a video call and share screen -- turns out they were interestingly calling our API via a proxy with cache enabled.

[0] https://formx.ai

[1] https://authgear.com

It turns out that the most effective and easy way for their sales rep to pitch your products to their client is if you have something complementary. For us, this is because they have similar products, so they can propose it to the client if they don't like the native one.

It also sometimes conflicts with the incentives of us.

Anyway, this guy was the go to guy for gas customers, and knew the database inside and out. So he created his own company, resigned as a full time employee, waited until the panic had set in properly.

Then he offered consulting services back to the energy company saying he'd take care of any database processing costs, or cloud migration costs or whatever, and moved the customer data for gas customers to his own system. Then he created an API, waited a while more and said he was going away again.... Or he could stay supporting this setup if the energy company agreed to a monthly fee and API usage. Then, as far as I know, he sat back and just watched the money roll in while he automated everything else about the job.

That sounds highly illegal.

The company could have called the bluff and passed on the consulting services.

> Maybe there’s a more charitable way to tell the story.

Yeah, management didn’t give the guy a raise so he quit and he could say no when they came begging.

If there’s anything that went wrong here it was management asking ‘what if we pay him more’ but not asking ‘what if we don’t’.

I’ve considered doing similar for one corp I’ve once worked with. The corp used an obscure hybrid cloud solution, unfortunately, the cloud provider didn’t really understand the corps needs (governance,devex,monitoring) making it impossible to do anything basic without manual action from an administrator. Pretty solvable with a couple of APIs and a few dashboards

The company couldn't even function without that person. Calling that an extortion is quite a leap.

How? The original company could have hired another employee to replace him, instead they entered into a b2b relationship with his company.

I've often considered something similar. I used to support a bunch of apps, where I thought "I could build most of these from scratch and they'd be better than what I'm supporting." Even if my employer didn't find value in them, and hired someone else to support the old junk, other companies would probably buy the new ones.

If the employee deliberately made the IT infrastructure worse, then maybe that would be fraud, but it sounds like he was the main person who was improving it.

I guess check your employment contracts and whatnot, but it's very common for people to leave a company and start consulting and/or freelancing in the expertise they gained at that company. And it's not too uncommon for former employers to be one of their first clients or get your first clients from relationships associated with your former employer.

All of this assumes you don't burn bridges and have the interpersonal skills, in-demand expertise, luck, timing, and interest in building out a small service-based business.

We don't care about them though.

What am I missing? Can you spell out some boundaries for what you're implying because I must be wildly missing it.

I put it in maintenance mode in 2019, so it's about 99% passive income, as I spend only about an hour or two on maintenance per year.

I'm surprised all my clients haven't switched to LLMs, but maybe I still outdo LLMs on price/accuracy since it's so niche.

I'd like to sell it to someone who wants to do something with it, but it would probably take me 30-40 hours to package everything up to hand off to someone, so I consider just the opportunity cost there to be around $5-10k, and I don't think anyone wants to pay $10k for an API that makes $200/mo.

What I wish I knew: don't use RapidAPI. They charge 20%, they have a terrible interface, and they let customers run up huge charges and walk away without paying anything. I wish I'd just rolled my own simple thing with Paddle.

[0] https://zestfuldata.com/

You get a real job to have health insurance.

Looks like that $200 is just side income, not their main job.

[0] https://mtlynch.io/resurrecting-1/

[1] https://stackoverflow.com/a/52304008/90388

[0]: https://x.com/DmytroKrasun

[1]: https://screenshotone.com

I didn’t consider wrapping any service.

What needed for scraping is a bit different for what needed to screenshot websites.

I need to have full control over my cluster to guarantee the best possible quality.

I signed up on my phone and tested in the playground.

It will fit perfectly into my workflow. I'm building a hyper-local directory site.

Getting good images for businesses is hard, so I'll use this to grab an image of their site as a place holder.

I can also add it to my AI workflow where I pass a website to OpenAI Assistant to extract data. OpenAI s not as robust with URLs as it is with images or PDFs. Often it won't visit then URL.

I can use this to get an image or pdf, pass it on and ask for the data back. OpenAI is better with files than URLs in my experience.

Good job!

Well done!

SEO, social media and other channels. I spent a lot of time on all of that.

docker run -p 3000:3000 browserless/chrome:latest

If you have one, very narrow specific use case, then maybe that’s not so bad. But sounds like a huge pain if you need that for any arbitrary site.

