Calmly News Killed

What it actually did

• A hard cap of ten. Not a suggestion, a recommended-reading section, or a wellbeing nudge. When you have read the ten, the feed stops serving.
• Summaries with the emotional loading removed. The adjectives that exist to make a headline hit harder are stripped; the facts of the story are not.
• Filters you set. Block a topic or a keyword you do not have the capacity for this week. It stays blocked until you unblock it.
• A reset after heavy stories. Some news is simply heavy. There are short breathing exercises built in for immediately afterwards, rather than a link to a different app.

What it was not

It was not a way to avoid knowing things, and it was not a mental health tool. Mental health is not about hiding from the truth; it is about having the capacity to handle it. A ten-story cap protects your capacity. It does not do anything else, and I would rather say that here than imply otherwise.

Questions

What happened when you finished the ten stories?

The feed locked for the rest of the day. There was no “load more.” That was the entire mechanism — the app was finished with you before you were finished with it.

Did it hide bad news?

No. It removed the adjectives written to make a story land harder, not the story. A plane crash still read as a plane crash. You could also filter specific topics or keywords, which did hide those, but that was a choice you made rather than something the app did on your behalf.

Can I still get it?

No. It was free on both iOS and Android while it ran, and it has been taken off both stores.

Was this a mental health app?

No, and it should not have been used as one. It was a news reader with a cap and a filter. If news is affecting you in a way that a shorter feed does not fix, that is a conversation for a professional, not an app.

Post-mortem

Killed products keep their URL here and gain this instead of a redirect.

What I believed

That news fatigue was a real, documented problem and that people would want a reader which handled it for them. That part I checked: I went looking for other people describing it in their own words, on forums and public threads, before I built anything, because I have shipped features nobody asked for before and I did not want to do it again. Then I found several apps already working in this space, decided the category was therefore proven and lightly educated, looked at what was shipping, judged that I could do better — and concluded the opportunity was large. The research was sound. The inference I drew from it was not. Existing products told me the problem existed. I read them as telling me there was money in it, and those are different sentences.

What I built

All of it, in three weeks, with three people. A designer in Vancouver on interface and product, me on development and testing, and a third person on marketing. Daily scrum, week one for a bare MVP, week two for features, UI and positioning, week three to ship. What went out was a website, an iOS app, an Android app and our own backend pulling the news. Every story was vectorised on arrival, every user wrote their interests out in plain text and was vectorised too, and the day's stories came back ranked by relevance as a deck of cards you swiped. The news API was free — I wrote a polling schedule that sat exactly on the ceiling of the free tier. The domain was $15 a year and the AI calls ran one to two dollars a day. Nothing about the build went wrong. Technically, procedurally, on schedule, it is the smoothest three weeks I have run, and it is the reason this post-mortem exists: doing the wrong thing extremely well is still doing the wrong thing.

What the evidence said

Seven and a half months, from 12 January to 31 August 2026. On the App Store: 723 impressions and 18 downloads. On Google Play, live from May: 2,400 impressions and 2 installs. Three thousand-odd impressions and 20 downloads in total, fewer than five people who actually used it, and nobody who stayed. The conversion rate was around 2.5%, which is not the problem — 723 is the problem. Seven and a half months of storefront works out at three people a day walking past. Marketing was a handful of Reddit posts and nothing else. There was also no way to make money in the app at all, and not because I decided against charging: making money was never a question I got round to asking, which I think shaped everything downstream of it.

What I would do differently

Reverse the order. Put up a landing page, do the SEO, collect a waitlist, and only build once there is evidence I can bring people to a page and hold them there. Whether I can generate and carry traffic is the first thing that gets tested, not the last — if I cannot get anyone to a landing page, finishing the app changes nothing except how much it cost to find out. I would also check willingness to pay directly, with something like Sensor Tower on the products already in the category, instead of treating their existence as proof of a market. I used to believe good wine needs no bush — that if the thing is good enough, people find their way down the alley to it. In an alley nobody walks down, the wine does not matter. Mine had three people a day.

The myth it broke

That bringing in a domain expert raises your odds. Our marketing lead came out of journalism, and it did not help — for two reasons, both of them mine. First, I put someone who should have been on product onto marketing, which is an allocation decision I made and got wrong, and the communication overhead that followed meant we could not pivot quickly. Second, and this is the part worth keeping: the product started from me. The direction and the assumptions were already set before they arrived, so the ground had a shape by the time their understanding of the industry showed up. Their insight had nowhere to grow. That is a structural problem, not a personnel one, and it means expertise only pays off when the thesis starts with the person who has it.

Other killed products →