How AI Watermarking Works? Can It Be Removed?
AI watermarking is not baked into model weights, it works at the generation level, by sampling different tokens, paired with a private key.
Decision tools and frameworks for autonomous living
AI watermarking is not baked into model weights, it works at the generation level, by sampling different tokens, paired with a private key.
China is already way ahead in the AI race, more than the Western media make it look. This is an account from an actual visit there.
"AI watermarking now traces Claude text back to you. Learn why it survives copy-paste, how it locks you in, and two ways to keep your writing [...]
We're not in the Singularity yet. The chevalier-in-the-supermarket test shows what real Singularity would feel like—and why AGI hype falls short.
I'm fortunate enough to have witnessed a few key moments in our recent tech history. One of them is web 3, the dot com revolution. Back [...]
See how to combine entry-level AI subscriptions and free models for Fable-level vibe coding, solid limits, and high productivity on an $80 monthly budget.
Publishing ten apps in four months sounds good. And it is good. It means the bottleneck is no longer building the app. With AI-assisted coding, small [...]
Pre-digital childhood gave you patterns younger developers never learned. Now AI removes the syntax barrier. Here's how to turn 50+ into your unfair advantage.
I kept the Claude Code interface I like and routed Grok's model underneath it via proxy. Same workflow, video generation included, $35/month instead of $100. Here's [...]
I shipped 9 iOS apps in 4 months with AI—and learned how to manage AI cognitive overload. Here are 3 actions that keep AI-assisted development sustainable [...]
Running local private AI models is now realistic, affordable, and urgent. Here's what hardware to use, which open-source models are ready, and why you shouldn't wait.
Harvesting attention isn't just for big tech. Learn how AI wait times created a new micro-pause — and why you should claim it before someone else [...]
I stopped prompting my AI agents and started writing loops instead. Here's what a 50-line Python audit loop taught me about trust, failure, and where humans [...]
After using Claude (Opus 4.6 to 4.8) for roughly 5 months, I'm experiencing with MiniMax 3. A full month to decide which is best for coding.
In 2010 the OS metaphor for self-improvement was an analogy. In 2026 it's literal — AI tools actually run as cognitive extensions. Here's what that changes.
10 iOS apps in 4 months, built solo with AI. Vibe coding for senior developers: specs first, two models, heavy verification, and marketing in parallel.
In 2018 the wrong tool was obvious — you could see the mismatch. In 2026, AI made it invisible. Here's where the granularity problem actually went.
Michael Grady built a production AI system for neurodivergent professionals using my Assess-Decide-Do framework and AIGernon on Zo Computer. Here's what he shipped.
Claude Mythos may indeed by a good model, but there's a lot of fear mongering around it. Maybe it's connected to the upcoming Anthropic IPO?
I turned 1,400+ blog posts into a conversational wiki powered by a Cloudflare Worker and an LLM. There's a button at the bottom right of my [...]
I've been blogging for 20 years
ask me anything