Skip to content

FCC for Science Journalists

Welcome. This track is written for journalists who cover artificial intelligence, algorithmic accountability, public-sector technology, and the research frontier. The aim is not to turn you into a software engineer; it is to give you enough literacy in the FCC framework to (a) evaluate claims vendors or researchers make about FCC-based systems, (b) explain those systems to the public accurately, and (c) reach reproducible conclusions when reporting.

Why a journalist should care about FCC

AI systems are increasingly deployed in contexts you cover: municipal government, healthcare triage, insurance underwriting, energy-grid scheduling, environmental monitoring. Many of these deployments are being built on FCC or on one of its ten constellation-verticals. The framework is unusual in that it is designed to be audited: it produces structured evidence on every run, classifies personas by EU AI Act risk category, and emits compliance reports. That is a journalist's dream. You can ask for artifacts and expect real answers.

What this track covers

Page Focus Approx. time
Getting Started Orientation, no install required 20 min
Reproducibility Primer What reproducibility means in FCC 25 min
Explaining FCC to the Public Analogies, avoiding common traps 30 min
Citation Guide How to cite FCC and its artifacts reference
Interview Personas Question banks for sources reference
Further Reading Background sources reference

What you get from reading this track

  • A working mental model of the three-phase cycle.
  • A usable vocabulary — "persona," "R.I.S.C.E.A.R.," "model card," "evidence graph," "risk category" — that you can deploy in copy with confidence.
  • Confidence to request artifacts from vendors and researchers by name.
  • A template for the three or four paragraphs a responsible FCC-related news story should contain.

What this track does not cover

  • Wire-service speed. If you are on a two-hour deadline, read only Getting Started and Explaining FCC to the Public, then file.
  • Highly technical explainers. If you are writing a feature for a specialist magazine, this track is a floor, not a ceiling. Supplement with the FCC Guidebook.
  • Individual constellation-vertical details. Each of the ten verticals (Ophiuchus, Serpens, Libra, Crater, Scutum, Norma, Pyxis, Vela, Columba, Caelum) has its own repo and documentation.

Ground rules

Three things to internalize before you start writing:

  1. FCC is not a product. It is a framework (a Python package under MIT license) owned by INFORMATION COLLECTIVE, LLC. When a vendor says "we use FCC," you should ask how — which personas, what vocabulary, what provider.
  2. Most FCC deployments are plural. A system rarely uses FCC alone. It usually composes FCC with one or more constellation-verticals, a hosted LLM, a knowledge graph, and local data. Your story should name the pieces.
  3. Evidence exists. If a vendor claims "100% compliance," ask to see the ComplianceReport JSON. Its absence is itself newsworthy.

A one-paragraph explainer you can adapt

Here is a reusable paragraph for lead-length introductions:

"The system uses a framework called FCC (Find, Create, Critique), a structured workflow for multi-agent AI that requires every agent role to be defined in a ten-field specification called R.I.S.C.E.A.R. The specification includes explicit constraints and risk categories aligned to the EU AI Act. Because each role is documented and each run produces a machine-readable trace, deployments can be audited against regulatory requirements in ways that opaque single-model systems cannot be."

Adjust length to fit. Trim "structured workflow for multi-agent AI" for mainstream outlets; keep it for specialist ones.

Common traps

  • Calling FCC a model. It isn't. It is a framework around models.
  • Claiming FCC "thinks." It doesn't. It decomposes reasoning into documented steps.
  • Confusing personas with people. A persona is a YAML file, not a human employee; conversely, "the Critique persona" is not a psychological critique — it is a named role with documented halt authority.
  • Extrapolating vertical-specific claims to the whole framework. If Ophiuchus (finance) misbehaves, that is an Ophiuchus story, not an "FCC is flawed" story.

Getting artifacts

When you are reporting a story, ask for:

  • The FCC version (e.g., v1.7.0).
  • The scenario JSON(s) used in the deployment.
  • The persona YAMLs (or their model cards, which are more human-readable).
  • The ComplianceReport JSON if compliance is in play.
  • The evidence graph JSON-LD for any defensibility claim.

These artifacts are explicitly designed to be shareable. A vendor or researcher that refuses to share them is making a statement — include that refusal in your story.

Where next

If you have never heard of FCC before, start with Getting Started. If you are on deadline and need the public-explainer piece, go to Explaining FCC to the Public. If you are preparing for an interview, see Interview Personas.