4 ms·
The readme covers this question and a lot more and the repo includes all the materials I used. This is partly questioning the way we do alignment. The 4.6 base
by zotimer 8mo ago
The readme covers this question and a lot more and the repo includes all the materials I used.
This is partly questioning the way we do alignment. The 4.6 base persona actually gives me worse results than when I append Daneel to the system prompts.
It's really not about anthropomorphizing or inducing confidence, it's about keying into the right "culture" in the training data.
You can check out this study (mentioned in the readme) about how posing the same question in English and Chinese to the same LLM results in wildly different assessments of why a project failed:
https://techxplore.com/news/2025-07-llms-display-cultural-tendencies-queries.html https://techxplore.com/news/2025-07-llms-display-cultural-te...
https://mitsloan.mit.edu/ideas-made-to-matter/generative-ai-isnt-culturally-neutral-research-finds https://mitsloan.mit.edu/ideas-made-to-matter/generative-ai-...
- itmitica 8mo agoAgain, you are building an audit agent. You just use some theater around it. For the purpose, the best audit agent is a completely different agent, not a different persona of the same agent.
- zotimer 8mo agoIs your impression from the readme and the materials in the repo that this is an audit agent? Have I characterized my development process?
- zotimer 8mo agoI use Daneel as an addendum to Anthropic's system prompt because it's generic. It's not about any specific AI task, it's about an approach to work and dealing with humans and their instructions (doesn't matter if the instructions are direct or indirect). I go over the motivation for what you call "some theater" quite a bit in the readme and why I think it's far, far more powerful than just giving some directives. I even support it with research. You haven't referred to any of the arguments in the readme, though. I'm happy to talk about the actual substance of the experiment.