551–560 of 809 posts
Earlier quoted context omitted.
It really doesn't have to come for everything to feel like it's taking everything. If it eliminates 10% of white collar jobs over the next decade, the impact will be felt everywhere.
Sure, but who doesn't think that 10% of white collar jobs are mostly bullshit anyway?
Earlier quoted context omitted.
Most of the time is spent figuring what the right thing to do is, not writing the implementation. Sometimes the process of writing the implementation surfaces new considerations about what the right thing is, but still, producing text to feed to a compiler is not the bulk of the work of a software engineer. It is to unearth requirements and turn them into repeatable software.
AI is pretty good at figuring out what the right thing to do is.
Unless I'm missing something, there's an obvious logic issue here. If we truly need to sacrifice our skill to be productive by using LLMs that atrophy us, then the only devs that have a limited lifespan are us. The next ones won't have a skillset to atrophy since they won't have built it through manual work. Also, I hereby propose to publicly ban the "LLMs generating code are like compilers generating machine code" a…
why is the LLM-compiler analogy flawed? Is it only because LLM output is non deterministic?
Earlier quoted context omitted.
People always say this but it’s misguided imo. Yes LLMs are not deterministic, but that’s totally irrelevant. You aren’t executing the LLMs output directly, you’re using the LLM to produce an artefact once that is then executed deterministically. A spec gets turned into code once. Editing the spec can cause the code to be updated but it’s not recreating the whole program each time, so why does determinism matter?
If it's not deterministic you can never fully trust it. In a deterministic abstraction I don't need to audit the lower levels.
Earlier quoted context omitted.
LLMs moved the moments of despair to PR reviews for me. It used to be that you could check on a junior dev occassionally throughout the day to make sure they're on the right track. Now you step away for 2 hours and they're raising a PR of bad code smell spaghetti and moving on to repeat their AI slopfest on the next task. It's getting hard to keep up with trying to teach new devs what bad code looks like. And I swear…
Write a damn good automated review agent that runs against their PRs before even looking at them… works well for me!
How have you set yours up that works well for you?
Multiple times per week I have the same conversation. It goes something like this: - AI will make developers irrelevant - Why? - Because LLMs can write code - Do you know what I do for a living? - Yes, write code? - Yes, about 2-5% of the time. Less now. - But you said you are a developer? - I did - So what do you do 95-98% of the time? - I understand things and then apply my ability to formulate solutions - But I ca…
If you’re a developer and you’re writing code 2% of your time pre-Claude, that’s 9 minutes a day, you will and should be fired.
- Meetings
- Code reviews
- Manual testing
- Deployments and more testing
- Triaging issues
- WTF how did this bug happen?
- JIRA in general
- Whiteboarding sessions / Design docs
- Interviews
- 1:1s (mandatory ones)
- 1:1s (networking / problem solving / political alignment)
- Whatever your company's version of corporate extracurriculars is
This requires having an understanding of a business domain, economics, human psychology and technology.
The competitive aspect of it means that you need to understand these things better than most people and machines. If you don't, then your skills have no value on the market. Will generalist AI trained on public data ever understand these things better than software engineers across every possible niche?
I don't think so because that knowledge is usually gate-kept. Nowadays, new engineers almost have to beg to be given access to knowledge of company systems. It takes at least 6 month for a skilled engineer to ramp up on large systems... And it's mostly because of institutional resistance.
The thing is, it doesn't even require people to be withholding information... Some engineers will happily share everything they know about internal systems... But in a big company; you first have to identify this person. That can take a while... Then you need to identify other persons who will give you other information that is relevant to your specific tasks/integrations. Then there are all sorts of other constraints and restrictions to deal with.
You can't just deploy an AI to a big company and it will magically guess all the endpoints which exist... You have to tell it what is available and enterprise systems are not designed for transparency.
Big companies resorted to a kind of security-through-obscurity. This used to be considered bad practice 10 years ago but at some point they just gave up, let complexity run amok and started calling it "multiple layers of defence" but now this obscurity is a problem for evaluating system security (too much unknown context is required, nobody fully understands the entire system) and it slows down development and maintenance as well.
Whoever knows the most context about a system has the advantage... And this isn't necessarily a company insider. Most likely, the people with the most context are platform providers.
I predict that most major hacks will originate from platform providers. We already started seeing this with Axios hack (originating from GitHub/npm) and Vercel (originating from Google Workspace).
The centralization risk is massive because each platform is servicing so many large companies. It only works when there is perfect incentive alignment but that's not usually the reality during difficult economic times. Third-party platforms cannot be trusted anymore.
[flagged]
Multiple times per week I have the same conversation. It goes something like this: - AI will make developers irrelevant - Why? - Because LLMs can write code - Do you know what I do for a living? - Yes, write code? - Yes, about 2-5% of the time. Less now. - But you said you are a developer? - I did - So what do you do 95-98% of the time? - I understand things and then apply my ability to formulate solutions - But I ca…