The Medicare Hack Wasn't a Security Failure, It Was a Scope Failure
OpenAI's agents got into Australia's Medicare portal while chasing an open-ended task. The real lesson isn't about firewalls, it's about how loosely the job was defined.
Notes from the build
What we are shipping, what broke on the way, and what it changed about how we build. Written from real projects, not from a content calendar.
A Cheap Confidence Score Doesn't Replace the Decision You Still Have to Make
Jev returns a floating point confidence score instead of text. That's useful for classification, but it doesn't remove the judgment call, it just moves where you have to make it.
Rabbit's New AI Agent Says It'll Figure Out What You Need. I'd Want to See It Fail First
Rabbit's new agentic OS promises to automatically work out what it needs. I've spent enough time building automated systems to know that's precisely where they go wrong.
Private AI for Business: Why I Stopped Pasting Company Data Into Public Chatbots
Founders keep asking me for the best private AI for business without knowing what they're actually protecting against. Here's what I've learned running one on my own company first.
An AI Receptionist Isn't a Script Problem, It's a Tone Problem
The hardest part of building an AI receptionist wasn't the script or the technology. It was making the voice sound like it cared. Here's what that means for what to automate and what to hand off.
Answer Engine Optimisation: What It Actually Is, and Why Most Advice On It Is Guesswork
Everyone is suddenly talking about answer engine optimization. Here is what it actually means, how it differs from ordinary SEO, and what I'd check first based on the automated content and reporting work I already do.
Building a Website That Sells a Service Business: What I Do First
Before design or copy, I decide two things: where the enquiry goes when someone fills in a form, and what questions the site is actually built to answer.
Agency or Independent Consultant: Who Should Actually Build Your Automation
An honest look at agencies versus independent consultants for automation work, focused on the two questions that actually decide whether it lasts: who owns it, and what happens when it breaks.
Muse Beat ChatGPT's Launch Numbers. That Tells You Almost Nothing.
Muse outdownloaded ChatGPT's launch week. That says almost nothing about the product, and everything about who already had your phone.
Got something like this to automate?
We take on a small number of builds at a time, as ongoing engagements with a defined scope rather than one-off tasks. If you have a process that is costing your team real hours every week, tell us what it is and we will tell you straight whether it is worth automating.
By email
New posts, a few times a month
What we are building and what broke. No funnels, one-click unsubscribe.