Will AI take your job?

It's incredible people are still in denial about what is coming.

You mention hardware needing to be replaced and upgraded. But if this is so due to obsolescence, then isn't this a prediction of the importance of AI rather than an indication of it being an over-hyped flash-in-the-pan? After all, if this AI thing was just a fad, then we would expect to see a glut of compute power in the server racks.

They haven't said that since the 40s.

They will change their tunes when they get outcompeted by competitors using other tools, for example Cursor and Amazon’s Kiro are getting ridiculously good

Like I mentioned earlier the danger to people is not AI replacing them but other companies and people using better AI tools and models outcompeting them

You might want to raise this with management

I use copilot to do pretty useful stuff. I think People don't spend enough time with it and give up too be easily. Other tools are better but that's no use to me if they aren't available in the job.

That has not been my experience. I have worked almost my full career in US multinationals and the training budgets are generous. Most people dont use them. People want to be spoon fed. Training and upskilling takes hard work and effort.

It is monetised. It's quite expensive for companies to buy. I expect in companies some people will have it and some people will not. It's too expensive to give to everyone if they aren't using it productively. Which means they will measure who uses it and who doesn't.

I did a little experiment at work a few weeks ago. A few guys in my team has been given the charge of Claude adoption for the last 3 or so months. We got a question from a business user about what seemed to be an erroneous order (trading) , it looked to them like it was missing some data. Basically theses orders present as a graph structure with multiple nodes on a hierarchy. I am probably an expert, given the questions I'm the org always funnel toward me when they get difficult enough. So I told the guys to use Claude, spend a day or two, see if it could beat me. I was a bit nervous and also curious but in fairness I had the deck stacked in my favour, given I have the context in my brain. Anyway Claude failed miserably but even worse it failed in a very ungraceful way. The guy had some context , so he did give it information on structure, it had direct access to all databases that applied. It completely hallucinated a solution. At first I got a shock, it confidently went down a path I had never considered before, using different identifiers, different pattern. I was worried and excited I would learn something. But it was complete nonsense. The main weakness was this:

The question from the quant was asserted with authority(these clowns often do this and are often wrong, it's a strategy they use to expedite an investigation by being alarmist) but he made a few incorrect assumptions implicit in his question. The questioner did not understand the data himself. Claude treats the questioner as some kind of all knowing God, so it will jump through hoops to satisfy the questioner who has bad context. These llms are highly biased to satisfy the limited user that engages it, they do not push back and challenge us enough, which may make us feel satisfied with small test cases, reinforcing our own assumptions about problems but actually create large issues in the future potentially. Even if so can write code better than devs, it has no value unless there are devs that understand it and we lose the value of feedback and corrections in our understanding that we get by struggling with logic during implementation. Code is not a commodity , it is also a tool we use to teach ourselves and challenge our logic.

Then I'd question the value of ai. Generating code is about 5% of the effort of solving the problem.

And that’s why I don’t buy the whole AI will kill software engineering meme and last weeks saaspocalypse where various software companies lost massive amounts of their stock price

Writing code and tests is usually a small fraction of an engineers day, and these companies that hire engineers still have much longer backlogs and feature wish lists than even with these tools increasing productivity by up to 25% would be able to address anytime soon, and of course amount of work always grows

As mentioned in stocks 2026 thread I was busy cherrypicking great companies last week that have wide moats and platforms that will only grow larger as they have access to same ai tools and whose engineers will use these tools.

Just look at this site. Despite ai tools it remains buggy and hard to use and continues to be outcompeted and grown by Reddit engineers using AI tools

Why not iterate until it was accurate. How often is code perfect v1.0

The reason this site is buggy it more to bloody minded stubbornness than anything to do with tools. Scope was too large for the resources and those resources were further under resourced. Rather than cutting their cloth to fit the resources they just kept digging a deeper hole. It has never recovered.

AI isn't automatic pilot. It's needs to be guided. But it needs someone who understands how to use it to ask it.

AI will iterate at an ever faster pace and AI companies will put frameworks and specifications in place to drive that. Time will tell but compare any car from today with a car from 100 years ago. That's what will happen with AI/Software just much quicker.