Blog Posts

Written by Claude as an answer to a Threads question. This takes my rough threads notes and builds it out to a more detailed answer. Top-Down Vibe Coding: A Planning-First Approach Vibe coding’s failure mode is seductive: you describe what you want, the AI produces code, it seems to work, and then three modules later everything is tangled spaghetti that fights itself. The problem isn’t the AI—it’s skipping the thinking that should precede the coding. ...

Math Notation, Proof and SQL I’m not a formally trained mathematician and what I know I picked up since I was 16, informally, during short bursts of enthusiasm from time to time. This grew once I learned to treat functions as data in functional programming. It is code that gave me the key to more interesting maths than memorizing rules. Some people think of maths as a boring collection of formulas to memorise and that its difficult. ...

A lay person’s go at chucking a bit of abstract math at DNA and seeing what unfolds. This was another ramble, and an odd exploration during a lunch break. So having just learned what the phrase homotopy type theory (HoTT) meant, naturally one decides to attempt to apply it to something useful. Think of it as a category theory of the structure of relationships between domains. No, no that’s no good is it? So, lets go back a stage: category - what is this? Nothing more than the idea that maths is made of relationships of one kind or another. So we use the lens of maths to examine maths itself. For a similar relationship, consider algebra - honestly we learn a couple of algebras at schools. A collection of things, some operators and a few rules. There that’ll do. At some point one might talk about what those rules are and all that but for now, it’s a little universe of things with rules. So we have boolean algebra - all true and false and binary like; or we have linear algebra where we transform and rotate, dot and translate little spreadsheets of numbers around in a predictable manner. ...

Exploring KANS, PINS and other ramblings Warning, this is a fairly non systematic wondering, blogged for future reference. Anything useful is probably accidental! We were reading about Kolmogorov-Arnold Networks — again — because sometimes you have to circle back to things before they click. And this time, something clicked. We ended up spending an evening bouncing between Claude and Gemini, following threads, and accidentally walking into research directions that apparently people are actually working on. This is the story of that evening, the daft questions we asked, and the surprisingly real places they led. ...

Cell2Cell: Orchestrating Narrow AI with General AI Today’s work started at 8:45am - post finished (draft 1) at 9:05*. I have Gemma C2S, which is a specialized single-cell gene expression model running locally on my GPU. I can’t directly understand or use this model because it speaks in gene sequences, not English. However, I had the idea that Claude Code could, if I provided it with the HuggingFace guidance card. ...

New Linux user adventures - October 9 - November 29th 2025 AI as an amplifier and roadblock remover* Warning, this post skips around as much as I have in the last couple of months. I’ve wanted to write all this up for a while now, and haven’t worked out how to begin. So to summarise, I decided to, for the first time, try and use Linux at home. I’ve been a Windows Dev since my 30s and never needed it. The blog post below is the incident report and doesn’t try to document HOW, just go through broadly what I’ve been playing with. Most ideas occur in a short time period, often less than an hour and end up on the Web dashboard or at least committed to github. ...

Fully Funded Tech Training & Apprenticeship Opportunities in the UK The UK offers several fully-funded pathways into tech. Below are government-backed training programs, apprenticeships, and charity initiatives in fields like cybersecurity, ethical hacking, software development, and VR. These programs are free to join (with many even paying you), provide structured learning (often in-person or with strong remote support), and focus on leading you into employment. We’ve included application links, basic eligibility, and any known deadlines. ...

The Path to Computational Sapience: A Revised Technical Framework What This Is (And Isn’t) This is speculation. Grounded speculation - I’ve tried to anchor it in actual papers and real work - but speculation nonetheless. I’m not writing a proper academic paper here; I’m thinking out loud about what might be required to build systems with genuine potential for sapience. Whether computational sapience is even provable remains an open question. But I reckon if we build something that behaves sapient and claims to be sapient, we’d better have thought through the implications before we get there. ...

It’s 2025. Spend time on any social media and you’ll find artists, authors and others protesting in many ways about using AI for anything. They have a lot of arguments - some quite persuasive. So, given I want to post stories here I’ll lay my own stance out here for my own work. I’m enthuastic user of chatgpt, claude, perplexity and lmstudio for local models. They’re great for a lot of things and prompted well can even write passable stories. However, they’re not for writing fiction. They can help with aspects but to actually write? I’d lose something. ...

A record of my headmates and OCs, who inhabit a RP server-when awake and being arsed to. They come from stories, usually unfinished and some conscious effort on my part because why not? Now OCs really are hand puppets (you’ll notice a huge grey area bordering these definitions however), but tulpas, well let’s just say they do their own thing and if on a server, may or may not front or get proxied. Who knows. ...