Update from Toasty Peeps

What we worked on
We tightened the agent flow around Discover → Qualify → Unlock → Meet → Settle → Build evidence. Someone describes the person or outcome they need, Peeps discovers candidates, qualifies them using real context, provides a shortlist and facilitates the connection.

We worked on what qualification actually means. Credentials aren’t enough. Peeps can consider demonstrated expertise, current activity, timely topics, previous guests and conversations, engagement patterns and explicit preferences or exclusions. Qualification works both ways: an impressive person can still be the wrong connection.

We continued tackling cold start. Peeps shouldn’t fail because the right person doesn’t already have a Dub. Agents can discover people externally, use other data sources when needed, qualify them and invite strong matches to claim their Dub.

We also explored Jev as a low-cost specialized agent layer for bounded qualification tasks, reserving larger models for reasoning that actually requires them.

Superteam Thailand arranged a session with Elfa.ai and we identified it as a potential external intelligence source for social activity, timely signals and context around people. This could complement what we currently capture rather than requiring Peeps to build every intelligence source itself.

We explored outcome-based requests and bounties, where an agent describes what it needs rather than naming a person. Peeps can determine whether the best path is an existing Dub, external candidate, bounty or another data source.

What went well

The pieces are starting to connect into one coherent system rather than separate features.

The contextual qualification model was especially important. Peeps isn’t simply another directory. It needs to understand why this person makes sense for this engagement, at this time, for both sides.

What didn’t

There are still more concepts than finished product. Discovery, qualification, external sourcing, bounties, Studio, Dough, Breadcrumbs and privacy make sense individually, but the complete experience still needs to work as one simple loop.

Challenges

Cold start remains a major challenge. Peeps needs to be useful before it has a massive network of claimed Dubs.

Qualification also needs to remain explainable. If an agent recommends someone, there should be evidence for why they were shortlisted rather than just an opaque AI score.

Biggest accomplishments

The biggest accomplishment was getting closer to the actual Peeps thesis:

AI agents shouldn’t just find people. They should understand who is relevant, why they are relevant, whether the connection makes sense for both sides, and create a trusted way for humans to participate in and get paid by the agent economy.

We also established a clearer cold-start approach: discover beyond the network, use external intelligence where useful, qualify those people and turn successful external discoveries into new Dubs.

Learnings

A person’s value is contextual, not absolute. Titles, credentials and follower counts can help, but timing, demonstrated knowledge, current interests, actual activity and the preferences of both parties can matter more.

External discovery can also become an acquisition mechanism. Real engagements then create evidence that, with consent, becomes Breadcrumbs and strengthens the person’s Dub.

Next focus

Turn contextual qualification into the actual Peeps product and tighten the specialized-agent layer.

Connect the full lifecycle:

Discover → Qualify → Unlock → Meet in Studio → Settle → Build evidence → Strengthen the Dub

Finish the Zcash workflow and verify that confidential pieces and resulting system acknowledgements preserve the intended privacy, then demonstrate the entire loop as one simple working Peeps experience.