2 months ago when I was recruiting for Frame 2.0 I realized how broken talent flow was in a field that would benefit from moving talent to important positions, fast. For instance, for Frame to get names of relevant profiles from another org's applicant pool (who have all consented to their profiles being shared with relevant AI Safety orgs) I have to first-

1. manually reach out to a point of contact at an org (like Bluedot).
2. Then I have to request them for their list (that they're incentivized to share as a talent-building org).
3. Then they will share their airtable.
4. Then I will manually search through their airtable.
5. Then I will export the right folks from their airtable to whatever my outreach platform is. 

This is way too long for 2026.

The process doesn't stop here. These applicants will have to then apply to Frame and answer some similar basic questions on Frame's application process (even though Frame does has a lot of different communicator specific questions).

Now, imagine the time it takes for a good applicant to flow through all these steps-

University EA group→BlueDot → ARENA → SPAR → MATS → junior researcher → senior researcher

I'm not saying that this is the path that every applicant takes. All I'm saying is that for a small field, that benefits from context of similar answers, and wants to analyze how effective things are- this isn't very efficient (I say this as someone who's filled 10+ applications myself). 

So, over a weekend (2 months ago), I vibe-coded a tool that I think should exist.

Refr

The goal behind Refr is to ensure talent moves fast and securely.

Here's how it would work-

If Juicebox, Common App, and an outbound CRM had a baby- it would be Refr.

1. The search would be agentic and fast
2. You can send friend requests to orgs that you can share talent pools with
3. Profiles will already be nourished by info from other application answers
4. You can hire directly from this platform
5. You can build automations for relevant folks by integrating this to your slack, email etc. 

Maybe? Maybe Not?

Now, this might be overkill and not work well in practice because talent management is hard. But as the Director of a fieldbuilding org I do think that if someone put some thought into this, it could be useful. Feel free to steal my github if someone wants to steal the code on this^ and apply to Surplus.dev  

What do you think is a more efficient/another way to solve this? Is this a problem your org faces?