Research social media through real browser operations

01 · Role in the system

What Jev does here

You give a research goal like find handmade art on Instagram. Jev chooses the platform, then chooses among a changing list of concrete read-only operations - search, open a specific profile or post discovered in earlier results, read comments, an explicitly requested media download, or finish - with each observed result fed back before the next decision. The division is strict: Jev picks from enumerated operations whose targets come from previous results; socai handles navigation, clicks, scrolling, and extraction, so Jev never chooses arbitrary DOM coordinates or generates shell commands. Runs end in cards, a table, and an evidence report, and every operation Jev chose is recorded in a history; the repository ships a real 64-second research report and recorded TikTok video evidence alongside its release-tagged, CI-tested code.

02 · Control boundary

Where Jev sits

Enumerated read-only operations with discovered targets to one choice per step inside the user's real Chrome; evidence and operation history persist into the final report.

Code owns the loop, permissions, thresholds, validation, and side effects. Jev owns only the bounded judgments described above.

03 · Known limits

What this evidence does not prove

• Drives your own logged-in browser sessions on your machine; platform terms of service and rate behavior remain your responsibility, and media download is gated on explicit request.
• Read-only by enumeration: posting, liking, or following are not among the operations Jev can choose.
• Research depth per run is bounded by the operation loop; the shipped example is one 64-second pass.

04 · Attribution

Public sources

This is a Community record: the project was published by a third-party community author.

• socai-io ↗Community · github · public · checked 2026-09-25
• socai-io ↗Community · demo · public · checked 2026-09-25