How we review

Most software reviews exist because someone wanted the traffic. This page describes what these reviews are for, how a verdict is reached, and the specific ways money is kept away from it.

The rubric

Every tool is assessed against the same four principles this studio builds by. They are narrow on purpose. A rating here is not a general quality score — plenty of excellent software fails these tests, because it was built for a different world.

1. 01

   Instant

   Local software responds at the speed of the device.

2. 02

   Private

   Sensitive work does not need to leave your machine.

3. 03

   Resilient

   Core functionality continues when the connection disappears.

4. 04

   Focused

   Purpose-built tools without feeds, engagement loops, or platform clutter.

Each principle gets one of three ratings: MEETS, PARTIAL or FAILS. There is no total, no score out of ten and no ranking. Aggregating four unrelated questions into one number destroys the only useful information in them.

A tool that fails every principle can still be recommended in the prose. The scorecard says what it is; the writing says whether it is worth using.

Testing rules

These are the conditions under which a review gets published.

• The tool is paid for. Full price, ordinary licence, no press account. If a maker offers a free copy, the review either does not run or discloses it in the specification block.
• It is used for real work for at least two weeks before anything is written. Not opened, evaluated and screenshotted — used, on the machine, for the job it claims to do.
• The exact version is recorded in the specification block, along with the operating system and the length of the test. Software changes; a review is a measurement taken at a point in time, and it says which point.
• Privacy claims are checked, not repeated. Where a tool claims local processing or no telemetry, network activity is observed rather than taken from the marketing page. Where that check was not possible, the review says so.
• Offline behaviour is tested by going offline. Not by reading the feature list.
• Corrections are made in public. If a verdict turns out to be wrong, the review is updated with a dated note explaining what changed. Reviews are not silently edited.

The money

Some reviews contain referral links. Every review states which, at the bottom, whether or not it has one. The rules around them:

• The link is added last. The assessment and the ratings are written before anyone checks whether a referral programme exists. A programme's existence never determines what gets reviewed.
• Negative reviews keep their links. If a tool has a referral programme and fails three principles, the review still runs and still says so. A referral link that only ever appears next to praise is an advertisement.
• Referral links are marked in the page itself and tagged rel="sponsored nofollow" so search engines are told what they are.
• No paid placement, ever. No sponsored reviews, no payment for coverage, no payment for a position in a list, no agreement to show a draft to a maker before publication.

If a referral programme ever made it hard to publish an honest verdict, the link comes out and the verdict stays.

Our conflicts

Out of Browser makes and sells software. That is a conflict of interest and it does not go away by being careful.

So: our own products are never reviewed here. Where a reviewed tool competes directly with something we sell, the review says so at the top. Where we have used a maker's work to build our own, that is disclosed too. You should read a review of a competitor with the appropriate suspicion, and the specification block is there so you can check the reasoning against the facts rather than trusting the conclusion.