Summary
In this article, we posit that vibe coding, though an exciting technology, is not (yet?) well suited to use by non-technologists.
No Code in contrast has been specifically designed with non technologists in mind. It is however limited in scope.
Bigger gains can be found by combining no code and AI, but, crucially, in well thought through ways. You need a design purpose and philosophy which centralises the business person as user and designer.
Background – the AI reality
Recently we’ve been coming across lots of talk online about reality checks about AI in general and uses of it like ‘vibe coding’. On the other hand, we frequently meet people at conferences who hold far more positive views, and see other articles which are very boosterish.
To set out our own stall before discussing, we’ve always been reasonably cynical about most of the hype, certainly when it comes to AGI (artificial general intelligence) and magic productivity gains from throwing AI at everything.
However we do realise and indeed promote the use of AI for specific crafted and well thought out things. It’s used in many parts of the Agilebase platform and customers can build their own AI workflows. There are some great examples of major time savings which we’ll no doubt showcase in coming blog posts.
For quite a while I (Oliver) have been of the opinion that the capabilities of Large Language Models (LLMs) have been through their exponential growth phase. They’re now pretty amazing at generating written text, which has many useful applications, when you’re aware of and can work with the limitations (non-factuality etc.).
However the curve of improvement rate is flattening out. That’s not to say that there won’t be more and other great advancements in the field of computer science, but in the short term at least, they’ll probably be something completely different to LLMs. I could be completely wrong of course, things may get more and more refined until they reach a new level. As someone once said, it’s difficult to make predictions, especially about the future.
This skeptical take has not actually been a commonly expressed view and makes it a bit of a strange and maybe slightly lonely place to be. LLMs are a really interesting and sometimes useful technology, but there’s a general assumption that they can do practically anything, an article of faith that their ability will rise exponentially and an insinuation that if you’re not jumping on the bandwagon, you’ll be left in the dust
Particularly, as an optimist, one doesn’t want to be a lone voice of restraint and doubt when there’s lots of excitement and positivity around.
As I say though, there is evidently more concern around AI now and a more nuanced discussion arising.
For example, looking at today’s ‘Tech’ section front page on The Guardian newspaper’s website (20th Jan 2026), there are 5 headlines referencing AI, 3 of which are
UK exposed to ‘serious harm’ by failure to tackle AI risks, MPs warn
AI companies will fail. We can salvage something from the wreckage
For those interested in digging in deeper, here are few other posts I’ve come across recently:
https://lucumr.pocoo.org/2026/1/18/agent-psychosis
- a report from a self confessed AI coding addict, asking “Are we all collectively getting insane?”
https://deadsimpletech.com/blog/the_problem_is_culture
- an insightful look into the effects traditional IT culture has had on the domain
https://www.theregister.com/2026/01/16/linus_torvalds_vibe_coding
- oppositely to possible expectations, for serious work “even experienced programmers using AI tools took 19 percent longer to complete tasks”
Vibe Coding
So with that background, I thought it might be interesting to look at one particular part of AI which has some similarities with the No Code platform, Agilebase, which we’ve been developing over many years. That is, ‘vibe coding’.
The idea is, as the article from The Register above explains:
‘With vibe coding, the “programmer” describes their requirements in natural language to an AI model. The LLM then generates the code. […] You accept the AI’s output largely as‑is and iterate by rerunning and adjusting prompts rather than editing the code.’
What’s very enticing to think is that you don’t then need to know anything about coding at all. You don’t need to understand it. Which gets to the nub of the issue. Democratisation isn’t the purpose of vibe coding, it’s just potentially a side effect.
Who uses vibe coding and how did it arise? Well, at some point people noticed that LLMs were good at creating working computer code – not surprising, since a programming language is actually a lot simpler grammatically and more structured, or ‘machine-like’ than natural language.
However, and this is what gets my goat so to speak, no one has made a serious attempt to make this accessible to non-technologists. Instead, they have found a great toy which they can be constantly amazed at, and keep playing with amongst themselves, getting more and more hyped up.
