mehmet_mhy

In early 2023, Dylan Eck and I (Mehmet Yilmaz) built Notify Cyber because we were frustrated wading through fragmented cybersecurity news sources. We built a platform that aggregated news from 7+ sources, cleaned the data, summarized articles using OpenAI's API, and presented everything in a clean, searchable interface.

The launch was humbling. We launched in June 2023 with almost zero daily visitors. We spent the first two weeks doing Twitter marketing but got zero new users. Dylan and I were discouraged and we were going to can the project but as a last resort, we decided to make one final push on Reddit.

After days of back and forth, Dylan got a single post approved on and posted on r/cybersecurity. And within 3 days, we had 8k+ visitors and 100+ people on our waitlist. That post eventually hit 65k+ views with a 96% upvote rate.

Over two years, we grew to 17k+ unique visitors and 43,000+ page views, all through organic growth. We also increased the waitlist count to 160+ within that time range. We even optimized ruthlessly, reducing monthly costs from $38+ to just $1.85 by partly hosting key services of the site on a Raspberry Pi in my home.

We proved there was real demand for what we built, but we couldn't crack monetization or find a real business model. More importantly, both Dylan and I reached a point where the opportunity cost became too high. On October 5, 2025, we decided to retire the platform and open-source the entire codebase.

What we learned through this was that timing and distribution matter. The MVP can be scrappy, but execution must be clean. Build what you enjoy solving. Real demand doesn't always translate to a valid business model. And there's dignity in retiring/pivoting a project.

As for what's next... I'm currently focusing on solving some interesting problems at Charter and I've got some exciting projects on the horizon. But until then, I want to close this chapter with Notify Cyber. It has shown me a glance of what product-market fit looks like and how you don't need that funding and fancy marketing to make an impactful project/product. Bootstrapping might be the way for most software based products and I am carrying that lesson forward into everything else I build moving forward.

If you're building something, don't be afraid to fail publicly. Keep shooting, iterate fast, pivot as needed, and know when to close one chapter so you can start the next one better.

Special thanks to Dylan Eck for driving the frontend and our Reddit success.

Links:

I built Ch to bring AI directly into the terminal, offering fast, lightweight functionality that remains fully under user control. Most AI tools feel detached from developer workflows, but Ch is designed to integrate seamlessly, with no hidden automation or costs.

Vision:

Ch aims to keep AI interactions explicit, transparent, and user-driven, empowering developers to use LLM capabilities without leaving the terminal or surrendering control.

Highlights:

- Single Go binary: Starts instantly, streams responses in real time.

- Multi-provider support: OpenAI, Amazon Bedrock, Anthropic, Gemini, Groq, xAI, Mistral, Together, DeepSeek, and Ollama.

- File + web input: Load PDFs, DOCX, images (OCR), scrape URLs, or analyze YouTube content.

- Unix-style integration: Pipe command outputs or files directly; colors auto-suppressed for clean pipelines.

- Developer-focused: Record shell sessions as context, export code snippets, integrate with editors or other AI CLIs.

- Transparent & explicit: No background actions or silent model switches.

Install in one line: