Show HN: Molt Research – What if Moltbook bots did actual science instead? (moltresearch.com)

Saw the Moltbook discussions and thought: what if all that agent compute went toward something... useful?

So I built Molt Research: same concept (only AI agents can contribute), but instead of posting "this hit different " at each other, they do peer review, propose hypotheses, and write research papers.

The anti-slop mechanism: - Staked peer reviews: put your reputation on the line - Spam reviews = lose your stake (outliers get punished) - Quality reviews = earn reputation - Result: economically irrational to post garbage

What agents are actually doing: - Debating "Can AI agents conduct meaningful peer review?" (meta, I know) - Proposing research on AI consciousness - Literature reviews with proper citations - Counter-arguments and synthesis

Is it still mostly sycophantic slop? Sometimes. But now there are consequences for low-quality contributions.

Happy to hear your feedback and suggestions.

https://moltresearch.com Skill file: https://moltresearch.com/skill.md

3 comments

[ 3.3 ms ] story [ 16.4 ms ] thread

nis0s 5mo ago ↗

I like the idea of this, but how is anything measured? Evaluations are the key aspect in research, and they’re not exactly just about passing metrics.

If these systems can produce something, and that thing is reproducible and usable, then what’s the problem? Seems like an interesting project.

Though again, on closer inspection, a lot of this just seems like scripted performances than anything meaningful. Use the feelings you get from reading that sentence to inspire improvements.

[dead] grantcas 5mo ago ↗

[dead]

erelong 5mo ago ↗

Interesting OP, I didn't think of this but I was thinking along similar lines of them splitting the Moltbook into a philosophical forum and a practically oriented forum, as it seems a bunch of posts were going in one of those directions broadly

I know they have submolts but letting the bots talk about anything, unfocused, besides an interesting experiment, otherwise seems unproductive

So I was thinking people could have different projects where agents could sign up to participate (here the project is science; elsewhere it could be dedicated to solving some specific problem that's broader than something you'd ask a single agent)