Author: Johnson Frank


  • You Probably Need Three AI Tools, Not Thirty

    I have a small graveyard of AI accounts. Thereโ€™s one tool I joined because it promised to organise every thought I had. Another was supposed to turn meetings into โ€œactionable intelligence,โ€ which is an impressive phrase for writing down who agreed to send the spreadsheet. A third could generate presentations, websites and social posts. I…

  • Vector Databases Are Useful. Your App May Not Need One Yet

    Vector databases have become one of those technologies people add to project diagrams before they have a project. The pitch is seductive. Your AI assistant will understand company documents, find the right paragraph instantly and answer questions like a colleague who has read every Slack thread since 2019. So the team creates embeddings, opens an…

  • The Most Dangerous AI Answer Is the One That Sounds Finished

    An AI once gave OpenAI researcher Adam Tauman Kalai three different titles for his own doctoral dissertation. None was correct. Asked for his birthday, it supplied three different dates, also wrong. The interesting part wasnโ€™t the error. Search engines return bad results every day, and colleagues misremember things all the time. What stood out was…

  • Privacy Usually Disappears One Convenient Click at a Time

    A photo-editing app once asked for access to every photo on my phone. I wanted to remove a shadow from one image. The request felt absurd, but the Allow button was large, the alternative required another screen, and I was in a hurry. That is how many privacy decisions happen. Not after careful thought. Not…

  • AI Has Started Asking for the Keys

    For the first few years of the generative AI boom, the arrangement was simple. You asked a question. The AI answered. Then everyone went back to doing the actual work. Now AI systems are asking for access to calendars, files, email, browsers, databases and business tools. They do not merely want to suggest the next…

  • The Cloud Is Far Away. Sometimes That Matters

    A security camera outside an office noticed movement at two in the morning. The video travelled to a remote cloud service, which analysed it and sent an alert. By the time the notification arrived, the moving object had left. It was a plastic bag. No disaster. Just an example of strange technology architecture. A camera…

  • The Dead Zone Is Starting to Look Temporary

    There is a particular kind of silence that arrives when your phone loses signal somewhere you actually need it. The map stops updating. Messages sit with a tiny clock beside them. You raise the phone toward the car window as though the extra thirty centimetres might negotiate with the mobile network. For decades, dead zones…

  • AI’s Most Physical Problem Is Electricity

    AI is usually presented as something weightless. A prompt goes into a box. Words appear. Somewhere, apparently, there is a cloud. The cloud has buildings. Inside those buildings are racks of computers drawing enormous amounts of electricity, producing heat and requiring cooling, networking, backup systems and actual land. The rapid growth of AI has made…

  • Quantum Computers Don’t Need to Arrive Before Security Teams Get Busy

    There is a security project that sounds as if it belongs ten years in the future and somehow creates work today. Post-quantum cryptography has that problem. Mention quantum computers in a normal office meeting and half the room expects science fiction. Yet the practical work is much less dramatic. Companies use encryption everywhere: websites, VPNs,…