Help

Reviewing FFTC:fwd Intersections.

This guide explains what the prototype does, how the generation workflow works, and how FFTC staff can help improve it by reviewing reports and leaving feedback.

Current stage

Testing, evaluation, and fine-tuning.

Intersections is a standalone FFTC:fwd module built for the Phase 1 launch window. It gives the Folio topic layer a way to show relationships, much like civic mapping shows relationships among organizations and systems.

The system is already useful, but it will become reliable through human-in-the-loop review from people close to the work. Staff feedback is the main improvement loop: generate reports, tell us where they feel strong or weak, then we tune the source layer, prompt, structure, and design.

The ask for FFTC staff

Generate reports, open them in admin, click Feedback, and leave clear notes. Content, tone, evidence, design, PDF layout, and functionality feedback are all fair game.

What the public tool does

  1. A visitor chooses two or three civic topics from the Folio set.
  2. They choose a purpose, such as learning, giving, or partnership.
  3. They may add optional context about what they are trying to understand.
  4. The system generates a polished civic analysis showing how the topics intersect.
  5. The report can be shared as a web page or opened as a compact print/PDF document.

What staff should do

  1. Start on the public generator and create reports with different topic combinations.
  2. Open the Reports tab and select the report you want to review.
  3. Read the report as a public visitor would see it.
  4. Click Feedback in the blue admin bar.
  5. Leave notes that are as specific as possible.

Phase 1 boundaries

The May launch version is built to get the core experience in front of staff quickly, then improve it through review. That matters because the Folio content is still developing and the best early signal will come from seeing real generated outputs.

  • This is not a directory of organizations. It is a relationship tool for understanding how community topics interact.
  • Phase 1 is intentionally practical: public topic selection, web report, share link, and compact print/PDF output.
  • Some Folios are still maturing, so early reports are not fully Folio-constrained.
  • There is no mandatory email capture or login in the public flow.
  • Future phases can add deeper FFTC-specific data, civic mapping relationships, themes, and more precise audience modes.

What feedback should cover

Feedback does not need to fit a narrow template. The goal is to surface anything that would make the report more accurate, useful, trusted, readable, or aligned with FFTC:fwd.

  • Is the civic analysis accurate and fair?
  • Does the tone sound like FFTC:fwd: clear, public-facing, useful, and careful?
  • Are any claims unsupported, overstated, too generic, or locally wrong?
  • Are important systems, organizations, terms, or caveats missing?
  • Does the report reveal useful non-obvious relationships, rather than just summarizing each topic separately?
  • Does the intersection map help, or does it create confusion?
  • Does the web report, print view, or PDF layout break anywhere?
  • Is the workflow understandable for a public visitor?

Best feedback includes the specific sentence, section, topic combination, or visual element that needs attention, plus the correction or concern if you know it.

What to test first

Good evaluation should include both normal public use and edge cases. We need to know which reports feel ready, which ones feel generic, and which topic combinations expose gaps in the current source layer.

  • Generate reports from strong Folios and from thinner Folios so we can see how the source layer behaves.
  • Try obvious combinations and strange combinations. The tool should be most useful when it reveals relationships a visitor may not have considered.
  • Add optional context in some runs and leave it blank in others.
  • Open the web report, the print view, and the PDF path when possible.
  • Leave feedback even when the report is good. Positive notes tell us what patterns to preserve.

How the synthesis works

Intersections is not a general chatbot. It is a structured generation workflow that uses selected topics, available source layers, and a constrained synthesis prompt to produce one designed civic brief.

1

Selected topics

The visitor chooses two or three civic topics from the FFTC:fwd topic set. Those topics define the report frame and determine which Folio and knowledge-base material is loaded.

2

FFTC:fwd Folio packets

Topic packets supply preferred definitions, local framing, language guidance, organizations, tools, articles, and source notes when that content exists. Folios are the preferred FFTC voice and local-grounding layer.

3

Curated knowledge base

Developer-maintained evidence entries add statistics, source-backed civic dynamics, caveats, and further-reading links, especially when a Folio is thin or incomplete.

4

Broad civic synthesis

For Phase 1, the model may use broad civic knowledge to connect concepts and explain systems-level dynamics. It should not use that broad knowledge to invent local facts, local organizations, URLs, or statistics.

5

Synthesis prompt

The prompt asks for a designed civic brief with one thesis, an intersection map, memorable pairwise relationship headings, a compact life-experience model, facts cited, next steps, and honest source notes.

6

Human review

Staff feedback is attached to the exact generated report so recurring content, tone, evidence, and design issues can be traced back to source material and prompt behavior.

What the prompt is trying to do

The synthesis prompt gives the model a fixed job and a fixed output shape. It asks for a title, subtitle, civic frame, intersection map, relationship sections, composite life-experience model, key insight, silo warning, civic meaning, next steps, facts cited, and source note. In admin report traces, reviewers can inspect the exact model, prompt version, system message, and JSON payload used for a report.

  • Write for a general public audience in the Charlotte and Carolinas region.
  • Produce a designed civic artifact, not a generic report.
  • Use Folios for FFTC voice and approved language when available.
  • Use curated knowledge-base entries for statistics, caveats, dynamics, and source URLs.
  • Use broad civic knowledge only for conceptual connections.
  • Explain pairwise relationships as real causal or practical interactions.
  • Avoid fabricated local facts, organizations, statistics, tools, or URLs.
  • Return structured report data that the web, print, and PDF views can render consistently.

What each admin tab is for

Reports
Inbox of generated briefs. Open a report to review the public view, inspect traces, and leave feedback.
Folios
Source-packet management. Status reflects content completeness: incomplete, draft, or complete. All topics remain selectable.
Help
This reference guide for staff review and the current development workflow.

How feedback improves the tool

Feedback becomes evaluation data. We use it to identify patterns: weak topic packets, missing evidence, confusing language, bad relationship labels, unsupported claims, tone drift, or layout problems.

Those patterns guide fixes to the Folio content, knowledge base, prompt, report structure, and visual design.