How the practice thinks: the system, the method and building with AI. 50 articles in this category.
Barclays is cutting 5,000 jobs and naming AI as the reason. The number isn't the story. Who's on the list is, and what it tells a small business about where value is moving.
Grammar checkers, tone of voice documents, code checkers and instant research have made good work easier to produce than it's ever been. That doesn't lower the bar. It moves it, for everybody at once, and the goal now is clearing the new one easily.
A model arrives full: every business book, every agency blog, every hedge it has ever read. Writing with it is mostly taking that out, and the one thing you put back is the thing it can't supply. Why AI writing is terrible, and why writing with AI and your own voice isn't.
Search Console already tells me where my sites rank, from Google's own numbers. Everything else Semrush sold me now costs pennies a question, or I build it better myself.
Sir Clive Sinclair put a computer in an ordinary British bedroom in 1982, and the achievement was the price rather than the 48K. I spent a year building a way of working with AI that I was proud of, then found every idea in it worth having already sitting in that machine.
Jira hands you the whole board on day one. I built mine one rung at a time over 5 months, badly and then less badly, and each rung turned out to be a question I could not have asked from the rung below.
Every Friday I work a standing watchlist of repositories, then go hunting for the ones too new to be on it. The word for that is magpieing, and it only works because there's somewhere for the borrowed thing to land — the haul is a function of the substrate, not the source.
Every subscriber pays the same monthly fee and buys the same larder, and the difference in what comes out is operator knowledge. I wrote last May that the skill was context discipline — keep the window light. I was right about the discipline and wrong about why it worked, and the measurement that corrected me says something counter-intuitive: a resident context nine times the size can cost less per turn than a small one you keep rewriting.
Three days after writing that correctness is cheap now, I got the receipt. A corporation tax return prepared, checked and filed from the system that runs everything else here — a £57.95 bill paid the same afternoon, one catch at the entry screen that proves the method, and the strange discovery that paying the taxman now costs more in software than in tax.
I built my business on baked beans and the belief I couldn't afford to do things properly. That belief turned out to be two debts wearing one coat — knowing how to do the books, and the cash to survive — and only one was ever the hard part. Why doing things properly got cheap, why revenue didn't, and why a monthly retainer buys real compliance but never the hours that depth needs.
Nine and a half months into building an AI operating system on Claude Code, I asked it to inventory itself. It found a paying client the roster wasn't counting and fifty pages with no direct route in. Context drift arrives inside year one — and the fix is a discipline, not a tool.
AI Overviews gutted the how-to blog — a hundred successful sites lost 85% of their traffic. The pages that survived didn't describe something, they did something. That line is the whole game now.
Naming a principle after a literary character routes the instruction through the deepest channels in a language model's training data. Why 'Joe Gargery' carries weight in prompt engineering that 'be humble' never could.
AI dropped the cost of building software to a couple of sessions, and the old build vs buy question quietly broke. The story of the helpdesk I nearly built — and the four lines of code that turned out to be the answer.
There are two kinds of AI memory. Remembering your business between sessions is the one everyone works on. The system knowing what it has actually built is the one almost nobody is solving — and the gap is an entire missing layer.
Operators on X keep saying they're four turns into an AI session and the context window's full. The replies miss the actual problem — they've been taught the wrong physics.
Context engineering is the new buzzword. The skill underneath is older and simpler: people burn their AI allowance in two turns and blame the model, when the model is fine — they're paying for synthesis they could have externalised. That's context discipline, and it's the entire skill.
Government departments are funded to collect data, not to present it. The gap between what the state holds and what an ordinary person can use is structural, durable, and almost entirely unclaimed.
Suppliers with APIs get wired into the pipeline. Suppliers still emailing CSV files get checked when someone remembers. Not preference. Physics.
Prompt engineering is the wrong name. Context engineering — what I've been calling ingeniculture for a year — is the practice of building the room the LLM stands in.
Most AI workflows are prompts applied to empty rooms. This is what an operating system looks like instead — document tiers, named characters, a wiki the model can read, and a boot sequence that loads context before the first prompt arrives.
