What we have learned
Anyone can collect failures. The part worth keeping is what the person who lived through one would tell you before you start the same thing. Each of these is our reading of what they wrote, with their words a click away.
17 of the 23 buried here came with one. The rest are a story without a moral, which is also how it goes.
Create an MVP and ship it, don't go too far with a client you don't have yet.
A personalized dashboard · antionmil
A better tool still has to beat the inertia of the ones people already have. The first question back was why anyone would switch from pip.
PyFlow, a Rust package manager for Python · the__alchemist · 2025
The only consumers were the client libraries the same team controlled — so the generality bought nothing, and cost processing time on every request.
A company API rebuilt around HATEOAS · gsibble · 2025
Technically impressive and almost unused, because the people it was for had never been properly understood.
18 months building for users they never met · onagengast · 2022
Unless you are going to put full-time marketing and sales behind a SaaS, you will have a very hard time beating the companies already there. Working well in-house is not the same thing as competing.
Testedrecruits and RomyLMS, sold as SaaS · silexia · 2021
Every third-party service the site stood on was a piece of it somebody else could switch off. Over 8 years, enough of them did.
A book price comparison site built on 8 APIs · bckygldstn · 2020
More than one company answered the pitch with 'why would we target them?'. The other two reasons they gave themselves: the wrong people were approached, and approached clumsily, by somebody who says plainly they are not skilled at marketing.
A $100-a-listing job board for over-35s in tech · stakkur · 2020
Find the customers first, then build. 2 years the other way round produced nothing to show for it.
2 years on a SaaS with zero sales · jcadam · 2019
Check the upgrade path before betting a company on a framework. The exit cost is the part nobody evaluates.
A startup built on the Sails framework · ruffrey · 2018
Whoever owns the platform has to be committed to it for at least as long as you are. Months went in before anybody asked whether Mozilla intended to keep XUL alive.
A desktop app built on Mozilla's XULRunner · _pdp_ · 2016
A second module system layered on one that already works is pure overhead — and it is charged on every change, not once at the start.
RequireJS on top of Angular · dceddia · 2016
The rough panel on the device beat the better tool somewhere else, because it sat where the work was already happening.
yhs, an embeddable HTTP server in C · to3m · 2014
A poll saying yes is not demand. After the first show of hands nobody was asked again — 6 months later it turned out the yes votes were being polite.
A hosted SaaS built on the strength of 1 poll · quaffapint · 2014
Shipping on a billion devices is distribution, not use. The two are easy to mistake for each other on a CV.
The Openwave WAP browser kernel · urbit · 2013
The problem was never the marketing — nobody wanted the thing. What worked afterwards was the reverse order: find what people are already asking for, then build around that.
A hosted CMS for developers · saturnflyer · 2012
The engineering was never the hard part. Getting anyone to use it was, and that work had not started.
A multiplayer drawing applet, Java, 2000 · bumblebird · 2009
Build the front end first and see whether anyone buys the subscription. If they do, grant them the one-off version by hand. A month of billing code before the first sale answers the question far too late.
Subscription billing for a $10 upgrade · jmathes · 2009
Got one of your own? Register the death and say what you would tell somebody about to do the same.