We are living in some interesting times. Gone are the days of reading thick books and getting good at software design and computer programming. Now you can use plain english and complete that million dollar idea you had by programming in plain english thanks to Flow-matic!
Well ok, that wasn't ever well adopted and you probably need to "google that" to even understand the reference.
But now, "anyone can build enterprise apps in days." with Odoo Studio Delphi! I wonder how many enterprises are still running them...
How about expressing your ideas through cutting edge technologies that have visual analogs like Hypercard/Hypertalk or M$ Visual Programming Language. So easy you can teach it in middle school!
The promises are so alluring, it is no wonder why people keep trying... to sell them to you.
How To Do Creative Work In 2025
I'm not opposed to using AI to assist, though I do personally draw the line at taking your entire code base and shipping it off to a 3rd party for analysis through IDE plugins and the like. Why? Well for starters, licensing! Do you want to actually own the work you produce? Do you even have permission to do this with the copyright holders of your codebase?
Instead, I use AI like someone uses search engines, like google, and hybrid content aggregators, like stackoverflow -- as something that can provide examples and to try to solve a problem I'm encountering.
If you have been doing work in this space, you probably know what I'm talking about. More productivity, but in no way is AI actually producing the output of my labors any more than some guy on stackoverflow that explained what cp argument makes a symlink.
The Disturbing Trend In 2025
It was about two years ago a client casually joked about using ChatGPT to solve a problem they were having with Salesforce Ecommerce (formerly Demandware). Well we tried it, and it certainly wasn't something that could just be taken and actually run, but it hinted at the various APIs and essentially what the "form" of the solution must be, and for someone who had never seen or programmed for this language or platform at all, this at least seemed helpful.
Now, however? Well, while nobody is explicitly stating it, I'm seeing people be essentially forced into using it.
Example Quote 1: "I'm not actually an Odoo developer"
A big trend I'm seeing are individuals who work in IT are now cosplaying as developers more and more. Note that I'm not talking about a Magento developer that has started to work on Odoo, but people who see these tools as that great playground leveling force that lets them make claims about what codebases do and how they are managed because "I analyzed this with an LLM". The problem here is that they cannot actually fix issues or improve the code base, but they are good at asking questions ... questions like "what does this file do", "how do you actually run this code", and "explain this to me line by line". They interrogate you like they do their favorite chatbot and give you the same level of appreciation to your knowledge that they do to the chatbot -- none.
Being upfront about using AI, and prefacing things like "I'm not an Odoo developer" absolves them of any inaccuracy or ignorance while demonstrating how proud they are that they're saving time/money by "not actually being a developer". It could just be my experience, but the more I am on the receiving end of their grilling, I start to question what their real world experience is and when I will see it expressed.
Example Quote 2: "Business goals necessitate me developing our own modules"
If you work in IT, you can fix my printer, right?
I feel like an unfortunate issue with living in an AI bubble is that management is sold on the idea that their existing IT or development resources need to justify their jobs by taking on coding.
The core problem here is that it doesn't work, and likely never will with the AI tools and techniques that the current breed of AI companies are trying so desperately to sell you. Just google "reddit vibe coding doesn't work", and you'll find real people trying to do real things and failing at it without oversight or last-mile assistance by someone they didn't hire because someone at a tech startup, who hired the best coders they could get their hands on, told them that their platform could do it for a fraction of the cost. Like so many things in life, things that promise "a fraction of the cost" are targeting consumers who are wistful and "don't have the money for the real thing".
Don't believe me? Ask the AI's themselves! Ask ChatGPT if you wanted to build a ChatGPT rival company, would you need to hire developers?
Follow The Money
I feel like the two examples just lead back to who "was sold" the AI coding myth. Was it an individual who sees it as a way to skip all the "hard parts" of programming and rocket straight to the finished product? Or was it the MBA's who watched a Youtube video and is going to save their company time and money by utilizing their resources better?
The former, at least from what I've seen, tends to be using this as a "selling feature" of themselves. That they embrace this and promise to be able to achieve anything!
The later seems to be an evolution of organizations that heavily embrace Odoo Studio. Difficulties with their Studio customizations are shrugged off during migration, and eventually their ambition takes them to needing to write their own modules. Presumably all of the examples that they get out of their favorite LLM includes starting with boilerplate modules. I've seen it, even if you ask for a "server action", they will still churn out pseudo-working python modules, usually with halucinated models or relationships that encourage you to "make that model!" to get it to work.
Now I'm sure people are going to think, "well you're just a software developer, so you are financially motivated to gatekeeper software development". This is of course true, but also missing the nuance of what it takes to actually design, develop, and maintain complex interconnected systems.
My parting example. Ask yourself, and your AI...
Which is better? Code likerecords = env['purchase.order'].search([('state', '=', 'draft'), ('order_line.product_id', '=', 123)]) orrecords = env['purchase.order'].search([('state', '=', 'draft')]); records = records.filtered_domain([('order_line.product_id', '=', 123)]
Here is the result on ChatGPT 5 https://chatgpt.com/s/t_68b0c08f241c8191b940b795e19f4b15
Maybe it will be different in 2026, or Odoo 19, but currently ChatGPT 5 will spew out loads of useful sounding analysis, but naively concludes the wrong answer.
History tells me that the adage "you get what you pay for", will always be true.