Build intuition

Dilemmas

Interactive explainers for the tradeoffs behind Fide’s cybersecurity research. They are toy models, built to show the shape of a tradeoff, and each links to the research question it serves.

FID-076 · available

Contain or wait

An AI defender can isolate a suspicious host now or escalate to an analyst. Move the threshold and the approval delay to see containment traded against disruption.

FID-088 · planned

When a patch counts as verified

How additional independent checks change the false-acceptance rate of AI-generated repairs, and what they cost.

FID-087 · planned

One compromised teammate

How false evidence spreads through a team of defending agents, with and without source checks.

FID-075 · planned

The budget you test with

Why a fixed compute budget can understate what an agent can do, and how rankings move as it grows.