When the cloud sneezes
”A computer once beat me at chess, but it was no match for me at kickboxing.” Emo Philips
Previously at the Volcano Base, I’d been thinking about boundaries and drawing a line in the sand. Since then, I’ve been wrapping up a complete business operating system overhaul. There has been a lot of coffee. But it's super cool.
Mission Briefing
When the cloud sneezes
Recommending software to a client is like recommending a café to a friend. You don’t need to know exactly how the coffee machine works, only that it won’t burst into flames when someone orders an espresso. If it does, nobody blames the wiring. They blame you for ever claiming it was “a nice little place.”
This week, one of my often-recommended tools collapsed in a heap at the exact moment a client needed it most. Not my fault, not their fault, just a reminder that modern software occasionally chooses to play dead in the middle of the road. Unfortunately, when it happens, my professional credibility is the one that gets scraped up with the roadkill.
The awkward truth is that every bit of software breaks. Every single one. Somewhere out there right now, a SaaS platform is wobbling off its chair while its users curse, refresh, curse again, and reach for the hammer. That’s not exceptional, that’s normal. The real calculation is simply: how much unreliability are you prepared to live with.
Which is why I take recommendations seriously. Painfully seriously. If I recommend software to a client, it’s because I’ve judged it to be sturdy enough for real work, not just pretty dashboards and marketing promises. I'm recommending it because it's not going to make me look stupid, too. But no tool is perfect, and no cloud is entirely storm-proof. When things fail, the best I can do is make sure my clients know I chose that tool for good reason, not blind optimism, and help them get through it to the best of my ability.
Classified Intel
Some interesting stuff I discovered on my adventures.
Anthropic has launched a controlled preview of a Chrome extension that lets Claude read, click and act on web pages while defending users from prompt-injection attacks. Trial access is currently limited to 1,000 users. (anthropic.com, simonwillison.net)
Why it matters: This is the kind of human-agent collision course we brace clients for: powerful automation inside our everyday tools, accompanied by digital safety minefields.
Genie AI. UK’s Legal Super-Assistant
Genie AI provides an AI-powered legal drafting and review assistant spanning over 120 jurisdictions, with strong privacy, security, and thousands of pre-built templates. (genieai.co, legaltechnologyhub.com)
Why it matters: It’s exactly the sort of tool that promises to optimise legal workflows without the usual vendor bravado, grounded in practical, contract-drafting muscle.
Building a Pro-Worker AI Innovation Strategy
he Trades Union Congress published a report urging that AI policy must centre worker empowerment, fair regulation, and equitable AI-driven innovation to avoid deepening inequality.
Why it matters: It’s a timely reminder that recommending software isn’t just about features. It’s about the broader social consequences, the context in which technology sings or stumbles.
Your data is not shared. Unsubscribe with 1 click.
Member discussion