The thesis
“Debaters were drowning in evidence. So we built the lifeboat.”
Backfile is the AI-powered debate preparation assistant. A local-first desktop app that parses Verbatim docs, indexes every card you own, and ranks the best counter-evidence against any opponent case.
02 · The gap
Policy debaters carry thousands of cards. Every round, finding the right one beats writing a new one.
Before Backfile, that meant Ctrl-F across folders of Word docs, partner drafts, camp packets, and caselist dumps. 2AC prep got eaten by menial labor: file organization, card hunts, document juggling. Strategic thinking came last.
The friction never made anyone a better debater. It just consumed the prep time that could have. So we changed the math.
03 · The Parser
The engine. Everything else is downstream.
Debate documents are chaos. Hand-formatted Verbatim files, partner drafts in someone else’s style, camp packets with rogue heading conventions, caselists pulled from twenty different teams. No two are alike. The Parser reads all of them and reconstructs the same hierarchy out the other side.
Pockets, hats, blocks, tags, short cites, full cites, card bodies, analytics, highlight runs. Identified by structural rules, not pattern-matching guesses. Two-tier classification: Verbatim heading-style detection runs first; a content-based fallback handles non-Verbatim files. Zero AI calls. Deterministic. Reproducible across machines and runs.
Stacks couldn’t index without it. Scout couldn’t rank without it. The Parser is the part of Backfile that has to be right before any of the rest can exist.
- ▪Rule-based, deterministic
- ▪Style-agnostic across formats
- ▪Verbatim & non-Verbatim
- ▪Zero AI in the parse path
04 · The rest of the stack
What the parser unlocks.
Stacks
Search every card.
Every parsed card lands in a local-first SQLite knowledge base with FTS5 full-text search. Sub-100ms queries. Tiered relevance — tags and cites first, highlighted body text second. Never lose a card.
Scout
Get the counter.
Drop in an opponent's 1AC. Scout reads the positions present — DAs, kritiks, counterplans, T-shells, case args — and ranks your own cards as counter-evidence. Position-Level for breadth, Line-by-Line for depth.
Export
Ready for your speech.
Curated arguments drop straight into your workflow. Verbatim heading structure, highlight runs preserved, pocket and hat hierarchy intact. Your speech document, already half-written.
05 · The product
A round-ready interface, not a research toy.
Live UI at backfile.app. Currently Mac desktop, Windows in development. Hover the rotator to pause; click a dot to jump.
06 · Story
Built by a 2AC who got tired of finding cards instead of writing arguments.
Mo debated at McDonough High School in Maryland — a top-tier national-circuit program with a Hall-of-Fame coach. He ran 2AC. The job, in theory: think hard, pick the right argument, win the round.
The job, in practice: spend most of prep hunting through Word docs for cards he knew he had. The administrative overhead never made anyone a better debater. It just stole the time that could have.
Senior year, Mo transitioned into a coaching role and watched his Tournament-of- Champions-level debaters do the same thing. Same hours, same friction, same missed thinking. Backfile is the tool he wanted.
Local-first. Fast. Built around how policy debaters actually prep — not how AI startups think we should.
07 · Receipts
What’s shipped, what’s under the hood.
Pricing (live)
- Starter · $4.50/mo billed annually · regional, beginners, LD
- Pro · $13.50/mo billed annually · national-circuit policy
- Team · $18/seat/mo annually · coming soon · squads, coaches
- 7-day free trial. No card required.
Under the hood
- Local-first SQLite knowledge base with FTS5 full-text search
- Rule-based parser — deterministic, reproducible, zero AI calls
- Claude Haiku for Scout’s position analysis and ranking
- No training on user documents, ever
- Next.js 16 · Stripe · Supabase telemetry · Vercel