the noise problem
i used to research product ideas the wrong way. open reddit, search a vague keyword, skim forty posts, bookmark twelve of them, and close the tab three hours later with nothing actionable to show for it.
the signal was always there. people complaining about the same workaround in five different threads. a recurring phrase that showed up in every frustrated comment. a clear gap that nobody had productized yet. but the noise-to-signal ratio was brutal, and the manual process didn't scale past "i'll figure it out."
i talked to other founders doing the same thing. same problem, different niches. hours of manual scrolling, no structure, just vibes and luck.
"we know the answers are in the conversations. we just can't spend two days digging through threads every time we want to validate an idea."
building for signal
so i started building something for myself. a pipeline that could pull posts from reddit, bluesky, and google search, score them for pain and relevance, deduplicate the noise, and cluster what's left into themes.
the first version was rough. hardcoded queries, no ui, output piped into a text file. but it was immediately more useful than three hours of manual browsing. i ran it before every product decision. it worked.
i kept shipping it until it had a proper web interface, configurable queries per project, a scoring system, and a report generator that would turn selected posts into quoted pain themes and product angles.
why self-hosted and source-available
when i build on a tool, i want to know what it does with my data and my api keys. i want to read the source before i depend on it. i want to run it locally and own the storage.
so that's exactly how threadlens ships. the full source is on github. the installer gets you running in one command. data stays in local sqlite. you control the api keys for every integration.
there's a docs site that covers setup, architecture, licensing, and storage in detail so the landing page doesn't need to repeat the manual. inspect it before you trust it.
why i still use it every week
i set a simple bar for every feature: if i don't use it myself before a product decision, it doesn't ship. threadlens runs before i commit to a niche, before i scope a feature, and before i decide if a pain point is worth a solution.
good research beats clever guessing every time. that's the whole point.
got questions or just want to chat? send me an email — i read every one.