The upshot is that although showing some promise, it isn’t suitable at all for use by anyone not already steeped in computer science and coding. As Iris Meredith describes in an article previous to the one above, https://deadsimpletech.com/blog/week_with_opencode, just one example is that
‘much of the difficulty in actually making use of real software is in building and deploying applications and maintaining the infrastructure backing them rather than simply writing [the code]’.
Will that change? Possibly in time, as the market matures, but we’re not going to see that sort of thing from Silicon Valley, we never do. It will take other people and industries (or at least people with other cultural north stars) to come along and make use of the tech, to the extent that’s possible, which may also change over time.
At Agilebase we’ve always had a more humanist ethos. When starting the company, we saw how traditional software developers were not serving their customers well. As the respected Standish reports at the time said, most large projects were failures, due to misunderstandings and miscommunications between the IT suppliers and the businessworld customers.
We wanted to give the business people themselves more control over creating their own software, to map exactly to the way they wanted to work. From our existing customer experiences, I think we’ve achieved that. It’s been done with graphical, easily understandable interfaces and lately with AI being woven in.
No Coding
So having dunked a bit on vibe coding, to be fair, what are the downsides of No Code?
First of all, it doesn’t remove complexity from the equation. We like to say that “no code is not no thought”. That’s in contrast to vibe coding, where the aim is to dumb down the user and trust the AI (which may or may not work!)
For example, if you are building a complex stock control system with lots of rules about stock re-order levels, best before dates, stock movements etc., you actually have to understand everything about it. You can build up workflows and calculations visually, seeing the connections and data flows, but it doesn’t remove the need for experience and knowledge.
Hence many customers find that they are best served with the help of an expert partner. Development is rapid and collaborative. People can dip in at any level, from making simple changes to the user interface, moving or adding fields for example, to becoming a complete system architect.
Secondly, the domain of usage is limited. With vibe coding, you can theoretically build any type of application you like. With No Code, what you can build depends on the No Code platform and its designed use cases.
In the case of Agilebase, that is
- business or organisational applications, as opposed to consumer or personal
- which work with structured data
- to run operations – everything which keeps the business ticking over and which supports growth. That covers a large amount of systems: stock control, ERP, membership systems, event management, ordering, staff rostering, approval systems etc. etc.,
Thirdly, No Code platforms can often be limiting in the range of real world complexity they can deal with. They are more suited to small scale usage, rather than more serious usage. That’s where we at Agilebase set ourselves apart. Our USP is that the platform has enough capabilities to grow with you as a business, from startup to scale-up, or grow throughout all departments of an enterprise. In many cases, our systems form the backbone of operations throughout the whole organisation.
Bringing things together
A while ago I was struck by a realisation that came from the chess world, back when Gary Kasparov was first beaten by an IBM computer. People saw that never again would any person be able to beat the best computer program. However, the interesting thing was that the combination of human and computer players still came out on top in competition, against pure human or computer.
I’m not sure if that’s still the case, I’m not a chess follower, but the same thing has since shown up in many other fields such as medicine. The most effective work is done with people who are also able to draw on AI tools.
It’s just an extension of the truism that the best teams are diverse ones. Adding AI agents and capabilities is another way of diversifying, but importantly realising that AI is not equivalent to a person – it is quicker at many things, but worse at many others.
Where we’ve used AI in Agilebase, it’s been bearing that in mind. A couple of examples are
- creating complex (or just simple) calculations. AI is really good at that (generating SQL code), but someone still has to understand what they’re trying to achieve when they’re creating the calculation. It’s a part of a wider whole.
- a tool for running prompts against data en masse from the database. For example, creating an AI workflow to classify all contacts into groups given their organisations and job titles, to allow salespeople to select and report effectively.
My favourite comment about software development came as an answer by Hazel Weakly on LinkedIn, to the question ‘what is good software?’
She says that software is “the shared collective understanding of a concept made concrete”. With collaborative teams including end users (the business domain experts) and experienced Agilebase technical partners, we can create that collective understanding and make it concrete, quickly and correctly, using the No Code and AI capabilities of the Agilebase platform – at least within our limited domain of operations software for businesses.
Here’s one example of that happening: