Filter for Telegram

Made a billing integration with Stripe. Filter allows adding many plans and bind them to telegram accounts. So this feature added a lot of UI complexity.

Since the previous update, I also built a proof of concept project that uses a webhook feature. In the process, implemented an export feature for Filter. This feature gives old messages in JSON format so client projects can bootstrap some history.

Did some technical improvements. Added more error handling, alerts so I would know when something goes wrong.

Send with Sengrid email to 40 registered users. And got one paying customer.

Now is time for marketing.

Got a Delaware single-member LLC - SITG LLC SITG - is 'skin in the game.' I'm a big fan of Nassim Taleb's philosophy.

I initially tried to get Stripe Atlas. But they don't accept non-us people for LLC. A bit bummer as I wanted to get free AWS credits.
I got in minutes EIN. Thanks to my time in SF where I got SSN so I could file for EIN online.

I applied for a business bank account in Mercury bank and waiting for their reply. When I get a bank account, I'll register on Stripe and will work on billing integration.

• built landing page https://filter.sitg.app

• 80% time/effort spent on animation - recreated Telegram mobile interface to explain how project works. The final result looks cool

• took open sourced policies from basecamp and adapted to my case https://github.com/basecamp/policies (learned that it's quite a task to make proper policies, gdpr compliance)

• spent some time learning Kotlin (Java was very verbose) and JVM dev environment

• made feature parity with the TypeScript version. This time project, UX felt like magic as forwarding was instant.

• started web interface

My original specialization is Front-end, so here I felt a real flow. I tried to use some external component libraries on top of React to built UI. But I mostly ended up implementing my components.

Any way this phase took a long time as building interfaces is not a fast process. I spent around 500 hours on UI.

• made backend dashboards
• created CI tasks
• showed to couple of friends demo, found some bugs immediately (took couple of weeks to fix them all)

• learned how to extract needed information using TDlib (not an easy lib to work with when you are new to it)
• made an MVP on top of AWS lambdas that check periodically new messages
• created mocks of UI that should configure filters

So I made a working version. And then I understood that I don't like how product works: I don't like a minute delay between posted and forwared message. I also didn't like that cost for many users on top of lambdas would be too high. I also didn't like that node.js don't support multi-threading and can't launch several TDlib instances.

I decided to make a server that holds all TDlib processes. It had to be a JVM (Java) based solution as Java has the best integration with TDlib.

I was researching ML for finances (portfolio management problem) and found that sentiment analysis is very important for automated trading decisions.

I knew that Telegram has many private/paid trading groups so I decided to build a tool that can help to extract some valuable insights from such groups.

I also thought about a use case where you want to read specific users/messages in crowded groups — this problem the product had to solve as well.