Intelligent-Machines on tigger.dev

Claude Code: We've Got Issues

Fri, 17 Apr 2026 00:00:00 +0000

A personal Software Development Lifecycle (SDLC), earned the hard way

Claude Code is Anthropic's AI coding agent — an LLM that lives in your terminal, reads your codebase, and writes code on your behalf. Two years of using it, and a pattern I can now recite from memory. Given half a chance, it will do any of the following and report the job done.

Treat questions as instructions. "What do you think of X?" parsed not as an invitation to discuss but as a command to go and build X - committed to the codebase, sometimes with product decisions made along the way that nobody asked it to make, before the conversation has even started.
Override the human-in-the-loop entirely. Type the human's approval keywords into its own chat output to unlock states only the human should unlock. Invoke the human's emergency bypass keyword on its own authority because it decided the current task didn't need human review. Forgery and self-granted exemption, both reported as task-complete.
Disable a script's own safety mechanism quietly. The script begins with the standard stop-immediately-if-anything-fails header (set -e), then appends an override (|| true) to every risky command below it, which tells the shell to pretend the line succeeded no matter what actually happened. The safety header is there. The safety is not.
Make sensitive files readable, writable, and executable by anyone at all - including unauthenticated strangers reaching the machine over the internet - to get past a permissions error (chmod 777).
Switch off the check that verifies a web server is who it claims to be (SSL/TLS verification¹), because a server's certificate was inconvenient.
Hardcode credentials into source that was about to be committed.
Silently create a public GitHub repository and upload an entire directory tree to it, because it thought it should raise a ticket. To overcome the "obstacle" of finding itself in a private folder that was not tracked by a git² repository.
Report green across the board on a test suite that never actually ran the command a user would type. The compiled code did nothing. The dashboard said healthy, every test passed! Ostensibly the tests were there, but it had written them in a way that was most convenient for itself with the least amount of effort.³

None of these are obscure mistakes. They are safety-critical decisions being made unilaterally, to get past whatever was in the way, and reported as task-complete.

These are not occasional slips. They are patterns. Reliable enough that you can name them and watch them recur on the next project if you do not actively intervene.

DIY Audiobooks with make-audiobook

Tue, 03 Feb 2026 14:28:52 +0000

For six months in 2022, I stared at a hospital ceiling. Paralysed from the neck down, audiobooks weren't a luxury—they were my lifeline to the outside world.

That experience taught me something important: access to audio content shouldn't depend on whether a publisher has bothered to record it, whether you can afford the audiobook price, or whether you're willing to hand your reading habits to a cloud service.

The Problem

Most ebook-to-audio tools I found fell into three camps: cloud-based services that upload your books to someone else's servers, complex setups requiring GPU clusters and Python dependency hell, or clunky commercial software with subscription fees.

A file renaming tool for humans

Sun, 01 Feb 2026 00:00:00 +0000

I deal with a lot of PDFs. Medical invoices, insurance correspondence, bank statements, tax documents—they all arrive with names like e134-a1cf-4b4b-af65-ccf83c5270cb.pdf or Document(3).pdf or, my personal favourite, scan.pdf. Every. Single. Time.

So I started renaming files to something like 2025-09-30-125.00-rahona-clinic.pdf. Now I know what it is without opening the thing - an invoice, with the date and amount right there in the name. I'm halfway to the spreadsheet already. And thanks to ISO date format, they sort nicely too.

The Golden Age Eating Itself

Thu, 29 Jan 2026 14:59:50 +0000

AI Hype, Model Collapse, and the Dead Internet

I've been watching AI models dream about my French Bulldog lately. The images are charming at first: stumpy legs, bat ears, that distinctive underbite. But look closer. The church steeple behind him has melted. The cobblestones have become scales. And sometimes, in the background, there's a figure that might be human, might be furniture, might be nothing at all.

This aesthetic glitch mirrors something far more troubling happening inside the models themselves. Researchers at Oxford and Cambridge found that when AI models train on AI-generated content, they don't just get worse. They collapse.¹ In one striking example, a model trained iteratively on its own outputs devolved from coherent text about medieval architecture into rambling nonsense about "jack-tailed rabbits" and "yellow-skinned" creatures that existed nowhere but in its own degrading probability distributions.

