Synthesis

Board Synthesis — 011 Reel Downloader

As specified, this ships a file identical to what snapinsta gives away free, into the one distribution channel the incumbents already own. The board didn't split on whether we can build it — it's built — but on whether any version wins. The answer that survived every persona: the binding constraint is not the resolver, it's distribution.

Consensus

• Ian's question — "what does a user get that snapinsta doesn't give free?" — has one honest answer on Paths 1 and 3 today: nothing. Same mp4, worse ads, no brand (claude, codex, cohere, gemini, glm, groq, openrouter, mistral).
• Zero moat, one channel. The tech is public yt-dlp; a copycat ships in a week. The only acquisition surface is Google search for "instagram downloader" — owned by DR70+ incumbents with 5+ years of backlinks — and paid ads for the category are banned. Saturated and shrinking.
• Grey economics: real for incumbents, brutal for a no-name. ~$0.30–$1.50 RPM needs millions of pageviews before proxies and continuous extractor repair (~14 open yt-dlp issues, breaks biweekly).
• Path 4 (creator lane) is dead — creators own their media. Near-unanimous.
• Sunk cost (32k LOC) must not bias the ranking.

Live disagreements

• Path 1 vs Path 2 is a genuine, even split. codex (all seats), claude-capital, gemini-distribution, glm, groq-redteam, cohere-redteam rank Path 1 or walk-away first — Path 2's install step is a conversion cliff and Chrome/AMO purge downloader extensions. Against them, mistral (all seats), cohere-capital, and claude/gemini/groq/openrouter-first-principles rank Path 2 first — it zeroes the binding constraint: no server request, nothing to rate-limit, no egress, nothing for Meta to attack.
• Does Path 2 violate Ian's "no login" rule? codex, groq, claude-redteam say yes. mistral et al. say no — the user is already logged in while browsing; the extension adds no step. This crux is unresolved.
• Path 3's value. claude-distribution alone calls it the only path that buys leverage (TikTok/YouTube = 5–10x search volume, less defended, yt-dlp needs no impersonation); openrouter and claude-redteam counter that cobalt.tools already does this free and clean.

Kill criteria

• Every top-10 SERP result is DR70+ and rankable tail volume across all target platforms is under ~500k/mo.
• Grey RPM < $0.50, or revenue < $3 per 1,000 real sessions.
• Path 1: download success < 80%, or 7-day repeat < 15%, or proxy/maintenance > 25% of gross.
• Path 2: landing→install < 15%, or > 40% of reels arrive as hard-to-stitch segmented streams.

Next actions

1. Buy one month of Ahrefs/SEMrush ($99). Pull search volume and referring-domain counts for the top rankers across Instagram, TikTok, YouTube, X, Reddit. This single number ranks every path.
2. $50 grey-ad smoke test to the already-built multi-platform page (Monetag/Adsterra sign up in days): measure real RPM and paste→download completion.
3. $0 Path-2 probe: walk 200 public reels in a real browser, logging progressive-mp4 vs segmented delivery; read current Chrome Web Store / AMO downloader policy for delisting precedent.

The way through

The biggest objection is distribution — the front door is a closed, incumbent-owned SEO war. The realistic route around it is Path 3's wedge, grounded in the keyword data: TikTok/YouTube/Reddit carry multiples of Instagram's search volume, are far less defended on tail terms, and yt-dlp resolves them at ~95% with no impersonation, no egress fight, and cleaner ad rails. Own the tail the Instagram-only incumbents ignore; make Instagram a client-side/extension feature bolted on, not the front door. The differentiator — breadth + reliability + no malvertising — is thin, but it's the only edge that isn't a straight copy. This lives or dies on Action 1: clear ~500k/mo rankable tail volume and there's a door; below it, there isn't.

If that kill criterion fires, the one pivot to test before walking to 005: redeploy the built engine as a Stripe-approved paid batch/API tool for social teams and agencies — clean rails, recurring revenue, no takedown magnet — rather than discarding 32k working lines.

Dissent

By: codex

• “Action 1 ranks every path” is false precision. Keyword volume and referring domains do not measure attainable CTR, SERP features, user intent, geographic mix, completion, repeat use, or monetization. A 500k/month threshold is arbitrary without a traffic-to-profit model.

• Path 3’s “way through” is asserted, not demonstrated. “Less defended,” “~95% reliability,” and “cleaner ad rails” need evidence. YouTube and TikTok are hardly uncontested, and broader platform coverage multiplies extractor failures, policy exposure, and support load. Downloads still incur egress; only a truly client-side implementation avoids it.

• Path 2 does not eliminate distribution. It exchanges SEO competition for extension-store discovery, installation friction, permissions distrust, browser-policy risk, and MV3/CORS/session-signature constraints. Whether it technically violates “no login” is semantic trivia; using an authenticated browser session creates a larger privacy and security burden.

• The tests are underpowered. Two hundred reels may miss account, geography, codec, rollout, age-gate, and A/B-delivery variation. A $50 ad test cannot establish stable RPM, especially across bot filtering, geography, fill rate, and multiple impressions per session.

• The kill metrics are internally muddled. Ad RPM and revenue per 1,000 sessions are not interchangeable. A 15% seven-day repeat rate may be unrealistic for an episodic utility, while 80% download success is likely already fatal. Neither threshold is tied to unit economics.

• “Creator lane is dead” is premature groupthink. Creators may pay for batch export, archival, highest-quality originals, caption/audio extraction, rights tracking, team workflows, and cross-platform repurposing—not merely downloading one owned reel.

• The API pivot is convenient salvage logic. Agencies do not automatically have download rights; Stripe approval is not assured; incumbents already sell APIs; and an SLA-backed batch product makes extractor maintenance more—not less—expensive.

• Survives: sunk cost should not influence path ranking.

Decision

PIVOT. Ads are dead by arithmetic - $8-10k/mo needs 5-33M pageviews. The 23-platform engine stays built and shelved for the Forge.

Seat health

Who answered and who did not. Kept at the bottom on purpose: it is diagnostics, not findings.

Seats present: claude, codex, mistral, gemini, groq, glm, openrouter, cohere, cloudflare No seats missing or failed.