We report on how games get made, sold and kept working. Not release dates, not trailers, not rumours with a question mark on the end.
Supply, silicon and the physical objects games actually run on.
Hardware
Everyone is still repeating a supply story from 2023. Three manufacturers told us the bottleneck moved downstream to advanced packaging eighteen months ago, and that nobody has updated the messaging because the old explanation is easier to give.
Hardware
Screws, adhesive, part availability and whether the battery can be replaced without heat. The results do not line up with the marketing, and one device that scores badly on teardown turns out to be the easiest to actually keep alive.
Advertisement
How the work gets organised, funded and occasionally undone.
Industry
Two years of sprint data from a thirty-person team that moved to thirty-two hours without cutting pay, including the eight months where it clearly was not working and what they changed.
Industry
Teams of forty to ninety people used to make most of what we reviewed. We went back through eight years of credits to work out where those teams went, and the answer is not simply that they closed.
Measured, not asserted. We test the defaults, because defaults are what most people use.
Accessibility
We measured the out-of-the-box caption size, contrast and background opacity of forty releases across four years. The median caption has shrunk every year since 2023, even as the settings menus have grown.
Everything in this section is reported by a named person who did the work. We do not aggregate, we do not rewrite other outlets with a fresh headline, and we do not publish a story built entirely on one anonymous source with no documentation behind it.
Where a story rests on figures, we say where the figures came from and how we handled them. The subtitle measurements, for example, were taken from clean installations on default settings at 1080p, photographed at a fixed distance, and measured in pixels by two people independently. When our two measurements disagreed by more than five per cent we threw the reading out and did it again. That is a tedious method and it is the reason the piece took five weeks.
We grant anonymity when someone would lose their job for speaking, and not otherwise. When we do, the article says what category of person the source is and why they are in a position to know. A quote with no attributable context is not evidence, it is decoration.
If we get something wrong we correct it at the top of the article, not in a footnote, and we date the correction. The original wording stays visible with a line through it where the change is material. We have issued four corrections since launching, and all four are still listed on the about page, because a corrections log you can quietly empty is not a corrections log.
We do not run leaks, unannounced hardware rumours, or stories whose only public interest is that someone wanted them kept quiet for commercial reasons. We also do not cover individual people who are not public figures. The test we apply is simple: does knowing this help someone decide what to buy, what to play, or how the industry that makes it actually works. If the honest answer is no, it does not run.