Writing.
Notes on the problems we actually work on: proving where software and writing came from, the boundaries of workplace monitoring, and keeping identity out of the products that depend on it.
AI24 Aug 20266 min read
Claude started signing its sentences. We have been signing our builds for years.
Two different industries arrived at the same answer within a decade of each other, and the second one is about to relearn every lesson the first one paid for. Here is what code signing already knows about provenance.
Engineering24 Aug 20265 min read
Why your code is barely watermarked
The watermark hides in the choice between words that mean the same thing. Code does not have many of those, which has a practical consequence for anyone hoping to detect AI-written source.
Product24 Aug 20266 min read
What AI watermarks actually mean if you run a business
Not a legal opinion and not a panic. A short, practical account of what changed on 2 August 2026, what it means for content you have already published, and the one policy worth writing down.
Security23 Aug 20266 min read
Why we sign every build, and what a signature actually proves
A code signature does not tell you software is safe. It tells you who is accountable for it. That distinction is the whole point, and it is the reason we treat signing as infrastructure rather than a release-day checkbox.
Product23 Aug 20267 min read
What a workplace productivity tool should not collect
Software that observes how people work carries an obligation most tools in the category do not take seriously. Here is where we think the boundaries sit, and why some of them are enforced in code rather than in policy.
Engineering23 Aug 20265 min read
One identity layer, three products, and why we did it the hard way
Every product could have grown its own login. We separated identity into its own service instead. That decision cost more up front and is the reason the other products stayed simple.