Oh okay so it does:

chrome --load-extension=path/to/ublock --headless --screenshot https://website.com

The cost with running a docker container is you have to manage that container. You’re paying with dev time not money.

    curl -X POST http://localhost:3000/screenshot \
  -H "Content-Type: application/json" \
  -d '{"url": "https://example.com"}' \
  --output screenshot.png

Can chrome even take 100 screenshots a second? What would be the max? I think it's interesting.

https://github.com/browserless/browserless/blob/main/LICENSE

https://capturekit.dev

If anyone wants to build “turn websites into APIs” and do it really well like OP, feel free to hmu.

Highly experienced full-stack and rust developer with experience at startups, G and Google X.

So if you want, "turn your website into an app", that would be interesting.

I don't think I'm authorized to share any of the specifics, so will keep it generic.

The API is a world-class machine learning model for a specific scenario. There's a public price list, and various customers manage to negotiate various discounts.

Our biggest challenge is that Google Lens (while much worse than us for our specific domain) is becoming good enough for the average potential customer.

I think one of the regrets is only doing the ML API and not the end-user apps. It turns out the people creating the front end pocket most of the money.

Well, yes. They're solving a pain-point for the paying customer. You're solving a pain-point for someone who is solving a pain-point for the paying customer.

You're one (or more) degree's removed from the source of the revenue.

I'm curious why this is not preferable? You can focus on your core competency and I imagine you have enough apps using your API that it more than makes up for getting only a small slice of revenue from each.

Would you be able to elaborate on this? I don't fully understand this statement.

You set API pricing at 1.5 cents per image analysis.

Another company creates an Android hotdog/not-hotdog app. They price it at $5/month. Each app user does an average of 60 food queries per month.

You get 60 API calls = $0.9 revenue. Let's say half of that gets used for compute costs. You're left with $0.45/month profit per user.

The company that made the frontend around your API gets ($5 - 30% app store cut - $0.9 your API cut) = $2.6/month profit per user.

> Would you be able to elaborate on this? I don't fully understand this statement.

the economic player with direct contact to the customer "owns" the customer and has a lot of power in negotiation with suppliers. the customer-facing players have the most information about their customers, and can offer adjacent products (want fries with that?)

Uber and Lyft make big money, not their drivers. Amazon and Ebay make big money, not their sellers. McDonalds makes more money than their food suppliers, and franchisees.

The exception is with something like intellectual property, let's use movies as an example. The owners of the content want to sell it widely and will do a variety of distribution deals for different distribution channels. However, if any distribution channel starts taking a big slice of the money pie, the terms of the contract renewal will be changed because without the content they are dead.

The challenge when exploring this topic: the incentive to stay under the radar. Those succeeding don't have much to gain from sharing details here. Worst case: it could invite competitors into their space.

Communities that thrive on growth (e.g., open-source) tend to share freely, but API businesses, especially ones which are easy to execute, often guard their edge.

A recent finding I had, while not necessarily an API: services which help you 24/7 stream a lenghty video file. YouTube live streams seem to work well for those lofi-types of channels, and there are services which are built to enable autopilot live streams.

At the end of the day, it is about what the customer is willing to pay for.

Back when The Pirate Bay was huge, music was essentially free. But Spotify came along and proved people are ready to pay for something better.

ImageMagick is an open source tool for resizing images etc. But some people successfully build API services or SaaS-services on top of it.

It works because people AND businesses pay for convenience.

What space do you have knowledge of? What pains do people have in that domain that can be solved with tech?

Always start with the problem. And start with an industry you know by heart or customer profile you truly care for.

For me that was software developers. I was that customer myself. I programmed a certain kind of solutions, realized there should be an API for this and built that API.

However, in the early days of Spotify the ”play any song with a click” was pure magic and nothing piracy could compete with.

A fun anecdote from that time is also that basically the whole Spotify catalogue was full of pirated music — they had not yet secured any music rights and I personally thought they never would succeed with that.

I know right this very second exactly how I could do a unique twist on my existing business and conservatively make another $1 million a year in profit within 2 years. But I already work 60+ hours a week on this business and I’m making tons of money, and the risk of revealing my secrets to another person to build the new business is simply too high.

Our MRR is ~500 000 EUR and our pricing model is pay-as-you-go (per SMS, per MMS, per phone call minute, per month for virtual mobile phone numbers).

The problem we solve is programmatic access to the mobile networks, specifically in Europe/Sweden.

We got out first paying customers through offline networking: going to hackathons, meetups and poking tech friends to find the first few early adopters.

Which is also our biggest challenge, it is hard to scale an offline based go-to-market method.

It has certainly been a painful struggle to get here and it still feels surreal it works so well.

Very custom contracts and back in 2011 when we started there was no such thing as a virtual mobile phone number.

