Home / Fellows / Case File 014
Sample case file · Red Horizon dispatch
One project, start to finish: the scenario a fellow wrote, the red-team that tried to break it, the countermeasure she built at the bench, and the disclosure trail that let us publish it. The hazard itself never appears. That's the point.
The brief was a single line on an index card: an outbreak with no origin. Nadia — a hospital microbiologist by day, a horror writer by the hours no one should be awake — took it and wrote a novella. Here is how it opens.
The novella never explains how the machine works. It doesn't have to — the dread lives in the microbiologist's dawning understanding, not in any instructions. That restraint is a fellowship rule, not a stylistic choice: the scenario names a failure mode, never a method. The threat class here is plain and already discussed openly in policy circles — cheap benchtop DNA synthesis that ships without meaningful screening.
A panel of scientists, a security researcher, and a bioethicist spent a day trying to prove the story couldn't happen. They couldn't. What they produced instead was a threat model that named the load-bearing assumption — the single hinge the whole plot rests on:
If the device will print anything, screening at the order desk is theater.
The red-team's job was not to make the threat more vivid. It was the opposite: to find the one place a defender could stand and make the story boring. They found it. If screening moves from the ordering website down to the device itself — the last point before synthesis — the garage printer stops being a plot device. The specifics of any bypass were documented once, in a sealed annex, and handed straight to the disclosure track. They are not here. They are not anywhere public.
No sequences. No synthesis routes. No device makes, models, or bypass steps. No "how the outbreak was built." Everything operationally hazardous from this project lives under managed disclosure with the relevant vendors and authorities. What follows is the defense — which is safe, useful, and yours.
Then Nadia put the novella down and picked up a soldering iron. Over a six-week sprint in a safety-reviewed community lab, capped by the 72-hour push at the heart of the fellowship, the countermeasure crew turned the red-team's one sentence into working defenses. Three of them, plus a brief.
github: emergence-institute/sequence-firewallOn the last day of the sprint the crew ran sequence-firewall live against a benign test harness — strings of harmless, publicly-known sequence standing in for anything real — and watched it refuse the order at the device. Boring, on purpose. The whole point was to make the scariest scene in the novella impossible to stage.
Before anything went public, the whole package went to the biosafety board and down the disclosure track:
That's the loop. A monster imagined so precisely that a defender could find its one weak joint — and then a fix built, demonstrated, and handed out, with the dangerous half burned on the way to press. Fiction with a footnote: solved, shipped, disclosed.
This is a sample // Case File 014, its fellow, and its artifacts are an illustrative composite built to show the fellowship format end to end. Repository names and figures are representative. The benchtop-synthesis screening gap is a real, openly-discussed area of biosecurity policy. Nothing on this page describes how to create, get, or misuse any biological agent.
If reading this made you want to write the scenario — and then go build the thing that kills it — that's the whole job description.