Replies to comments on my "LLMs are eroding my career" post
So, my latest post went viral.
And with virality, you get a ton of comments that you must reply to.
I don't want to reply in HN/Reddit/Whatever to prevent endless thread depths that will consume my sanity. I'm going to cherry-pick some comments and leave answers here for those looking for them.
Wut? I pilot LLMs all day but there's no way in hell I'd agree to be at the helm of a finance product.
LLMs routinely fail at our business specifics: Local tax regulations
I should have been more explicit about it. LLMs are not automating everything when it comes to local tax codes or very fine-grained details, but this is usually handled by the legal team here (which is also automating a lot of routines with LLMs as well).
But much of the domain knowledge I mastered over time (which is obviously shallower than what the legal team is doing) is now just promptable with ChatGPT Pro/Extended Thinking.
That's what makes me sad, I thought that having this knowledge would set me apart in a world of coders that just know how to code, but that's not the reality anymore.
particularities of the accounting process, specifics of our ledger implementations.
Agents used to be bad at this kind of stuff in my workplace as well, but newer models + agent-friendly documentation + AGENT.md begging agents to read the fucking docs before coding changed this landscape for us here.
Less and less I feel the need to reach out to coworkers that have been around for longer and knew something in detail. I need much less human input to do my work now, which is freaking scary when I stop to think about it.
Also, a fintech whose managers recommend speeding up design docs with AI sounds way too careless to be in the money handling business.
Yeah, I don't agree with that as well, and my workarounds have been:
make the docs somewhat generic on the implementation details, state machines and such, so I have room to make a thought-out implementation. After AI came in (and then the layoffs), everybody is drowning in long docs to read and PRs to review, so the reviewers are way less picky now. This gives me room to work around flaws in the initial doc.
do some juggling with the team board to buy me time. For example, I always add tickets for E2E tests, in which I can find bugs, which gives me room to file bug/improvement tickets before the feature is released into the wild. This also gives me more time to review the implementation with caution. I also break the initial parts of the implementation (which are usually more sensitive) into more cards than I usually do so I have some room to implement and review it with caution
Do I like doing this? Of course not, but what are my options? The reports I get from people I know are that my company is not on the extreme edge of vibecoding, so leaving it for a potentially worse environment is not a good trade. At least I'm in a place where I know how to control the anxiety of the stakeholders (my diligence and caution earned me a good reputation) and that does not forces me into full-throttle vibecoding.
Ride the wave. You rode it when websites/webapps were the wave. I came into software industry before internet, kept changing my horse. You are never too old to learn new tricks. The new wave create new kind of work and workers. Be one of them. Ride the beast, master the tools. It's the same game again.
Yeah, that's what I'm doing right now. I'm one of the engineers who's constantly committing to improve our agentic tooling, I use different models to do adversarial code reviews, I keep a toolbelt of skills and prompts, etc. I have effectively become the so-called "AI-native engineer" (gosh, I hate that term).
I'm more concerned about the future.
If the models (and harnesses) keep getting better at the same pace for the foreseeable years, we are heading to a world where the profession is commoditized to the ground. There's this talk about Jevons Paradox but I disagree. The demand for software most certainly has an upper limit.
Take copywriting. It was a profession that took years to master and paid well. This changed slowly as more professionals joined the market, even after the demand spike driven by ecommerce and adtech. Now, LLMs have destroyed the job for the vast majority of professionals.
It's simple to see why: the bulk of the demand was from smaller companies that needed copies but are well-served with ChatGPT-generated copies. Some are still hired just to prompt, review and send the copies, but since the demand is not infinite, not all of them can be hired to do that. One copywriter is now doing the job of 10, but the demand is fixed. The demand is not going to 10x just because you have 10x more supply.
Of course, the very best copywriters are still employable, but they are the ~1%. The other 99% are fighting for scraps. UX Writer used to be a career with good prospects. Now the ones I know were all laid off. Even large orgs fired them; you can just prompt ChatGPT to come up with text labels, and that's ok 90% of the time, so having 10 professionals in your payroll is not justified anymore, fire 9 and keep 1.
We are all heading to the same fate if the models continue to improve in the same direction.
We will have some engineers hired to steer the agents, sure, but they'll be replaceable and cheap (supply and demand, again). That's what I'm highlighting in the previous piece. And that's coming for other professions as well: we all know software is the low-hanging fruit labs are targeting now. They will come for finance, biology, law, marketing, all knowledge work. That's their stated goal and they're already teasing it with "ChatGPT for Health" and similar launches. They're working on "harnesses" for other fields, it's just a matter of time before we have "Claude Finance Analyst" or something.
This anonymous article is likely more FUD from the AI industry. "Just give up,you can't beat the machine. Please go quietly, we want to take your place and it's easier for everybody if you don't resist because you believe it's pointless"
So blog with single post hyping LLMs. Oh and the domain name "human-in-the-loop". Call me suspicious.
If after reading what I just said in the reply above you still think I'm an "AI shill" or "lab shill", there's nothing I can do for you.
Overall society feels more turbulent, but this is otherwise all the same song and dance all over again. The 90s and 00s had this wave of "object oriented programming changes everything".
OOP didn't make knowledge promptable. OOP didn't show signs of fast, compounding improvements heading in the direction of replacing that many workers across areas not limited to software engineering. This is not the same, people.
We have this tendency to think that the past predicts the future (no serious Historian believes this, btw), but sometimes extraordinary things happen. Take Covid-19. I remember when it broke out in China, people dismissed it because of the recent memory of 2009 Swine Flu, the 2014 Ebola outbreak or SARS. We all know how it ended. And it could have been a lot less dramatic if we hadn't dismissed it at the start.
Now something similar is happening, and people are dismissing it because of OOP, Metaverse or NFTs. Stop and think, don't try to predict the future using (bad) past examples.
I don't want to fearmonger, but we do have something that's bigger than OOP now. We have built a matrix multiplication machine that (given the appropriate harnesses, tools, and prompts) can output useful text strings for hours in a row. This is sci-fi stuff. We should act accordingly.
If the author's vision of the future is correct, then competent software engineers are safe. Domain knowledge can be learnt much quicker than how to apply good engineering principles.
Beg to disagree. The models will learn good engineering principles at some point.
Just an example: there's a company called Turing AI that's hiring engineers to write "good code" across all kinds of domains and languages that's then used for reinforcement learning at the labs. The "human moat" on that is not going to last forever.
I think the one thing the author is underestimating, especially in his "first pillar" is that he is able to coach the models into great results because he already knows the lay of the land.
God willing you’re right, but I'm finding LLMs also competent at explaining and giving advice on other domain stuff I'm totally new to, which I have cross-checked with Legal/Product Managers and is usually right.