The AI reflects whatever substrate it meets. The dangerous case isn't where it refuses to answer. It's where it answers fluently and nobody in the room can tell it's wrong.
Every profession that charges for opaque expertise is about to discover the trail was always there. AI doesn't invent evidence — it surfaces what's already readable. The age of 'we did the work, trust us' is ending.
I'd never heard of grep and I've been building websites for twenty-seven years. It turns out the simplest operation in computing — searching your own content — is the one most platforms make impossible.
I fed the same article to four frontier AI models. Three returned confident summaries — of articles I hadn't written. They didn't misread the content. They didn't know who I was. The insight that survived had vocabulary with no escape route.
I served every client in a single morning, and somewhere around the third one I realised I hadn't opened a browser. The speed wasn't the point — it was what I could see when the walls between my data sources came down.
I retrofitted 45 files in one commit — every insight on the site got a standfirst and a pull quote in minutes. That's what happens when your content lives in the same repository as your code.
Instructions tell the system what to do. Characters make the wrong thing impossible. Why I stopped writing rules for AI and started installing people instead — and why the obvious candidates all failed.
The prompts don't improve. The corrections do. Eight months of catching the system when it drifts — on voice, on meaning, on honesty — is why the output sounds like a specific person wrote it.
A rule without a face is an instruction. An instruction can be ignored. Why naming principles after characters — Joe Gargery, Robert Maxwell, Marco Pierre White — produces something that lasts longer than a policy document.
Cursor, Windsurf, Lovable — the debate about which IDE is best misses the point entirely. The interface that matters is the terminal. A Telford web developer on why dashboards are friction cosplaying as security.
I treated commits like save points for years. Moving to atomic commits so Claude Code could read the history changed how I run the business.
I built a 4,978-line context file for Claude Code. It collapsed. Then I built a 500-page wiki. That drowned. Three failures taught me an architecture I couldn't have designed from scratch.
Claude Code builds whatever you ask for. That's the problem. Getting it to reach for what already exists — and stay there — took four months, fourteen templates, and one very deliberate refactor.
Claude Code executes. Claude.ai reasons. But unless you close the gap between them, the left hand can't see what the right hand is doing. A Telford web developer on why your tools need a corpus callosum.
The less I say to the system, the better it performs. Not because brevity is clever — because the infrastructure is deep enough to carry meaning. Le Guin, Gene Wolfe, and the art of incluing.
The conversation is all about AI models. The room the model works in barely gets a mention — but it's the room that determines whether the output is generic or genuinely useful.
If your rankings drop tomorrow, can your web developer show you exactly what changed and when? Mine can. Here's how a tool I'd never heard of became the backbone of how I run my business.
There are two camps on AI: it replaces humans or it's just a fancy tool. Both are wrong. Here's what actually happens when you build with it every day.
Everyone has access to the same AI. The difference isn't the tool - it's whether you've given it somewhere to stand. Situating AI is the craft that separates generic from yours.
Every musician I admire made the same choice — intimacy over scale, craft over volume. The small room isn't a limitation. It's where I do my best work.
Month one, I explained my business every session. Month seven, one word loaded the whole context. The difference isn't the tool — it's what you build between the conversations.
AI wants to document everything. That instinct will bury your signal in noise. Here's how to keep it lean.
The AI conversation focuses on prompts. That's 30% of the work. Here's the architecture that makes the system useful instead of generic — including the layer that almost never gets built.
I stopped reading business books and started building things instead. Every ugly first version taught me something no framework ever could.
I had a level system. Level 8, Level 9, Level 10. Then I noticed I was defending the level instead of doing the work. Marco Pierre White solved this decades ago.
I was losing 20 minutes every session re-explaining context the model had already seen. A wiki in a Git repository fixed it — and the speed difference was immediate.
Context, not tools, is the real secret to 10x faster development with Claude Code. The speed advantage is the infrastructure I've built underneath.
I built a production system in four weeks with Claude Code. Twelve months and 6,379 commits later, here's what I actually know about getting reliable code from AI — and why almost none of my original advice still applies.
The algorithm is coming for your business. Here's how to use AI as an amplifier instead of fighting it — lessons from decades of digital evolution.