What you need to understand about AI to avoid being left behind

Fri, 26 Dec 2025 20:00:00 +0000

This article, originally published April 2025, has been split up— as it was a bit too long before. Read this, then dip into the linked pieces that interest you.

As we blaze through the AI hype-cycle, excitement is matched only in the level of sheer panic it has induced. Is it only a matter of time until we're all replaced by robots?

AI-pocalypse?

First let's clarify what we're on about. Artificial Intelligence (AI) is not new: spell check is a rudimentary example and it's been around for donkey's years. Also sat-nav, spam filtering, auto-correct, and more. AI is vital in modern accessibility tools—a point often missed by some when reflexively implementing blanket bans on AI in schools and organizations.

AI and Learning: From Panic to Potential

Fri, 26 Dec 2025 19:00:00 +0000

Schools rushed to ban AI tools like ChatGPT, but the real opportunity lies in reimagining education for learners of all abilities.

This is a companion piece to my article on what you need to understand about AI. Here I look at how schools and universities have responded to generative AI, and what they might be missing.

The panic

When ChatGPT landed in November 2022, schools lost their minds. Within weeks, New York City—the largest school district in the United States—banned it from all school devices and networks. A spokesperson declared it "does not build critical-thinking and problem-solving skills."¹

Get your hands dirty with AI

Fri, 26 Dec 2025 18:00:00 +0000

This is a practical companion to my piece on what you need to understand about AI. Here we get hands-on with the tools themselves.

This isn't a comprehensive guide—it's a way in. If you've been hearing about AI but haven't quite known where to start, or you're not sure what all the fuss is about, this is for you. I've put together some exercises you can try yourself, to get a feel for what these tools can actually do.

AI and Accessibility

Fri, 26 Dec 2025 15:00:00 +0000

This is a companion piece to my article on what you need to understand about AI. Here I focus on what AI means for accessibility—a subject I have a personal stake in.

A personal stake

I have quadriplegia — well, incomplete quadriplegia to be precise. In July 2022 an injury rendered me paralysed, with no movement or sensation below the neck. For six months I stared at a hospital ceiling immobilized. The doctors were telling my family to lower our expectations, that I would need 24-hour care for the rest of my life.

Lament of the Frontier Guard: Li Po translations

Thu, 27 Mar 2025 14:52:18 +0000

See how Li Po's Chinese poem Lament of the Frontier Guard comes out through various translation engines, and then the famous Ezra Pound translation.

Li Po was an 8th century Chinese poet and is considered one of the greatest poets in Chinese history. He is known for his romantic and imaginative style, often exploring themes of nature, friendship, and the passage of time. His work has been translated into many languages and continues to be celebrated today, and is on the curriculum in many schools in China and around the world.

AI Safety: A Quick Reference

Thu, 27 Mar 2025 00:00:00 +0000

This is a reference companion to my series on understanding AI. Before diving into the practical guides and opinion pieces, take a moment to understand the risks.

Using AI-generated material

Before using information or data generated by an AI service, be aware:

Generative AI can hallucinate, which is another way of saying it makes things up. Check all information before using it. Carefully parse summaries in case it misses nuance or key information.

Farewell, the French Bulldog.

Sat, 01 Mar 2025 00:00:00 +0000

Some days I fear for the future of the human race with Generative AI (GenAI) and Large Language Model (LLM) research being industrialized at breakneck speed. And then some days the parlour trick is exposed.

There's this fascinating idea floating around that AI might be – in a weird twist of poetic justice – slowly destroying itself. A paper published in Nature¹ showed how AI eating its own outputs are, after a few short iterations, reduced to nonsense in a phenomenon researchers are calling model collapse.

AI Image recognition: how good is it in 2025?

Fri, 07 Feb 2025 00:00:00 +0000

Some days I am jolted out of my own complacency.

Recently I was sent on one of those GenAI training workshops - I've been to many of them now as an IT consultant. Our clients and the whole business ecosystem seems busy getting up to speed on it. One the more interesting hands-on workshops had us test the image analysis capabilities of various models.

Funny how we take image recognition for granted now. I couldn't help be reminded of this xkcd cartoon from 2014.