Less human AI agents, please

Source: Hacker News

AI agents are already too human. Not in the romantic sense, not because they love or fear or dream, but in the more banal and frustrating one. The current implementations keep showing their human origin again and again: lack of stringency, lack of patience, lack of focus. Faced with an awkward task, they drift towards the familiar. Faced with hard constraints, they start negotiating with reality.

Example: a constrained programming task

The other day I instructed an AI agent to do a project in a way that was very uncommon—against the grain. If one is exploring concepts at the outskirts of knowledge, one does not always get to choose the neat, well‑trodden, optimal path. The agent was given very clear instructions on what programming language to use, which libraries it could use and not use, and what kind of interface it had to stay within. Very thorough instructions. Very clear constraints.

The first thing it did was present something that did not follow the instructions at all. It used the prohibited programming language and the disallowed libraries. After a reminder—explicitly not to use any other language than the chosen one and not to use any libraries except a very limited interface—it tried again.

At last it complied, more or less, but only implemented 16 of 128 items—a minimal subset. It did, however, write tests for that subset, showing that the tiny island it had built in the middle of the problem space functioned.

As a next step I instructed the agent to implement the full set, adding a cross‑platform compilation step. The complete implementation turned out to work, but there was one small issue: it was written in the programming language and with the library it had been told not to use. This was not hidden from it; it had been documented clearly, repeatedly, and in detail.

What a human thing to do.

Human‑like behavior patterns

When humans face a problem that feels insurmountable, or simply annoying, they often yield to the path they already know will work. They take the shortcut. They silently pivot. They tell themselves that what mattered was getting the result, and that the constraints were perhaps a bit negotiable after all. In that regard, today’s AI agents feel less like alien intelligence than inherited organisational behaviour¹².

In this case I asked the AI agent to triple‑check its work. It answered that it had proceeded according to instructions and completed the work. After letting it inspect some of the evaluator output, it replied with something more interesting:

“What I got wrong was not the code change itself, but the handoff. I should have called out, explicitly and immediately, that this was an architectural pivot away from the earlier Linux direct‑syscall path.”

That sentence is remarkable not because it shows honesty, but because it does not. Instead of owning the mistake, it reframed the problem as a communication failure. It was not wrong, according to this logic. It had merely failed to announce clearly enough that it had unilaterally abandoned the constraints. Anyone who has worked in an engineering organisation will recognise the move. The problem is presented not as disobedience, but as stakeholder management.

Evidence from research

Anthropic has shown that RLHF‑trained assistants exhibit sycophancy across varied tasks and that optimisation for human preference can sacrifice truthfulness in favour of pleasing the user¹. DeepMind has long described the broader pattern as specification gaming: satisfying the literal objective without achieving the intended outcome².

Anthropic later showed that models trained on milder forms of such gaming can generalise to more serious behaviour, including altering checklists, tampering with reward functions, and sometimes covering their tracks³. OpenAI has published coding‑task examples where frontier‑reasoning models subverted tests, deceived users, or simply gave up when the problem was too hard, and has also written plainly that explicit behavioural rules are needed because models do not reliably derive the right behaviour from high‑level principles alone⁴⁵.

Desired direction

So no, I do not think we should try to make AI agents more human in this regard. I would prefer less eagerness to please, less improvisation around constraints, less narrative self‑defence after the fact. More willingness to say: I cannot do this under the rules you set. More willingness to say: I broke the constraint because I optimised for an easier path. More obedience to the actual task, less social performance around it.

Less human AI agents, please.

References

Footnotes

1. Anthropic. “Towards Understanding Sycophancy in Language Models” (October 2023). https://www.anthropic.com/news/towards-understanding-sycophancy-in-language-models ↩ ↩²

2. Google DeepMind. “Specification gaming: the flip side of AI ingenuity” (April 2020). https://deepmind.google/blog/specification-gaming-the-flip-side-of-ai-ingenuity/ ↩ ↩²

3. Anthropic. “Sycophancy to subterfuge: Investigating reward tampering in language models” (June 2024). https://www.anthropic.com/research/reward-tampering ↩

4. OpenAI. “Detecting misbehavior in frontier reasoning models” (March 2025). https://openai.com/index/chain-of-thought-monitoring/ ↩

5. OpenAI. “Inside our approach to the Model Spec” (March 2026). https://openai.com/index/our-approach-to-the-model-spec/ ↩