The story

I have a confession: I can't read long articles.

And I definitely can't watch 2-hour long videos. My attention span is basically a sieve, and my browser is a graveyard of "tabs I'll read later." We both know I never read them.

For about a year, my workflow was painful. Every time I found a video I wanted the "gist" of, I'd hunt for a free transcript tool, copy the entire wall of text, and paste it into ChatGPT or Claude with a messy prompt. It worked, but it felt like extra work to avoid work.

Then the AI world exploded. Models started reading videos directly. In early 2025, I spent two weeks building the first version of this. I called it Key-Insights. It was a disaster. Google blocked my VPS from grabbing transcripts, and I hit a wall. I failed, so I quit.

Fast forward to early 2026. I decided to try one more time. This time, with some help from the latest AI (thanks, Gemini), I rebuilt it from scratch. I called it Fewwords.

It's not just a "make text shorter" tool. It's smart. It doesn't just summarize; it understands intent. The engine runs through a multi-stage process: it extracts the raw data, classifies exactly what kind of content it's looking at, and then picks the most relevant strategy to give you a summary that actually makes sense.

To make this work, I built three distinct pipelines: one for articles, one for short videos (up to 20 minutes), and a heavy-duty one for long-form videos (up to 3 hours). Reddit support is coming next.