These days the operators are a bit more aware of the value of A2P (application-to-person) versus P2P which was/is very much on the decline. That also means the operators have capacity built of ready to be used without new investments.

As for our own software, we are mostly a Python based shop. And we use tons of open source. The most heavily used components we replace over time with cusomized software as these are generally fully not suited for our needs.

The true value for our customers is our technical support, our operator conectivity, the robustness of the platform, and lastly the nice REST API with it’s debugging capability.

We basically have full international support these days, but are strongest in Northern Europe.

For instance, if you want a virtual mobile phone number in Sweden that supports both SMS and voice, Twilio can’t provide that but we do.

Plus we have nicer API if you aks me :)

Company name is 46elks. The country code to Sweden is +46 so the name is an hint our service is in the telephony space.

When we launched 3 years ago our differentiator was that we could train both cheaper and faster by running on TPUs, these days GPUs have mostly caught up, and open source models are not as competitive as they once were.

It’s making ~5k/month these days, not bad as we’re no longer actively working on it, but a fraction of what we were doing a year ago.

The main challenge for us was the non-technical part. We built an API-first product because we love the tech and felt it’d allow us to focus on that part. But we still had to do marketing, sales support etc which we didn’t enjoy or excel at.

Now we’re both back in larger companies where we can focus on doing ML. It was satisfying to build a working business from scratch, no regrets, but I’m definitely happier now.

Biggest challenge was getting the first few customers (is there anyone for who this was not the case?).

https://borgcloud.org/speech-to-text

As the other commenter mentioned, this seems like an extremely competitive space. But I did a quick Google for speech-to-text and your solution is a few times cheaper than most others.

I know you may not want to share your secret sauce, but I'm curious as to how you managed that. I'm guessing you use Whisper, is the trick to just run it on your own hardware instead of paying AWS prices for compute?

Will probably have to at least match their price in STT API and take a hit to revenue. But we are working on an end-to-end voice offering, and Groq does not seem to have anything in that area yet, so hopefully will stay afloat.

Btw, looks like you made a typo in your first paragraf writing text-to-speech instead of speech-to-text.

Thank you, fixed!

It took quite some efforts to have a service of really high quality, resilient and bug-free but I'm nearing this and it's a real pleasure.

To answer all your other points :

- What's your API: An API to convert HTML documents (URL/Raw) to PDF - How much MRR: About $12k - Pricing model: Subscription plan with limits in the number of converted documents that depends on the plan. The rest is the same for all the plans (no features unlocked at certain level). - First paying customers: Mostly from IndieHackers and Quora at the beginning, and a successful launch on ProductHunt when it was still big. - Problem I'm solving: Not having to manage your own servers to handle conversion. It can be done locally for sure, but as the service grow, the need for resources increases too and can lead to issues down the road. With PDFShift, it's just set up and forget it. - Biggest challenge: I was initially at OVH Strasbourg, when it caught fire, and lost a lot. I had to rebuilt the database from data I had around and gave access to everyone for free until I had all sorted up (took 2 days). This was a good lesson to learn to not save the backup on the same service and at the same place! - I would definitely do it again. (I'm thinking of this frequently for other services too :D)

@wenbin posts on here about it

Same thing for home improvement - Lowe’s in stock indicator is wildly wrong to the point that I sometimes call and ask an employee to go physically look at the item.

Sure, but alternatively, using the same data you could find out what the whole list costs at each store and just go to the cheapest one.

Privately, I would like to use an API, which gets my Lego bricks inventory and returns all sets which I could build. Further it should show me sets, which I almost can build and show me missing pieces. Also e.g. it should show me sets with slight color differences, e.g. I have bricks in yellow but originally it requires orange. Things like that...

I think the first one should be very much doable, but I am not sure how I would build the "almost complete" and the "similar colors" features.

Edit: scratch that, just intersect the set of available pieces with each set. If there are fewer than n--or m%--missing pieces, suggest those.

After two weeks off looking for a solution, we’re going to vibe code something basic and use deepgram. If we just skipped the shopping around, the solution would be live already..

We thought long and hard of creating an SDK but did not go that route for now. Our main problem is that if you have to support Windows machines you're putting the finger into something really complicated.

And then depend on the end user's specs, etc.

So I'm really curious, in your case are you planning on deploying to Apple machines (m1+)? Or also set it up on Windows machines?

Will even auto detected who spoke, transcribe, video records and AI summarize it (with defined actions / next steps).

You can set it up to auto perform once the first person joins the meeting.

This, and there are also 3rd party bots that work with Teams (and presumably other video chats as well) that will join the meeting and do their own transcription. I've seen a couple of on meetings with vendors lately, it'll join and be called something like x company transcription bot.

Only read.ai has webhook support, but all these tools only post notes to integrations. we want the full transcript to generate extract of what we think is important.

Have not tried recall.ai

https://chatter-box.io/

I run a web service whose primary purpose is cleaning up messy, SEO-enshittified data from Google, eBay, etc. After years of fine-tuning my own heuristics, I threw a super-cheap LLM at it and it massively out-performed my custom code. It's slower, but the results are well worth it.

Those kinds of ideas can still become real businesses, especially if you have a great way to get awareness of your solution out to people who aren't actively looking for that, but you're better off trying to find problems where people have tried throwing money at every solution out there and are still yelling for help.

Tl;dr try to sell "pain killers" not "vitamins".

at Postman, we have a network for companies to reach developers and distribute their APIs on our platform. worth checking out: https://www.postman.com/explore.

you need to handle monetization, but the network gives you eyeballs

It's basically a service where you can download highly processed Bitcoin data in CSV, XLSX, or JSON formats, updated hourly.

I offer affordable subscriptions paid directly in satoshis to my self-hosted Lightning Network node — no intermediaries!

If you’d like to check it out: https://bitcoinisdata.com/

I’d say that this kind of projects are not different than any entrepreneurial endeavor, and the biggest challenge is usually acquisition, even though the technical part was / is hefty too.

It’s been some years already. Back then he had an office and a couple of developers. Seemed to be doing well. It has probably grown since then.

The Swedish employment office have a publicly available API for open positions. It is subsidized by their main branch because they see value in more companies / sites building business on top of their data.

https://data.arbetsformedlingen.se/

The business was very valuable across a lot of industries - gambling, encryption, advertising, security, adult entertainment, etc. - so there was a lot of demand that also helped smooth out the demand up/down cycles. If one market was cold, another was hot. But basically it's a lot of work and a lot of hand-to-hand combat. This is the best way to learn and get passionate customers. Show up, sell to them, and convince them they need you. You'll learn so much by doing this. And don't use the excuse that you're an introvert or not good at selling. If you want to be an entrepreneur, you need to learn and improve. No one is the best at anything on day one - you won't be either. But you'll get there if you keep at it.

Being the absolute best in the market meant that even having much more better funded competitors ($50m+ for competitors against our $12m in funding) meant we tended to win all the time. And before you ask, if I had this to do over again now, I could do this company for a LOT less money given how commoditized things are. I can tell you the time I almost spent $1m on a storage array until I found a cheaper vendor for $250k. Oh, that storage array was for 1TB of storage. So yeah.

Feel free to ask me anything. If there are enough people who have questions and want to do a chat I'd be happy to host a video call and get peppered with whatever questions you might have.

I’m just now really trying to put myself out there and get into sales to further develop our business. We’ve been lucky with some early network sales, but for the next chapter of our business (https://resonancy.io) I need to build the sales engine and I feel very much out of my depth.

Just telling myself its a practice and trying to chip away at it one week at a time.. At this point identifying potential customers and getting meetings is a real challenge. Been trying Apollo and doing cold outreach but is a slow grind. Also trying my best in linkedin..

What we did was more labor intensive and a function of these kinds of tools not existing (initial dot-com days). We would read the tech press, etc. and identify companies that might have a use for our tech. Then we would email the top person (guessing their email address usually - and most companies use a small number of variations for addresses) with a custom written, but short, tailored pitch to why they could use us and see if they wanted to talk. No pitch deck or attachments. Literally 2-3 sentences. Something that if even someone is going to quickly delete, something in those 2-3 sentences might catch their eye and make them reply instead of delete. We landed customers like Google (who we ended up later suing - that is a different tale), Doubleclick (who ended up being bought be Google), PayPal, Xbox, and a bunch of others through this method.

Happy to chat more. Startups are hard and all of us should help one another out to make the road a little less hard if possible. Keep at it - I know you'll get there. Hard work will get you there, shortcuts (usually) won't.

What I mean is: did anything ever break or behave badly ... and then when you investigated you discovered "Oh, look at what these people are doing ..." ?

It's doable, monthly updatable for free, but requires a bit of processing and understanding of the geo structures.

Does the Google Maps API not do this or is it too expensive?

An example of what you look up a postcode for (and the input data) would be useful.

1. How did you find your first customers?

2. How did you determine a price?

3. What are the payment plans based on? Usage? Flat rate? Something else?

4. How much are you currently earning MRR and what are your costs like?