I genuinely don't understand why some people are still bullish about LLMs

As others have pointed out already, the hype about writing code like senior engineer, or in general acting as a competent assistant is what created the expectation in the first place. They keep over-promising, but underdelivering. Who is the guy talking about AGI most of the time? Could it be the top-executive of one of the largest gen AI companies, do you think? I won't deny it has occasionally a certain 'star-trek-computer' flair to it, but most of the time it feels like having a heavily degraded version of "rain man". He may count your cards perfectly one moment, then will get stuck trying to untie his shoes. I stopped counting how many times it produced just outright wrong outputs, to the point of suggesting literally the opposite of what one is asking of them. I would not mind it so much, if they were being advertised for what they are, not for what they could potentially be, if only another half a trillion dollar were invested in data-centers. It is not going to happen with this technology, the issue is structural, not resource-related.

> Just amazing, doing things we dreamed about for decades.

Chatbots like in the sci-fi of your nostalgia? I never dreamed about that shit, sorry.

Im not saying LLMs are not impressive or useful — Im pointing out that corporations behind commercial AI models are capitalising on our emotional response to natural language prediction. This phenomenon isnt new – Weizenbaum observed it 60 years ago, even with the simplest of algorithms like ELIZA.

Your example actually highlights this well. AI excels at language, so it’s naturally strong in teaching (especially for language learning ;)). But coding is different. It’s not just about syntax; it requires problem-solving, debugging, and system design — areas where AI struggles because it lacks true reasoning.

There’s no denying that when AI helps you achieve or learn something new, it’s a fascinating moment — proof that we’re living in 2025, not 1967. But the more commercialised it gets, the more mythical and misleading the narrative becomes

When digital cameras first appeared, their initial generations produced low-resolution, poor-quality images, leading many to dismiss them as passing gimmicks. This skepticism caused prominent companies, notably Kodak, to overlook the significance of digital photography entirely. Today, however, film photography is largely reserved for niche professionals and specialized use-cases.

New technologies typically require multiple generations of refinement—iterations that optimize hardware, software, cost-efficiency, and performance—to reach mainstream adoption. Similarly, AI, Large Language Models (LLMs), and Machine Learning (ML) technologies are poised to become permanent fixtures across industries, influencing everything from automotive systems and robotics to software automation, content creation, document review, and broader business operations.

Considering the immense volume of new information generated and delivered to us constantly, it becomes evident that we will increasingly depend on automated systems to effectively process and analyze this data. Current challenges—such as inaccuracies and fabrications in AI-generated content—parallel the early imperfections of digital photography. These issues, while significant today, represent evolutionary hurdles rather than permanent limitations, suggesting that patience and continuous improvement will ultimately transform these AI systems into indispensable tools.

My take is that if you expect current LLMs to be some near perfect, near-AGI models, then you're going to be sorely disappointed.

If that disappoints you to such a degree that you simply won't use them, you might find yourself in a position some years ahead - could be 1...could be 2...could be 5...could be 10 - who knows, but when the time comes, you might just be outdated and replaced yourself.

When you closely follow the incremental improvements of tech, you don't really fall for the same hype hysteria. If you on the other hand only look into it when big breakthroughs are made, you'll get caught in the hype and FOMO.

And even if you don't want to explicitly use the tools, at least try to keep some surface-level attention to the progress and improvements.

I honestly believe that there are many, many senior engineers / scientists out there that currently just scoff at these models, and view them as some sort of toy tech that is completely overblown and overhyped. They simply refuse to use the tools. They'll point to some specific time a LLM didn't deliver, roll their eyes, and call it useless.

Then when these tools progress, and finally meet their standards, they will panic and scramble to get into the loop. Meanwhile their non-tech bosses and executives will see the tech as some magic that can be used to reduce headcount.