Anthropic and Google are watermarking AI-generated text, OpenAI has announced plans to follow. We explain why text watermarking is a fundamentally flawed paradigm: copyright paradoxes, false positives, a backdoor for user tracking, degraded output quality, and a detection system that catches honest users while giving bad actors a free pass.

We have released Agent Redactor, an MIT Open Source local security proxy that detects and redacts PII, credentials, and proprietary data from AI agent API requests before anything leaves your machine. We explain why we think the indifference toward sending sensitive data off-premises is dangerous and why open source is the right foundation for privacy software.

In the span of 8 days, we were banned from X, Reddit and Threads. 12 Days later we were banned from Linkedin. All of these accounts are around one year old or older. Each platform has employed different methods of deception, misdirection and outright lying.

A 58-year-old mathematical puzzle was solved by GPT-5.4 Pro in a single prompt. We explore what this means for the nature of creativity itself, from Edison's lightbulb to Shakespeare's plays, and ask whether synthesis is fundamentally any different when done by a machine.

Our first blog post. We are husband and wife builders of Negative Star Innovators. Here we present our thinking, discuss our achievements and failures, and build trust with our readers as we release products under the Negative Star Innovators brand.

We have released Paste Redactor, a tool that redacts Personally Identifiable Information (PII) from your clipboard. It processes redactions 100% locally which means your clipboard contents never leaves your device. We discuss who would use it, why the feeling of indifference is dangerous, and where we are now on our privacy journey.