Martin Keywood·Oct 14Has Someone Really Found the Point Where an AI “Goes Bad”?I came across a paper recently with a title that was pretty much guaranteed to get my attention:
Martin Keywood·Oct 2DeepSeek 4.1 Flash Might Be More Important Than You ThinkWe’ve Been Obsessed With Bigger Models. DeepSeek Just Made Memory Requirements A Lot Smaller.
Martin Keywood·Sep 23Five AI Answers to Your Child’s Homework. Two Are Wrong. Can You Tell Which?I’m exploring a simple idea: helping non-IT people become comfortable with AI, not by teaching them how it works, but by teaching them /…
Martin Keywood·Aug 13Someone Asked an AI to Book a Gym Class. It Accidentally Exposed the Next Big AI ProblemWhat looked like a harmless booking request may become one of the most important AI stories of 2026.
Martin Keywood·Aug 2MCP, A2A and the Rise of Agentic AI: Are We Finally Building the Internet of Intelligence?A few months ago, I found myself trying to connect multiple AI systems together. Sounds simple, right?
Martin Keywood·Jul 22The AI Bubble Nobody Wants to Talk About: What If China Doesn’t Need to beat the USA in the AI race?The real disruption isn’t that Chinese AI is catching up with the USA. It’s that nobody may care who comes first anymore.
Martin Keywood·Jul 10DSpark Isn’t a Better Model… So Why Is It Making LLMs 80% Faster?Or: are we quietly entering the “post-model innovation” era?
Martin Keywood·Jun 23AI Agents Are Becoming Synthetic Employees — And We Are Nowhere Near Ready to Manage ThemIt’s been a while since I’ve written something like this.
Martin Keywood·May 31Can AI Use Computers Like Humans? (Surely That Would Change Everything !!)This is one of the most important emerging areas in AI right now.A response icon1A response icon1
InGoPenAIbyMartin Keywood·May 19Thinking Machines Lab Introduces Interaction Models — Are Turn-Based AI holding us back?I’ve spent the last few years building, testing, and occasionally swearing at AI systems that almost feel natural to use.