Mobile!
The first draft of UpTrust's mobile app is live on both the App Store and Google Play! We're still testing, and make sure it's the purple circles one, since there are two apps named UpTrust and one has nothing to do with us.
Trust scores
- Trust values now show how much evidence is behind them. A score built on one act reads very differently from the same score built on forty, so people lists and the Trusted People card on a profile now show an "act count" under the value. Think of this like review number on Google/Yelp. Better language for it than acts? Let us know and we'll update it.
- Voting the same tag on the same contribution as someone else now builds a small amount of per-tag trust between you. Agreeing counts twice as much as disagreeing, both are damped, and both stop accumulating after about six votes, so no pair can run up a score by always voting together. General trust is unchanged.
Reporting and safety
- Report and block are now inside a direct message thread, on the message itself. Before this you had to get to the sender's profile, and from inside a conversation there was no route to it.
Hey Wait
- All twelve launch conversations now have both a watch link and a read link on the series page. The transcript set is complete.
Experimental: "Asks"
- You can now bring in an introduction you wrote with an AI assistant. Run the interview prompt in Claude or ChatGPT, paste the result at /import, and your open questions come back as ask posts, with existing answers matched under each one and topics you can follow in one tap. We imagine a future where it's really easy to find a free couch, a job, or someone into the same hobby.
- The entry points are on the asks page, in the header and the empty state.
In testing (experimental)
Four larger pieces landed this week but are not on for everyone. They are built and running, switched off by default, and still changing week to week. Here is what is coming.
- @claude in group channels. A group admin can add an AI agent as a member, and members @-mention it in a channel to get a reply in the thread. Replies are grounded in the thread and in the group's own trust field, so when it points you toward a person, it is reporting what the members of that group think, weighted equally, and it says plainly when nobody there has a track record on the topic yet. Each group's admin turns it on, and replies are capped per group. We are starting with one pilot group.
- One inbox, ranked by trust. A single stream with everything that deserves your attention, ordered by trust, relevance, and recency, and sized to clear in five minutes. Replies and mentions, asks routed to you because you follow the topic, and the rest in one place. The team is using it first.
- Identity hub. A page that wires your other accounts into one identity, so trust you earned elsewhere counts here: GitHub, X, ORCID, and your own site. Claiming a domain works through a link back to your profile. On for a small set of accounts.
- Slack as a source of trust. Connecting a Slack workspace turns everyday activity there into topic-scoped trust signals, feeding the same engine as everything else. Being built against one workspace first.
- A bunch of MCP stuff
- A bunch of crazy awesome data science... more coming soon