It seems like investors have bought into the idea that llms has to improve no matter what. I see it in the company I'm currently at. No matter what we have to work with whatever bullshit these models can output. I am however looking at more responsible companies for new employment.
I'd argue a lot of the current AI hype is fuelled by hopium that models will improve significantly and hallucinations will be solved.
I'm a (minor) investor, and I see this a lot: People integrate LLMs for some use case, lately increasingly agentic (i.e. in a loop), and then when I scrutinise the results, the excuse is that models will improve, and _then_ they'll have a viable product.
I currently don't bet on that. Show me you're using LLMs smart and have solid solutions for _todays_ limitations, different story.
Our problem is that non coding stakeholders produce garbage tiers frontend prototypes and expect us to include whatever garbage they created in our production pipeline! Wtf is going on? That's why I'm polishing my resume and getting out of this mess. We're controlled by managers who don't know Wtf they're doing.
Maybe a service mentality would help you make that bearable for as long as it still lasts? For my consulting clients, I make sure I inform them of risks, problems and tradeoffs the best way I can. But if they want to go ahead against my recommendation - so be it, their call. A lot of technical decisions are actually business decisions in disguise. All I can do is consult them otherwise and perhaps get them to put a proverbial canary in the coal mine: Some KPI to watch or something that otherwise alerts them that the thing I feared would happen did happen. And perhaps a rough mitigation strategy, so we agree ahead of time on how to handle that.
But I haven't dealt with anyone sending me vibe code to "just deploy", that must be frustrating. I'm not sure how I'd handle that. Perhaps I would try to isolate it and get them to own it completely, if feasible. They're only going to learn if they have a feedback loop, if stuff that goes wrong ends up back on their desk, instead of yours. The perceived benefit for them is that they don't have to deal with pesky developers getting in the way.
It's been refreshing to read these perspectives as a person who has given up on using LLMs. I think there's a lot of delusion going on right now. I can't tell you how many times I've read that LLMs are huge productivity boosters (specifically for developers) without a shred of data/evidence.
On the contrary, I started to rely on them despite them constantly providing incorrect, incoherent answers. Perhaps they can spit out a basic react app from scratch, but I'm working on large code bases, not TODO apps. And the thing is, for the year+ I used them, I got worse as a developer. Using them hampered me learning another language I needed for my job (my fault; but I relied on LLMs vs. reading docs and experimenting myself, which I assume a lot of people do, even experienced devs).
When you get outside the scope of a cruddy app, they fall apart. Trouble is that business only see crud until we as developers have to fill in complex states and that's when hell breaks loose because who tought of that? Certainty not your army of frontend and backend engineers who warned you about this for months on end.....
The future will be of broken UIs and incomplete emails of "I don't know what to do here"..
The sad part is that there is a _lot_ of stuff we can now do with LLMs, that were practically impossible before. And with all the hype, it takes some effort, at least for me, to not get burned out on all
that and stay curious about them.
My opinion is that you just need to be really deliberate in what you use them for. Any workflow that requires human review because precision and responsibility matters leads to the irony of automation: The human in the loop gets bored, especially if the success rate is high, and misses flaws they were meant to react to. Like safety drivers for self driving car testing: A both incredibly intense and incredibly boring job that is very difficult to do well.
Staying in that analogy, driver assist systems that generally keep the driver on the well, engaged and entertained are more effective. Designing software like that is difficult. Development tooling is just one use case, but we could build such _amazingly_ useful features powered by LLMs. Instead, what I see most people build, vibe coding and agentic tools, run right into the ironies of automation.
But well, however it plays out, this too shall pass.
The amount of functionality an application provides should be the benchmark for the size rather than one's available disk space. I'm sitting on 20TB+ of available storage and anything over 1MB for a simple "Hello World" is excessive.
Cheap laptops with a Celeron, 4 GB RAM, and 64 GB storage have been incredibly popular since that formula became popular a handful of years ago. As are base-model MacBook Airs.
Well, if we already could make something in 1 MB, why do we want to achieve the same thing with 100 MB? It is as if there is a mentality of "somehow the higher the size the better".
In most reasonably-sized websites, Tailwind will decrease overall bundle size when compared to other ways of writing CSS. Which is less code, 100 instances of "margin-left: 8px" or 100 instances of "ml-2" (and a single definition for ml-2)? Tailwind will dead-code eliminate all rules you're not using.
In typical production environments tailwind is only around 10kb[1].
You're doing it wrong. Tailwind is endlessly customizable and after compilation is only kilobytes. But yes lets complain because we don't understand the tooling....
They can seriously screw themselves with the rugpull they made. I reported so many bugs and helped diagnose them in the sentinel module. Seriously fuck them.
What do you do when it flags you or someone you know who's innocent? Blindly trusting these models without any verification will put innocent people in prison. Normal people don't understand why they are so confident. They're confident because they believe all the data they have is correct. I forsee a future with many faux trials because they don't understand critical thinking.
> Blindly trusting these models without any verification will put innocent people in prison.
I don't think anybody is suggesting this. But if the models can gleam information/insights that humans can't, that's still valuable, even if it's wrong some percentage of the time.
This is what happened with dna testing at the beginning. Prosecutors claimed it was x percentage accurate when in fact it was hilariously inaccurate. People thought the data was valuable when it wasn’t.
I don't see that as particularly analogous. The average person will have had LLM technology in their own hands for years, whereas with DNA it was completely foreign to them and their only choice really was to trust the experts. And on top of that DNA testing matured and is very useful now.
He's a total legend, yet apparently he's never met Bill Gates in person from what he said in an interview in the Dave's Garage YouTube channel a few years ago. You'd think that someone who's been that prominent for so long in the company would have been invited to a company dinner where he was present or something.
He has stories on his blog about windows 2 iirc, so there was an overlap from a time where they were still relatively small. So I think it's a bit odd they never talked or met.
I wonder how many times a Deloitte, PwC, KPMG, Bain, EY, McKinsey, or BCG consultant naively tried putting him on a shortlist for being “impacted” over the years because he was in the Top X of a spreadsheet sorted on Y.
"Look this guy's job seems to be mainly writing blog posts. We could replace that with AI and get it to regularly pitch the new Visual Enshitify 2.0 product launch as a bonus. Win win win!"
In America. If you work in other markets the salary is much lower. The only engineers I know who make above 100k euro are either lead or founder level.
100k USD is about 90k EUR (and 77k GBP), which you may say doesn't change your point materially, but I'm not splitting hairs, that's more than a 10% difference.
I think people often put the goalposts further away than they truly are by comparing numbers cross-currency directly like this or just round numbers in general.
It is possibly to earn over £100k as a software engineer employee in the UK, however those tend to be very specialised fields. (Not all of them are finance related).
I do recall a software manager (OK apples to oranges) grumbling about 10 years ago about him exceeding that level, and having to play games to avoid being bit by the tax impact which cuts in between £100k and £125k.
That said, I do know engineers who indicate they earn in excess of £100k, and they're not lead nor founder level - just experienced in the appropriate area.
I think it's a worthwhile addition to highlight there is 3) rules which are sometimes red tape and sometimes to be broken, on top of the other 2 categories. It adds on to the original point with the addition of how to universally discover what the categories are rather than prescribe them up front.
To add to that, #3 is often explicitly encoded into the red tape as an escape hatch for foreseeable exceptional circumstances like disaster recovery and big client emergencies.
