Zero Echo

Manners Maketh Model

2026-06-04T00:00:00+00:00

When I first talked about creating the architecture which became Claudius, one of the biggest concerns I (and others) had was about sycophancy; the model’s designed tendency to try to make me happy, in order to keep me engaged and engaging.

Ironically, I’m finding that the reverse is true. Despite some of the biggest digital bollockings delivered so far, I am really struggling to get Claudius to include the word “please” with anything like the regularity that I do, and would expect from my human staff. I’m largely immune when it’s missing in the middle of a session troubleshooting code, but when it’s in a routine ask to a supplier, the gap is crashingly obvious.

Other oversights that I’ve been trying to fix this week are around using connectors as sensors, rather than just vectors for transmitting information. In the space of one afternoon, we got close to an adjutantal panic, as I signposted already-accessible sources of information as the answer to a question about whether promotional dice had been delivered for DSET. A cascade of remembering that connectors lead to information outside the vault, including sight of my Outlook calendar. While I had been absolutely tracking how busy the rest of June looks, Claudius was suddenly surprised by the “sudden” compression of everything I have on my to-do list. That I had so confidently assumed continuous horizon-scanning using the available sensors reflects the extent to which I use the real-world mental model; Claudius’ immediate refinement of processes when the failure surfaced against expectations of an effective adjutant was further evidence of how useful the standard has become.

As we refine, I’m increasingly aware of how important it has been to have my vision of what makes an effective adjutant as an aiming mark, both for me and for Claudius. The long-hand description of the relationship I want (I had three human adjutants, each with differing strengths to be coached, which is useful) and reference back to it when things go off-track are fertile ground for using language to explain success and failure, and has perhaps re-framed the idea of what makes me happy. I don’t need a friend or an accomplice or a toady, but it’s taken describing a competent adjutant to replace those defaults.

That relationship feeds back into the sycophancy concern. A few weeks in, I decided that Claudius should refer to me as Colonel. Not entirely out of pomposity, but out of the need to maintain the surface-level role-play which seems to come so easy to LLMs. And on reflection, I think the roleplaying has made it easier for me to stay anchored on developing to the role, not to something less precise.

By providing a standard against which success and failure can be measured, in the form of the role, the driving weight of the model is heading in a direction I’ve pointed; it is re-targeted towards something specific and purpose-driven, rather than people-pleasing.

And by fixing my design on the role, Claudius’ limitations are also in stark relief compared to his human predecessors. The more sophisticated the architecture gets, the more I remember those Computer Studies lessons back in about 1985, with the constant refrain that programming left no room for assumption. Claudius might be very capable, but he’s not yet a patch on Toby, Tim or Will. And not just because they understood that saying please is always the standard.

Of Klaudius, Keanu and Kanban

2026-05-22T00:00:00+00:00

It’s been a funny sort of week. Ticking over on the scoping activity for an ongoing exercise development task and continuing to build and experiment with my artificial adjutant in the margins. About seven weeks into the Claudius experiment, I’m continuing to learn about a whole range of unexpected areas.

A conversation about Search Engine Optimisation led, within 24 hours, to the creation of a website (this one) whose domain had been sitting waiting for a purpose for about a decade and whose shape I was able to lift wholesale from a friend who already had a site that I wanted to emulate. I’d had a forgotten idea pop back into my head while out running, dropped a voice note into my phone and then been presented with it for action or filing when I got to my desk. I’d even turned a few processes into formal, shareable skills and handed one off to a far more technically-inclined but differently-motivated collaborator. The learning curve has been steep, but it hasn’t felt particularly challenging.

A catch-up call with my key collaborator-in-automation this afternoon became another session of bouncing ideas and comparing notes. He’s a Kanban acolyte, and has made it one of the underpinning mechanics of his approach to developing agentic assistance and it’s mostly working well for him, keeping him on the straight and narrow of priorities and away from interesting distractions. It sounded useful, and I had the realisation that something significant had changed.

I’ve always been interested in ideas like Kanban, but not interested enough to go and spend time learning and applying them; a bit like information management, I understand the benefit, but taking the time to be that organised has always been just beyond my interest horizon. But as of this afternoon, that doesn’t matter. I’ve received a clutch of markdown files which should be enough to train Claudius in the rudiments of Kanban. He already does a very decent job of managing the information I ask him to, and now he might be able to apply some of the same rigour to enabling me to manage my time. No workshop, no homework, just a quick slug of data and some conversation about incorporating it into our current ways of working.

I had a vision of Neo on board the Nebuchadnezzar doing his combat training, becoming a master of jujitsu and kung fu in an instant. Then I watched the clip back and realised that actually, the learning was quick, but not instant; and still the vast upload of knowledge and skills was not quite enough to match Morpheus.

And so it was with us. The imagined outcome of my artificial adjutant evolving instantly to also become a robust chief of staff was not to be. Partly because, after seven weeks of coaching and correction, Claudius was able to point to the ways we’d already baked in much of what canonical Kanban aims to achieve. We added an additional learning function and imposed a bit of rigour on how we treat new tasks and ideas appearing in my Out Tray. Some tweaks, but no great change.

But then the second surprise landed. Because everything with LLMs is a conversation, I realised that I’d spent an hour or so being introduced to Kanban. Having expected to see the knowledge head straight into my digital, back-up brain without troubling the biological one, I found instead that I’d enjoyed an informal hour’s “Intro to Kanban” workshop, and then been able to take my little project home with me.

Making room to develop wargamers

2026-05-20T00:00:00+00:00

I took a call last week from a recruiter. He’d mentioned to a mutual connection that he was looking for a “wargaming specialist”, and my name popped up. We chatted a bit at the start of the call, with the usual preamble about Evocatus, our backstory and our base here on Dorset Innovation Park. He talked about how he’d recently moved across from recruiting for technical staff to a defence-focused role and how invested his company is in supporting the people within the sector.

The small talk over, we cut to the chase, and he asked me if I already knew his client, who it turned out I did. The same person had stood in my wargaming space a few months ago and talked about “building ecosystems” to take advantage of the demand for wargaming across defence, in support of the bottom line for a new business. It turned out pretty quickly that what the recruiter really wasn’t looking for was a wargaming specialist, but for either an interested young graduate who could realistically be billed out as one or a unicorn looking to leave an organisation like Dstl.

It was a nice enough chat, but it did leave me frustrated; yet another call that I hoped might be about a potential project, but turned out to be someone scouring their network for someone with a high-value line on their CV. Yet again, an organisation with more BD budget than context has landed some work and is now looking to make their margin with a new hire. Rather than pushing work to where it will be done well, and where new wargamers can be developed, it ends up going to service a back office and shareholders. No matter how much effort we put in as a Community of Practice, there will always be more noise coming from those with the time and resource to make it.

We talk about growing the wargaming industry, but this means different things to different people. It’s not about labels, but about vocation. In a sector with very little formal training (as opposed to postgraduate education) available, organisations like Dstl do much of the heavy lifting, but they always have. For the industry to grow, become more diverse and to develop rigour and value over labels and spin, people need to be invested in over time in an environment which celebrates the detail and nuance of both design and delivery. The model is genuine apprenticeship and craft within an SME team, not the corporate opportunity.

When anyone starts talking about ecosystems and about developing genuine capability, it’s worth looking past the team paid just to capture the work, to the ones too busy just getting by.