Nobody sane and smart I know is advocating for AI now. All keep saying 'one day', 'soon' etc. Our megacorp very effectively blocked all of this and we don't have anything internal yet, but we are in specific market.
And basically all I saw so far was a snake oil from people trying to push it or benefit from it, even here. Trust me bro... well no, show me some hard facts and we can start building trust. Sure assistants are nice but thats about it.
Its just too flawed now to be dependable and reliable, it feels like typical 80/20 or 90/10 situation where inexperienced juniors are all over their head how future is now, and seniors just meh and go and do some work.
Advocate or not, it’s here to stay. And if you’re not adjusting to it, you’re doing yourself a disservice.
Last week I did the amount of work that would’ve taken me give or take a month. A significant part of it was writing an API client for a system I needed to use. Pretty run of the mill stuff. Doing this ‘by hand’ just takes time. Go look at the docs, type out your data structures, wire things up for the new call, write tests. With “the robot” once the framework was largely in place, you just paste API docs for the endpoint you need, and it’s done in a minute. With tests and everything.
Now, you could argue that this is not “senior level work”. Sure, maybe. But the thing is, you need to use it to learn how to use it effectively, and more importantly to get the intuition about what it can help with best.
So this is my thesis now - if you’re a senior who neglects this new tool, you’re going to be out of a job soon-ish.
> Last week I did the amount of work that would’ve taken me give or take a month. A significant part of it was writing an API client for a system I needed to use. Pretty run of the mill stuff. Doing this ‘by hand’ just takes time. Go look at the docs, type out your data structures, wire things up for the new call, write tests. With “the robot” once the framework was largely in place, you just paste API docs for the endpoint you need, and it’s done in a minute. With tests and everything.
That just points to an inefficiency. Could be tackled in other ways than involving an LLM to produce essentially what's being done elsewhere every day over and over again. A framework automating and hiding all this would be just as effective. Perhaps even cleaner than all that duplication that the LLM created for you.
In other words, that month of busywork that you just saved is inherently unnecessary to do. But progress is not linear in the number of lines of code that you produce. If you think hard about a good architecture and design, coming up with that after 2 weeks of hard thinking, that could be 90% of the work. The remaining 10% are writing all that down. That could take 3 more months, and it taking more than 10% of the time points to the existing inefficiency in tooling / framework / ... But making that more efficient isn't necessarily achieved the best by using an LLM and writing it all out. There are still huge redundancies, which is what made the LLM possible. Once you boil these down in some common frameworks or tools, the LLM will also just produce the same few lines that you'd need to produce to get those 10% done in then just 10% of the total time.
> So this is my thesis now - if you’re a senior who neglects this new tool, you’re going to be out of a job soon-ish.
Work(ed) for a company super big on a force of agents. Brought them into my workflows early on, before the company even pivoted. Ran them at home on private hardware (that 3090 had to do something), at work when access was opened up. Provided valuable feedback on hallucinations (e.g., fabricating the existence of a MongoDB CLI module wholesale, documentation and all) and barriers to adoption. Pitched an integration of four disparate systems across ~20 data points to create tenancy in a core product line from scratch, with said AI at the center of it for customer service and workload creation, reducing engineers to approvals only for ~20 tickets a week averaging ~40hrs of work, freeing up said engineers for actually valuable work instead of handholding customers through routine tasks. Got great feedback and enthusiasm on it.
RIFed this year.
Doesn't matter whether you're bullish or bearish on it, the only thing that ever really matters is if scrubbing your line item from a spreadsheet will net someone higher up than you a bonus. That's what's driving this AI mania, and what needs to be addressed via policy before these employers gut themselves and the economy in the process.
Oh, 100%, and I knew that going in. Still, I'm the type of worker (sucker? sap?) that will go down with the ship if it helps the rest of the folks get into lifeboats. I know I can take care of myself when stuff hits the fan, and I'm never one to half-ass my output because of my own cynicism.
I'm hoping that pays off for me someday. Thus far, it's just been a lot of burnout/layoff cycles.
> Still, I'm the type of worker (sucker? sap?) that will go down with the ship if it helps the rest of the folks get into lifeboats
> engineers to approvals only for ~20 tickets a week averaging ~40hrs of work, freeing up said engineers for actually valuable work instead of handholding customers through routine tasks
Not trying to be too judgy here but to me it sounds more like you are the one helping sink the ship than help everyone to life boats
Something to think about. Agreeing to bring AI into our workflows is actually digging out the foundation beneath our own feet (and our coworkers feet)
Sure, but what choice do you have as an individual? Try and adjust to the new reality and maybe get RIFed anyway, or not adjust to it and get RIFed with 90% certainty. Power looms won.
Totally agree with you on the line-item point. My point there was that if a 50% RIF is in the works, the people who haven’t figured out how to become more efficient using AI would be more likely to be affected. Statistically.
But again, hope you’re doing well and everything has worked out for you!
I’m a month in, and it’s rough as hell. But I at least have a runway for the first time in my career, so I’m profoundly grateful for that.
Still optimistic that my generalist skillset and adaptability to new technologies (like AI) will make me an asset for IT teams who can’t afford multiple specialists in rough economic times like these.
But a solution that takes an OpenAPI spec and generates all of those things from templates would be much more reliably correct than letting an AI do it, and has already existed for years.
I think AI is a game changer but not for things that can already be automated rigirously without AI.
Eventually only the arquitect will be required, no need for seniors, the offshoring team is the cloud instead of some folks on the other side of the planet.
This is the future to prepare, not AI generating a couple of files.
Maybe. But that future is definitely not yet here. Generating a couple of files though is here. And that compounds. So people who’re still typing those files by hand… yeah.
Can you give some examples where “architect, no coders” is here already? Asking, because I’d rather try and adjust to the new world, than pretend it’s not happening!
You have to balance this time savings with all the times the LLM is just making shit up. I can't the number of times I ask "how do I do X?", and says "call function Y" or "Use command line argument Z". Turns out Y doesn't exist and CLI doesn't take argument "Z". The frustration from that lying is immense. I'd rather take a little longer on some tasks and never deal with this lying.
I see what you mean. The way I approach this is that I think of it as a very (very, very) elaborate auto-complete. Something that takes some text, and continues with the most likely continuation. Which is what it basically is, right?
But that means that whatever rather impressive internal representation of knowledge it has encoded, that knowledge itself can’t be trusted on a “factual” basis.
This is where you have stopped, but you shouldn’t. Because you’re missing what it’s great at. And it’s great at “continuing” text in a longer context. So just use it with a longer context. Use the Search functionality in ChatGPT when asking it things such that you see where it got it from. Use it in a codebase context where you point it to the relevant parts. Give it examples it should follow, etc.
For every task I have I try using AI in some way and form first. Because the downside is limited - I waste 5-7 minutes figuring out that it’s not going to work, and this is something I definitely need to do myself. But the upside is unlimited - the task is done in minutes.
If I waste 5-7 minutes multiple times every day, that quickly adds up. It's easy to waste an hour or more fiddle farting with these unreliable tools. It's easier to just do it myself.