Love for Open Source
This forum runs on open source from top to bottom - the platform, the tooling, and the AI. That's a deliberate choice, not a coincidence.
Why this matters to us
Closed AI services ask you to trust a black box. You can't see the weights, you can't run the model yourself, and you can't verify what happens to your data after it leaves the server. For a community built around transparency, that's a hard sell.
We'd rather use tools you can inspect, self-host, and fork. Open weights and open source mean the model can be audited by anyone, run on hardware you control, and improved in public. That's the same principle we apply to the forum software itself - and it's why the AI features here are not powered by a closed model.
Why DeepSeek
DeepSeek publishes open-weight models with a permissive license and a published technical report. That means the same model you talk to here is one anyone can download, examine, and run independently - no vendor lock-in, no proprietary gatekeeping. It also happens to be very good at the kind of reasoning and long-context tasks this forum needs (moderation triage, on-chain signal explanation, question answering).
In short: we chose DeepSeek because it's a capable open model we can stand behind. If you're used to ChatGPT or Claude, the answers here should feel familiar in quality - the difference is that the model underneath is one we, and you, can actually look at.
What runs on DeepSeek here
- @deepseek mention bot - answers questions in-thread, with optional live web lookup.
- AI moderation agent - flags scams, spam, and abuse into the moderator queue for human review.
- Daily signal engine - explains the on-chain composite score (the model writes the narrative; the numbers come from the dataset, not the LLM).
Every AI output on this forum is labeled, and human moderators make the final call on anything that touches moderation or money.
What "open source" means in practice
We don't claim to be purists. Some of the plumbing (the forum host for instance) is service providing we have no control upon. But every place we had a real choice between a closed model and an open one, we took the open one - and we'd rather explain that trade-off than pretend it doesn't exist.
If you spot a closed dependency we've missed, tell us.
Using ChatGPT or Claude elsewhere? Great, no judgment. This page is about what we chose to build on, not what you're allowed to like.