-
Prompt injection and the perils of leaving it to AI to handle AI
I’ve written recently about how citizens’ adoption of AI is putting stress on public services that rely on friction to reduce demand. I’ve since had many fascinating conversations with public servants who recognise the issue, and are working out how best to respond. One of the most common ideas I hear is to use AI… — read more
-
AI-powered planning: for good and for ill.
I just read an excellent article by James O’Malley looking under the bonnet of the UK Government’s recent work using AI to improve the planning process. The works is making excellent use of LLMs (plural) to solve nicely focused problems. James’s article is well worth 7 minutes of your time. One worry? Were I to… — read more
-
Playing with open weight LLMs.
Checking back in on the state of open weight LLMs after a couple of months away and the best are much better than in Jan – notably gemma4 and qwen3.5. Unlike AI agents like OpenClaw, it’s pretty safe to play with them at home using tools such as Ollama and AnythingLLM. Warning: using them may… — read more
-
Introducing OverBanded.com, a total nonsense made with AI
I got the Eurostar back from Paris yesterday. So I made a new thing. This time I’ve combined data from the excellent new Energy Performance Certificate API with council tax bands and historic individual property prices to try to work out which nearby properties might be in the wrong council tax band. It’s total nonsense.… — read more
-
Beaten by the bins
I’m beaten. Me and my LLMs have been beaten by the bins. I’ve spent a bunch of time and tokens failing to create a web app that lets anyone in the UK share their address and find out when to put their bins out. There is simply too much variation in the coding and UX… — read more
-
‘Demand Management’ is doomed. Here’s how to get over it.
Anyone responsible for a public service needs to start modelling how to adapt their policies and processes for the scenario when AI Agents reduce the cost to citizens of appeals, challenges, recalculations etc to near-zero. ‘Managing demand’ by making processes confusing and time-consuming will no longer work. AI Agents are relentlessly dogged. I’m afraid you’ll… — read more
-
Letting the Agents run wild – seamless Public Services?
A few days ago I shared a video of an AI Agent hosting a benefit entitlement interview. The AI Agent intuits a lot in the course of this conversation, but still has to ask loads of tricky financial questions about earnings, rent, existing benefits etc. Mistakes are inevitable, often causing huge headaches. But what if… — read more
-
OpenClaw MCP-enabled benefits entitlement interview
Here’s a video of a benefit entitlement conversation hosted by an AI Agent (OpenClaw Opus4.5) using my new MissingBenefit.com MCP server to do the calculations. The circumstances are fake. The jokes are real. To explain a bit more about what’s going on here. The OpenClaw AI Agent is using Claude Opus 4.5 to act as… — read more
-
In the future, our AI Agent overlords will all speak MCP
I got the train back from St Ives to London yesterday, so I bashed out an MCP server for MissingBenefit.com, the UK benefit calculator I made the other week. But what is an MCP server? I mean, blimey, what is *MCP*? Well, Model Context Protocol (MCP) is an open-source standard introduced by Anthropic in November… — read more
-
AI agents will join up government before government does
AI agents will join up government before government joins up government. This will be a mixed blessing for citizens, with big implications for trust, equity and accountability. Let’s say you move house in the UK, and want to inform relevant public services. Well, good luck. You’re on your own. There’s no single, joined-up place to… — read more
-
Subscribe
Subscribed
Have a WordPress.com account? Log in now.
