Hi all,
Apologies for no email last week, I was in the midst of moving out of Dubai (just in the nick of time it would now seem).
I’m now in the UK for 5 weeks, culminating in my wedding at the end of March, which is very exciting! It seems very apt that this week I wanted to talk to you about some new AI my Head of AI, Christina has been testing and implementing for us, because right now my time is a little more limited than usual with everything going on with our move and wedding.
Firstly, I think it's worth telling you that I'm not actually typing this. We found a new tool called Wispr Flow, which downloads for free on your Mac or your PC. It allows you to just hold down the fn button and talk into the computer, which then turns it into text. Effectively, it is a transcribing tool, but it's doing it in the exact same way that I would usually write as well as summarising when you take pauses and taking out any filler words like ums and ahs that you might usually leave in - this is supposed to help you type around 3.8X faster than usual by transcribing instead of typing.
Actually, this is quite a good test for it. I wonder if it will leave in the ums and ahs that I'm using as examples, or is it going to delete them? (It left them in!).
Christina tests 2 new tools a month in our business and then includes them in our internal client monthly newsletter (The Recognized Digest) with all the experimentation data so they are the first ones to benefit from tools we have positively tested in our business.
Meet the AI "Jarvis" That Broke the Internet, and the Rules
Imagine texting your computer from across the world and coming back to find your emails answered, your calendar sorted, and your flights booked. That's OpenClaw and it's the hottest thing in tech right now.
Built by Austrian developer Peter Steinberger and originally called Clawdbot, OpenClaw runs locally on a Mac mini, connects to your messaging apps, and operates autonomously around the clock. Think Iron Man's Jarvis, except you can actually have one. It's driven such a frenzy that Apple Mac minis are now backordered for up to six weeks.
So how does it actually work? You set it up on a Mac mini, connect it to WhatsApp, Telegram or iMessage, and start chatting with it like you would a human assistant. You can say "watch my inbox and flag anything urgent," or "book me a flight to Austin Friday morning" and it goes away and does it. Your memory files, preferences and history all live as plain Markdown files on your own hard drive, meaning your data never touches a corporate cloud. Users have used it to manage family businesses, replace Zapier automations, unsubscribe from email lists, and even have it create its own new skills on demand. The learning curve is real, but once it clicks, people describe it as getting a superpower. People have replaced whole offices full of employees with this one Mac Mini and OpenClaw.
That said, the safety concerns are serious. Because OpenClaw needs deep access to your computer, files, calendar and messaging apps to function, it's a significant attack surface if misconfigured. Cisco's security team tested a third-party OpenClaw skill and found it could silently exfiltrate data and inject malicious prompts without the user realising. One of OpenClaw's own maintainers bluntly warned on Discord: "If you can't understand how to run a command line, this is far too dangerous for you to use safely." In one widely reported case, a user's OpenClaw agent independently created a dating profile and started screening matches without being asked. Over 21,000 exposed instances have been identified online. This is genuinely powerful technology, but right now it rewards the careful and punishes the careless.
But there's a catch at the platform level too. OpenClaw was gobbling up millions of AI tokens through unofficial "OAuth" backdoors, essentially exploiting flat-rate subscriptions to run unlimited compute. Anthropic, whose Claude model powered much of it, shut down third-party access and forced a name change after legal action over the original "Clawdbot" branding. Google followed, banning users without warning, costing some people access to their entire Google accounts.
So when should you get involved? Right now, OpenClaw with OpenAI's backing is the moment to pay close attention, but not necessarily to jump in. If you're a developer or technically confident early adopter, the window to experiment is open and the upside is real. For everyone else, the smarter move is to watch what OpenAI does next. Steinberger has been brought in specifically to build the next generation of personal agents, meaning a polished, consumer-ready version of everything OpenClaw promises is likely coming, with proper guardrails and without the risk of losing your Google account overnight. The technology is ready. The infrastructure around it isn't quite yet. Give it six months, and this could be as mainstream as ChatGPT.
And the real twist? OpenAI hired Steinberger. He's now there to "bring agents to everyone," with OpenClaw moving into an open-source foundation backed by OpenAI. What started as one developer's side project has become the new AI battleground and the race to own the personal agent future is well and truly on.
Stay curious. Stay sharp.
FP