Bad Code Smells

I smell a bad smell a-risin’
I won’t deploy code today
I think the logic might be frightening
I see a rewrite on the way.¹

Note: this blog post is intended as a discussion piece, an editorial if you like and not an investigative or pure informational/educational article. The views expressed in are the authors and not wholly reflective of the company, other employees or even the author themselves after they’ve had enough time to think about it and for a different opinion.

M’ dearest colleague, Mike Whitaker, was chatting online in a social media channel² thataway –> and he made this comment:

And it struck a chord with me as he’s absolutely right, it is about knowing how to recognise a smell. Another way of thinking about it, is how to sense when something isn’t quite right, a sense that elicits an emotional response tied to it. Now this leads me into an argument I like to indulge myself by repeating when discussing AI is of keeping people involved in all stages of a process.
Now by process I mean pretty much any thing that is produced, this might be about software, but it could be about holiday planning, researching or comics. Because it isn’t just about the competency, or the ability to do something with a degree of accuracy and efficiency; nor is it just speed, reproducibility or cost. it is about the experience and the emotional response, and by that I mean that having a lot of one will trigger the other.

This to me is fundamental, keeping people, holding onto the learned experience, joining it to the experience of others and growing experience. I think it is going to expose some huge failures in how we see these new technologies implemented.

Deploy for Cheap, Maintain at a Cost

Like a herd of gullible lemmings being steered by the Disney Land Rovers^[a] of big tech towards the inevitable high cliff there is a rush, a movement, an evolution, a progression (that throws in a fourth item and screws with the notion that ‘generative AI wrote this’) to let automated tools handle most of the process of constructing a project.

One of the natural homes of this suicide plunge for the progress of an emerging technology is software development, but like an untreated fungal infection it is spreading to other areas. This leads immediately to two major issues that are already growing in their reach:³

1. We are letting automation replace people. Often junior or introductory roles, but increasingly to mid-level employees in a growing number of disciplines. ⁴
2. We, as a company, business and species, will shop around for the lowest cost, the best deal, the two for one or the buy one get one free (which it isn’t as it has the conditional requirement that you have to buy one).

One: Great Replacement

Taking a casual glance at employment figures for graduates, school or college leavers and those with limited years of experience (as well as those too old or costly to keep) will already show that there is a dearth for employment prospects. There have been a number of large layoffs of technical and creative staff in industries that has wiped out many junior and mid-level posts. The software, journalism, animation and film industries, the early targets, resemble the battlefields of an early twentieth century conflict, with both sides entrenched but only one side has access to the machinery of destruction. The notion that a single set of agents controlled by one experienced operator prevails in both the propaganda and the balance books of many organisations and they seem to be happily deploying a scattergun strategy as they trim their numbers.

This is seemingly the wave of ‘great replacement’, the phrase so often misused, clearly morally repugnant, and therefore apt when livelihoods are rationalised by profits. There is an unproven mantra that this AI automation will reduce all overheads while simultaneously speeding-up work or production. In the right hands there are gains to be made, and productivity increased, for some, but it takes experience to know when and how to do that.

What it is guaranteed to do is leave a large number of people dispossessed, angry and unemployed. It should be noted that the waves of redundancies in the last year in the tech industry might have less to with AI than widely claimed and more a bid to look good to shareholders and investors or to justify the need to cut costs with a useful scapegoat.^[b]

The only reprieve that seems to be presented from large organisations and authorities to those lost in this emergent wasteland focusses on attempting to open new markets where we can all be the lone pioneers; or to insist that the abandoned souls retrain in the hope of showing value in an imagined, poorly defined, future as an expert. Before that expertise also falls to the inevitable silicon superiors.⁵ Both of these approaches, while having some validity, will only address a fraction of the problem and a large swathe of people will remain out in the wilds beyond the wall.

Two: Built by the Lowest Bidder

For newspapers, publishers, software houses, consultants, designers, technical engineers, pick a discipline any will do… the wave⁶ that we are all part of heralds something depressing. Our skills are to be valued against an automated system, and worse they are going to be used in a bidding war with anyone who can switch it on or off or write an elegant prompt.

How do we justify charging as much as we do for our experience, ability, commitment to procedures, security, governance, trust and other qualitative metrics when someone who has less observance of those will build a similar product for a fraction of our price?

