Funnily enough, Lemmy does allow AI-assisted code contributions, though they’re not encouraged. This is mentioned in the code of conduct:

Use of so-called Artificial Intelligence (AI) is allowed only if it is explicitly mentioned. Additionally all LLM-generated text or code must be manually reviewed by the author before submission (no vibe coding allowed).

This article is not advocating AI driven development. It’s arguing a human understanding the problem, and how best to solve it, is the most important work.

I’m not anti-spec. Far from it. I think rigorous, well-structured specs are about to be the single most important artifact in software. Read the book. Adopt the format. Use the tools. The mechanics matter and I’m not waving them away.

But the spec is a vessel. The recipe is not the cooking. If we pour mediocre understanding into a perfectly-formatted spec, we get perfectly-formatted slop, faster and at scale. The format is necessary. It is nowhere near sufficient.

The real work - the part no tool will do for you - is the human work of digging in. Of actually understanding the problem and the person you’re solving it for, deeply enough to have an opinion about what right feels like. Of capturing not just the requirements but the taste. The why. The elegance. The exact sweet spot of how it should work.

I agree that the AI generated image is trashy, however the article is a cautionary tale about the pitfalls of relying on agentic coding, instead of collaborating with other developers.

But there is always a ceiling on how far a single-player game can take you, even with agents. Software that lasts, software that grows, software that people can actually depend on – that is built by groups of people exercising judgment together over time. By teams developing shared taste, shared mental models, shared sense of what their product should be. None of that happens through individual prompting, no matter how clever the prompts.

Agile came from toyota?

My understanding is that Kanban came from Toyota, which is an agile way of working.

This is a fascinating article about the history of software development. For me the key quotes are:

The thing that killed Waterfall was that discovering your spec was wrong months later, after lots of code had been written - and fixing it cost a fortune because writing code was the most expensive part of the process.

The key reason Agile was invented was to account for the high cost of writing code, so yes, that part of the Agile value proposition is no more.

The risk isn’t that AI development is inherently Waterfall. The risk is that organizations with latent Waterfall instincts will use spec-generation as license to do the bad thing they always wanted to do — front-load requirements, skip customer validation, equate a fancier document with a better outcome, and ship one massive thing every quarter.

This quote from the article really sums it up:

And to be clear, I don’t care whether you typed the code yourself. I care whether you understood it before you shipped it. I care whether you can explain why the bug happened, why this fix is the right fix, what the model might have missed, and what would make you roll it back.

Could you elaborate on this?

Thank you! I’ve updated the post with the TL;DR from the article.

An acronym for domain-driven design.

Depending on your level of programming experience, you might find the exercises at Exercism quite useful.

In case anyone is curious, this is the original post on X.

Could you give more context about what Vercel features you need - is the site statically generated, or do you also need Vercel Functions?

There are several European based alternatives to Vercel. It’s also worth having a read through or posting to !web_hosting@programming.dev

Your comment reminded me of this article, The Software Quality and Productivity Crisis Executives Won’t Address, which discusses the lack of technical leadership when it comes to tackling technical debt, and that the solution is usually a rewrite.

Instead, most organisations don’t tackle technical debt until it causes an operational meltdown. At that point, they end up allocating 30–40% of their budget to massive emergency transformation programmes—double the recommended preventive investment (Oliver Wyman, 2024).

I have noticed the repository lacks CONTRIBUTING.md. If you want to set some rules about contributing, I would have added them there, instead of creating a Markdown file specific for agents. I’m very much of the philosophy that you should write documentation for humans, which has the added bonus that it will also be consumed by agents.

I agree but it depends on how teams create and refine their tickets. For example, you could have high level tickets, and someone picks one up and creates an implementation that’s not an appropriate fit for your architecture.

Thank you for not assuming my motivations. Could you please elaborate on what you mean by “oneshotted”? I share a lot of articles, so I’m not surprised you recognise my username.

I don’t specifically seek them out. I follow quite a few different programming blogs, and I am just sharing what people are posting about, and it just so happens a lot of people are posting about this topic.

Headless does not mean “no screen anywhere.” It means you are not required to use the company’s app or site to finish the job.

You might say: “Book a flight and a hotel in Tokyo.” A helper (with hooks into services, e.g. MCP or other agent APIs) talks to airlines and hotels for you. You might never see their homepage or their “join our club” popup.

Whilst I can see where the author is going with this, I can’t see some tasks, particularly booking concert tickets, being done by AI agents. Whilst it may be convenient for end users, it’s also open to exploitation by scalpers.

Not sure if you were even looking for paper reviews.

I didn’t write the article, I just shared it because I thought it was interesting.