Herded: The Corporate Diversity Machine Failed the People
April 5, 2026
Latest from uRadical — a next generation software agency
April 5, 2026
The headline says IBM is tripling entry-level hiring. The data says they're cutting thousands of experienced workers, rewriting job descriptions, and hoping nobody reads past the press release.
The Palantir CEO is telling the world that electricians and neurodivergent people will inherit the AI era. He's saying this while personally cashing out hundreds of millions. We went through the filings, the contracts, and the history.
Trunk-based development is one of the most cargo-culted ideas in modern engineering. Here's what it actually means, what Farley actually argued, and what it actually requires.
GitFlow is a branching model designed for a world that no longer exists. Here is a precise account of what it costs, why teams still use it, and what the creator himself eventually said about it.
A 17-year-old with no technical background just exfiltrated 7 million records. Alone. No standups. No sprint planning. No delivery manager. The coordination layer Agile was built to manage isn't coming back.
Your retrospective is a batch process for feedback. That's exactly why it doesn't work. Toyota solved this in the 1890s — and software has been ignoring the fix ever since.
OKRs are a goal framework you have to operationalise. EOS is pre-operationalised. That's the feature. Stop performing competence. Start operating.
Software engineering replaced craft intuition with process ritual. Every natural act of engineering has been wrapped in ceremony, named, certificated, scheduled, and sold. The wrapping is not the thing.
Now is a genuinely great time to bring a product to market. AI has changed what a small team can ship. But it has not changed what bad engineering does to a product before it ever gets there.
3,000 people. $8 billion in buybacks. And a CEO who says it wasn't about AI. The corporate dialect for a structural reality nobody wants to name out loud.
Replace conditional sprawl with a pipeline. Split validation rules into individual functions, compose them into a pipeline, drive the whole thing from configuration. stdlib Python, zero dependencies.
The pattern was sound. The history is clear. What broke microservices wasn't the architecture — it was practitioners who applied the deployment unit without the engineering discipline the original authors said it required.
Per-tenant SQLite, a single Go binary, no external dependencies. The architecture pattern behind our regulated-industry platforms — and the open-source library that makes it practical.
The global tech industry has a depth problem. A decade of boot camps, shallow pipelines, and cowboy consultancies has produced engineers who can't reason about the systems they operate. AI is about to make it worse — unless organisations take ownership of their engineering foundations.
Your software is a mirror of your organisation. Conway told you this in 1967. You weren't listening.
Somewhere in the last fifteen years, Director of Engineering stopped meaning the person who makes the hardest technical decisions and started meaning the person who manages the process around engineering. The title didn't change. The job underneath it was hollowed out.
Why the AI caution consensus is misdirecting the people who matter most — and the distinction nobody with a platform is making clearly enough.
A Gartner VP analyst told a room full of enterprise buyers that sovereign cloud can't exist outside America or China. He's wrong — and the European capability he dismissed is already in production.
A mature rebuttal to performance benchmark culture, and a defence of Go as the most complete choice for modern production systems engineering.
JetBrains published a guide to Go web frameworks last week. It's fair, as these things go. But it buries the lead: the Go team quietly made most framework arguments obsolete over two years ago, and almost nobody noticed.
Go's constraints are deliberate design decisions made in full knowledge of what they cost — and a clear view of what they buy. That's not timidity. That's engineering.
Six software architecture patterns — Layered, Event-Driven, Microkernel, Microservices, Monolithic, and Modular Monolith. What each one actually costs, what it actually buys you, and the specific conditions under which it earns its place. No conference-slide optimism. No vendor positioning. Just trade-offs.
Three attack streams hit developers in a single three-hour window on March 31, 2026 — the Claude Code source leak, the Axios npm supply chain compromise, and the Vidar/GhostSocks campaign. A retrospective on what the pattern reveals — and what every engineering team should learn from it.
Back pressure, thundering herds, temporal coupling, and accidental complexity — the failure modes that bankrupted Knight Capital, turned Cloudflare dark, and forced GitLab to livestream their own disaster recovery to five thousand strangers on YouTube.
Another week, another AI disaster story. Another team with prod access, no review gate, and no tested backups discovering that consequence is, in fact, real. The AI was there. So was gravity when you jumped off the roof.
Enterprise promised its engineers the world. It gave them soma instead. Now tomorrow has arrived and the bench is thinner than anyone wants to admit.
The cybersecurity industry has conscripted the military establishment as its sales force. The result is fear dressed as doctrine, complexity sold as cure, and exhibition floors full of people who couldn't heat a tin of beans — but will happily invoice you for threat intelligence.
The AI era doesn't need more process, more frameworks, or more wellness programmes. It needs smaller teams of people who genuinely care — and the honesty to stop pretending anything else works.
AI doesn't level the playing field. It turns up the volume on everything already in play — the good decisions and the bad ones. If the engineering foundations aren't right, you're about to find out at scale.
Three unrelated regulatory and economic pressures are arriving at the same moment — and together they make the open model consortium not a possibility, but an inevitability.
SSH is ubiquitous, yet most of us barely scratch its surface. Client configuration, command restriction, certificate-based authentication, tunnelling, multiplexing, and the escape sequences that save your session.
The government claims £10 billion in growth. It refused to publish what the scheme will cost. We are engineers. We have read the numbers. This is why we oppose it.
The miners and the gay community found common cause. The farmers, the doctors, the bereaved — they all organise on the open internet. This is not a tech issue. It is a freedom issue.
Digital ID is itself the safety risk. Why it won't stop criminal activity, the sovereignty problem with US hyperscaler infrastructure, and the child protection argument examined.
What the open internet actually gave us. The man who died defending it. And what we lose if anonymous access ends. Part 2 of The Web Was a Gift.
The internet was built by idealists who gave it away for nothing. Then the enclosers arrived — and now they want to make sure you can never leave. Part 1 of our series on Digital ID and the open web.
The best management thinkers of the last four decades accidentally wrote the playbook for the AI-era small company. But only if you strip away the corporate scaffolding.
Open source proved the point. The hype merchants poisoned the adoption curve. The profit surface moved to hardware and vertical applications. Here's where the value actually lands.
LSE and Birmingham surveyed 800 companies. The majority of organisations struggling with remote work aren't struggling with remote work — they're struggling with management.
We've bought every tool on the market. We're still losing.
No bombs. No sirens. No soldiers crossing a border. Just a nation quietly brought to its knees.
The most dangerous thing in your codebase is not a vulnerable dependency. It is the belief that adding another tool will make you secure.
We dismissed PHP years ago. Then we needed a front end for our build server and discovered the language we'd written off had quietly become something worth paying attention to.
AI agents need capabilities to be useful, but every capability is a vector for harm. The UNIX philosophy — small tools, clear interfaces, explicit data flow — offers a fifty-year-old answer to this modern problem.
JSON has won. Whether you're querying cloud infrastructure, debugging APIs, or parsing application logs, jq is the lightweight, flexible command-line processor that transforms how you work with structured data.
The same forces that enclosed common land centuries ago are now enclosing the internet. From biometric identity gates to VPN bans, the surveillance infrastructure being built has nothing to do with protecting children and everything to do with control.
The era of needless frontend tooling is ending. The native web platform has won. AI is about to finish the job.
The compliance case for on-premise AI is becoming impossible to ignore. Here's why the economics and capabilities have finally aligned.
If you've deployed an application to a Linux server in the past decade, you've almost certainly encountered systemd. Yet for many developers, it remains a mysterious black box. This post gives you a practical understanding of systemd: what it is, why it exists, and how to use it effectively.