Are AI Agents like von Hammerstein's industrious and stupid? | Hacker News Reader

Are AI Agents like von Hammerstein's industrious and stupid?

14 pointsmultidude10 comments

"I divide my officers into four classes; the clever, the lazy, the industrious, and the stupid. Each officer possesses at least two of these qualities. Those who are clever and industrious are fitted for the highest staff appointments. Use can be made of those who are stupid and lazy. The man who is clever and lazy however is for the very highest command; he has the temperament and nerves to deal with all situations. But whoever is stupid and industrious is a menace and must be removed immediately!” ― Kurt von Hammerstein-Equord

At first i was thrilled at the incredible speed of my new coding partner. Chatting with the agent felt like talknig to a real person, which blinded me a bit and made me forget many rational rules of software development. A few refactorings and many learnings later i find resonance in HN where people critizize which at the beginning i was so enthusiastic about.

I have this incredibly industrious partner, blazingly fast but mindless and oblivious. Am i the guy Hammerstein told us to avoid, or is it my Agent(or worse, is it the combination of us both)?

Imagine an industrious and stupid one with an AI Agent.

Some of this AI brings to mind the old Despair "Incompetence" poster (https://despair.com/products/incompetence)

When you earnestly believe you can compensate for a lack of skill by doubling your efforts, there's no end to what you can't do.

For those too young to remember, once upon a time people went into an office to work. And sometimes HR put up motivational posters. And this spawned humorous demotiviational posters.

If you're still confused, they're memes that printed, framed, and hung on a wall.

Despair is always with us. The "Adaptation" poster was prescient:

The bad news is robots can do your job now. The good news is we're now hiring robot repair technicians. The worse news is we're working on robot-fixing robots - and we do not anticipate any further good news.

I have gone back to manual coding to remove this AI induced laziness because i discovered immediately the LLM starts making some useless engineering decisions it does it like it knows what it's doing and also very deceptive and sometimes it's just easier to do it by myself other than trying to get the AI to do it,LLMS get so confidently wrong.

And it sounds so knowledgeable and authoritative in its replies. And i like being told "great idea"... OK, that is a joke i don't like it that much coming from a mindless agent. But in the dialog it becomes so convincing that i struggle not to anthropomorphize it. Remember "Her" with Joaquin Phoenix? i wont fall in love with it but i might entrust it with my root passwords?

Hammerstein's matrix works because armies have rank, evaluation, and consequences. The classification is useful because there's a system behind it that can observe, judge, and act.

But if "stupid" just means can't adapt and repeats what it's told, that's every script and cron job ever deployed. We used to call that automation. Humans can be stupid too, and we don't fire all of them. We put reviews, audits, and liability around them.

AI agents have none of that. A junior dev who breaks prod gets a postmortem. An agent that modifies 50 files gets a git revert. The problem isn't stupidity. It's that there's no structure around it.

All I know is that when I look through my clients' Jira boards, I see an explosion of bugs, outages, and other problems, starting 2 years ago and ramping up so fast that none of the teams can keep up.

If AI were working as well as people claim, I'd expect the opposite - work getting delivered faster than the product managers can spec it. Bugs speeding right through the SDLC process. Jira boards sitting mostly empty.

As I'm not seeing that, I'm inclined to agree that this whole industry is falling into "industrious and stupid"

Or maybe you are building more features than two years ago, which is reflecting positively on other KPIs.

You were building one house (feature) per year. After finishing the house you found a few issues (bugs). You get your plumber, electrician..etc to fix each issue.

After two years, you start building 10 houses per year. When you finish, you find twice or x times as many issues (bugs) as before. Does that mean your workrate and productivity declined or that things are going downwards?

I'm not saying AI is the joker and solving everything. I think there is a bigger picture to consider whether AI is having a good or no impact on the company you are working for. Did you release more features than before, what about MRR, CSAT,..etc?

The stupid+industrious problem gets worse in 1-on-1 setup. AI has no competing voice to push back against its own industriousness. Maybe the fix is putting it in a room with 30 people who disagree

AI is simply not an officer. It's a tool. At best, it's a stupid and lazy officer, mere information routers and supervisors.

It sounds smart, but it's autocomplete. If a group of students uses it to get ideas for an essay, the whole class will write the same essay, starring the same "Sarah Chen (no relations to Marcus Chen)".

That said, throughout history we've delegated decision making to tools. Bone reading, palm reading, face reading, dice, cards, horoscope, diviners, prophets. AI is just the latest, and 50 years from now we'll be shaking our heads at how the people in 2025 blindly trusted them.

The AI agent is certainly more industrious than you, and I want to believe you are smarter than an LLM. But relying on it too much makes you dumber.

This quote is exactly why giving shell privileges to any LLM is idiotic.

Reply on news.ycombinator.com