This is perhaps simplified as the bespoke, the artisan, the craftsman versus the quick, fast and easy automation; the individual versus the army of clones. But these are not production line items, they are not exact replica, they are not a facsimile but a doppelgänger. It might look like the original but it isn’t. But for many if the end result looks like a pig, smells like a pig, cooks like a pig, and even dances like a pig, the maker will be long gone before you realise it tastes like fakon.⁷

For me the less human experience involved in deploying and maintaining a complex system, the fewer people who have hands on and interactive engagement with it during its evolution and construction, the higher the likelihood of a serious, or even catastrophic problems entering the process, you will be placing all your eggs in a basket you didn’t make, don’t operate, don’t understand, don’t own and can’t control.⁸

To add to this, I worry about the motives and professional discipline of anyone who wants to build everything at speed and for the lowest possible outlay. Their primary motivation is just to gather as much money in the least time and for the least responsibility. There is little job satisfaction in that and there is little adherence to creating something reliable. When everything is a cost-cutting exercise the person with the biggest blade wins and they do it by making cuts you can’t see.

Our experience makes us take time to think things through, to check the output of automation even if it is close to perfect almost every time, even if its quality is good. Because we want to ensure we have trust, because we want responsibility, and in doing so we extend our governance and reliability.

And that smell…

While you are happily enjoying your cheap fakon, and by enjoying I mean chewing with an unimpressed grimace on your face, let’s address that “bad code smell”. I would be a liar if I said we could always tell the fakon from the bacon. But if you ask a chef who has cooked a thousand times they’d probably tell you that they can tell bacon from facon in a moment with their eyes closed and their head removed from their shoulders.

A good cook might be able to save a sauce that’s split, a chef will make it split on purpose and for a good reason. They can cook a souffle a thousand times and might only have only one that collapsed because they took too long shouting at the KP even though the ingredients and conditions vary. More importantly they can instruct others how to cook by sharing experience, create new recipes, will know if the ingredients, or end product, needs adjusting because of slight variances. Try to get an agentic system to share its method, or the route to how it did something. Chefs don’t just look at a potato and statistically determine its size, water content, etc to give a cooking time and length, they will use their experience and feeling to know how long and how hot and the end product is unique and also familiar.

It is likely that every single one of you reading this will be expert in something. A hobby, a skill, a musical instrument, whatever. And every single one of you will know that sometimes it doesn’t ‘feel right’. That is the thousands of tiny signals that you get, the ‘bad smell’, it is born from a lived experience and it isn’t something you can teach, only show, you cannot define it enough that a probability will match that feeling. It is a knowing that comes from learning, doing, and repeating, it is from the action and not just the replication.

We need new tools. It is wonderful that we have new lands and new opportunities to experience. But we need to be the explorers and students, we need to bring along our juniors so they get the same experiences and come to understand and learn how to smell.⁹

Tools: This editorial was written entirely in markdown using Obsidian. No other tools were used in the creation except a search engine and a brevity of wit.

1. With apologies to John C. Fogerty and Creedance Clearwater Revival.
2. It was LinkedIn if you were curious, in a chat about a previous blog post on this site.
3. There’s probably more than two, scrap that, there are more than two but these are the two I am focussing on in this article.
4. Management and senior employees will be next in some organisations as departments and projects are coalesced and rationalised in what I assume will be the continuous pursuit of greater profits.
5. And that future will never be properly defined while we are in a mad growth phase of the technology that is less like an Industrial Revolution and more like the expansion of an empire, or the colonisation of unknown territories. Few laws, few watchers and a whole slew of new opportunities and directions to travel.
6. I resisted the urge to use a turd tsunami and say it was part of a shumfing-storm, except of course by writing it here, I didn’t.
7. How exactly do you spell fake bacon, i.e. Facon/Fakon/Fakeon, without trying to be dismissive of vegan/vegetarian/plant based alternatives, which for once I wasn’t trying to be?
8. Note that I am not saying that we remove using the tools altogether, just how we use the tools and what steps we place to understand what’s going on. You could easily argue that very few people understand a complex system, but we do know that ‘someone’ has a grasp of it. Now we are building tools where the builders of those tools don’t know how the tool works, they have less understanding of how it built whatever you asked of it.
9. I know that employing people while introducing tools is going to increase costs, that’s inevitable. Which is why I advocate for keeping, training and investing in those people and introducing new technologies at a more refined, and knowledgeable, pace. Turnover efficiency can come from productivity and expansion and not from just simply reducing outgoings.

[a] https://en.wikipedia.org/wiki/White_Wilderness_(film)
[b] https://hbr.org/2026/01/companies-are-laying-off-workers-because-of-ais-potential-not-its-performance
https://economictimes.indiatimes.com/magazines/panache/ai-didnt-steal-your-job-then-who-did-yale-study-busts-the-myth-of-the-great-automation-apocalypse-speculation-not-disruption/articleshow/124322702.cms?from=mdr