Skip to work
← All work

SCHROLLAR

Academic discovery reframed as a feed, so the scrolling reflex produces a research trail instead of nothing.

2nd runner-up of 25+ teams at GDG HackFest: Axis, March 2026

March 2026 · 2nd runner-up, 25+ teams

runner-up
2ndrunner-upout of 25+ teams at GDG HackFest: Axis

The idea

Students already scroll for hours. Research requires the opposite posture: deliberate, structured, effortful. SCHROLLAR's argument was that the gap is interface, not discipline: that if academic material arrived in the format the scrolling reflex already accepts, the reflex could produce something.

The mechanism

A feed of research material where the relationships between claims are grounded by a DeBERTa-v3 natural language inference model, so what surfaces next is connected by entailment rather than by engagement. Passive scrolling converts into a structured research workflow: the trail you leave behind is the output.

The result

Second runner-up out of more than 25 teams at Google Developers Groups HackFest: Axis, March 2026. Built under hackathon time pressure. This is a competition build, not a product with users.

Screenshot

Schrollar's research feed: 15 schrolls over 141 papers, each entry grouping papers under one research question with its paper count and date.
Schrollar's research feed: 15 schrolls over 141 papers, each entry grouping papers under one research question with its paper count and date.

More work