dif.sh keeps feature flags and A/B tests as markdown files inside your codebase. Most tools in this category describe what their dashboard can do, while dif.sh opens with a claim about where flags should live in the first place. It's a principled stance that challenges what a developer already assumes about how this stuff is supposed to run before offering the simple fix in the same sentence.

The subheading carries the explanation: one .md file per flag, versioned in git, reviewable in a PR, and read by your coding agent at the start of a session so prior decisions travel with the work. That last part is why the argument holds up now, because a typical agent won’t open a vendor dashboard and read what you decided last month.

A headline that argues will lose you the readers who disagree with it, which is usually a good trade, since the ones who stay have already accepted your premise before they read a single feature.

Why this H1 works:

• Takes a principled stance
  a claim about where flags should live, which a reader can agree or disagree with

• Challenges a working assumption
  it assumes you keep flags in a vendor dashboard and tells you that's the wrong place for them

• Plain category language
  "flags", "experiments" and "repo" are all a developer needs to place the product

• Self-selects the audience
  anyone happy with their current setup leaves early, which is cheaper than converting them later
  
  

Follow David Herzog on Product Hunt

Visit dif.sh ↗︎