Methodology Declaration

Founder Application Suite

Sonnet 4.6 generation from your minimum inputs + your Founder Dossier + public data on the program. You always review, edit, and submit your own work — we never submit on your behalf.

1 / 5

Overview

The Founder Application Suite drafts complete applications to accelerators, contests, VC firms, grants, incubators, and government programs. Pricing is 99 credits ($40) per Single Application — one program, all questions, fully drafted.

Three steps. Pick a program from the directory. Provide your minimum inputs (or leave blank — the drafter pulls from your Founder Dossier). Get back a complete draft scoped to the program's known success patterns and your founder story. You review, edit anywhere it doesn't sound like you, and paste into the program's official form yourself.

What this product is NOT. It is not an automated submission system — we do not submit applications on your behalf. It is not a guarantee of acceptance — programs evaluate on private criteria we don't have visibility into. It is not a replacement for your own judgment about what story to tell about your company. It is a drafting assistant that does the typing-up so you can spend your time on the parts that matter — talking to customers, shipping product, and prepping for interviews.

2 / 5

How it works

Step 1 — Pick a program from the directory

Sprint 1 ships with 30 programs hand-curated from US accelerators, talent-first programs (Antler / EF), Japan-government tracks (JETRO / J-Startup), Plug and Play Tokyo, Silicon Catalyst Japan, EU programs (Endeavor / APX / Founders Factory), pitch contests (TC Disrupt / Forbes 30 Under 30 / SXSW / Web Summit / Slush). Each program carries a public success_pattern_summary describing the typical accepted profile based on public data.

Step 2 — Pick your founder + cofounders

The lead founder must already exist in your account as a person_profile (run them through People Finder + enrich-person first). Cofounders are optional. The drafter reads each person's career trajectory, employment history, education, skill signals, notable publications, podcast appearances, and personality signals from their stored Founder Dossier.

Step 3 — Provide minimum inputs

Per question, you can either (a) provide a direct input and the drafter uses it as the spine of the answer, or (b) leave it blank and the drafter composes from your founder dossier + public program data. Hybrid is fine — direct input on the questions that matter most, dossier-fill the rest.

Step 4 — Sonnet 4.6 drafts every question

For each question in the program's application_format, the drafter calls Sonnet 4.6 with: your founder data, the program's success_pattern_summary, your matched minimum_input (if any), and a strict ruleset. Banned: invented data, hyperbole, corporate-speak, name-dropping investors not in your dossier, fabricated traction. Required: specific evidence, numbers where your data has them, specific employers / projects / customers when those facts are present.

Step 5 — You review, edit, submit

The draft review screen shows every AI-generated answer alongside an editable textarea. You can revert to the AI version, save your edits, copy all Q&A pairs, mark the application submitted (capturing the final text), and record the outcome (accepted / rejected / interview / etc.) when you hear back.

3 / 5

Privacy posture

What we store

The minimum inputs you typed into the application form.
The AI-generated answers and any founder-edited overrides you persisted.
Pointers to the founder + cofounder person_profiles you selected (we don't copy their dossier data into the draft row — we re-read at generation time).
The submission status (draft / submitted / accepted / rejected / interview-invited / interview-completed) and any outcome notes you record.
The optional confirmation URL you enter when you mark the application submitted.

What we do NOT store or share

We do NOT share your draft with the program you're applying to. The program receives only what you paste into their official form yourself.
We do NOT use your draft to train models. Drafts are inference-time output; no fine-tuning corpus is built from your data.
We do NOT share your draft with other users by default. The is_shareable flag is off; the public_token field is reserved for a future "share with mentor" feature where you explicitly opt in per draft.
We do NOT contact the program on your behalf or send your draft anywhere — the only network call from the draft review page is to your own browser's clipboard when you click Copy.

Deletion

You can delete any draft at any time via the standard supabase client (or by emailing support). Deleted drafts are removed from the database; we retain no archive copy. The Founder Dossier the draft was generated from is governed by the separate person_profiles privacy posture (also user-controlled deletion).

4 / 5

Honest Limits

We don't know the program's private decision criteria. The success_pattern_summary attached to each program is synthesized from public data (program website, alumni interviews, accepted-batch announcements, public commentary). The actual scoring rubric inside the program — partner preferences, batch composition needs, internal politics — is not knowable from outside. Any prediction we surface about acceptance probability is probabilistic, not deterministic.
Sprint 1 uses a universal Q&A template across programs. The eight default questions (company name, one-sentence pitch, problem, solution, traction, market, team, ask) cover ~80% of accelerator + contest applications. Per-program customization (e.g., YC's exact 11-question form, Antler's talent-first questions) is a Sprint 2 follow-up.
We don't pattern-match against a comparable_founders_reference seed in Sprint 1. The "founders accepted to YC W25 with profiles like yours" pattern-match is a Sprint 3 follow-up that requires building the reference seed first.
We don't generate non-text application content. Video pitches, photo attachments, deck uploads — out of scope. The drafter generates text answers only.
Drafts are best-quality when your Founder Dossier is rich. Run the lead founder through People Finder + enrich-person first; thin dossiers produce thinner drafts. The drafter explicitly refuses to fabricate when the data isn't there — you'll see "[insufficient public evidence]" or a one-line honest fallback rather than invented traction.
Outcome data is captured but not yet used. Sprint 3 will use outcome signals (which drafts got accepted / rejected) to refine future drafts — Sprint 1 just stores them.

5 / 5

Corrections + Right of Reply

Three commitments modeled on the BBC editorial corrections process:

Identification window. Errors in a draft (factual misstatement about your founder pulled from the dossier, mischaracterization of a program's success pattern, etc.) flagged within thirty days of generation are corrected within five business days.
Re-publication, not silent edit. Corrections to your draft are persisted as founder_edits_applied entries — the original AI-generated version is preserved alongside, time-stamped, with the founder's edit recorded.
Program right of reply. Any program named in our directory may submit a correction to its success_pattern_summary or application_format via corrections@mentionfox.com. Verifiable corrections are applied within five business days.

Bottom line. The Founder Application Suite is a drafting assistant — fast, cheap, evidence-anchored. It is not a magic acceptance machine. Founders who pair this with rigorous dossier prep and direct minimum-inputs on the highest-stakes questions get the strongest drafts.

Methodology v1.0 · Published 2026-05-04 · Verifierce / MentionFox← Back to programs