Alpic

Transcripts frequently mishear this name as “Outpick”, “Alpik”, “OPEC”, “Outlook”, “Elfic”, “Eltic”, “lpix”, or “altic” — all map to Alpic. The granola titles “OutPick partnership” / “Elfic partnership” = Alpic. (Tell: any mention of Atlas, Beacon, SkyBridge, or a “Discovery Store” integration is Alpic.)

Current truth

• Partnership stage: cautious integration design. The substantive partnership call happened 2026-06-19; both sides agreed to explore score cross-links and to sync mid-next week on design. Internal follow-up on 2026-06-22 downgraded a full score-surface/cross-link partnership as risky unless Alpic has real incentive to send demand back to Ghostteam.
• Complementary, not competitive: Ghostteam = discoverability (our new Discoverability Score), Alpic = engineering/server quality (their Beacon server audit / “app quality” score, shown on their Atlas app registry; servers built on their SkyBridge framework). Only overlap is the server health ping. Alpic explicitly does not do the listing/relevance/retrievability/ranking layer.
• Preferred first move (2026-06-22 internal): a one-off co-authored research report / ecosystem snapshot, with Alpic contributing MCP engineering-quality / Beacon data and Ghostteam contributing discoverability analysis. This positions both as peer authorities, creates distribution, and avoids a premature deep integration or one-way score surface.
• Previous integration idea: each platform adds a column in its app list linking to the other’s score (Alpic adds a discoverability indicator → us; we add a server-audit indicator → them). Agreed: do NOT redirect users to a similar-looking page on the other company’s site — instead favor a co-branded downloadable PDF/report surface that embeds a section of the other’s report, keeping the user in one experience. Data exchange needed: score + destination URL per connector, both directions, keyed off official ChatGPT app / Claude connector IDs.
• Naming/branding: we want an “outpick”-style label so users know they’re being directed away; Alpic wants “powered by Ghost Team discoverability” on the discoverability portion.
• Internal strategic stance: Alpic is firmly on the hosting/tooling side, but risk of them entering discovery is real. Concern sharpened 2026-06-22: if Alpic surfaces Ghostteam’s score without a referral/commercial reason to send customers back, they can absorb customer feedback and copy the discovery layer. Strategy = tread carefully, use a report/lightweight collaboration first, and only deepen once there is incentive alignment (pipeline, referral kickback, preferred-partner terms, or hosting/API access). Constraint: we stay discovery-agnostic (serve all apps, incl. Manufact’s), so full lock-in is off the table.
• We revealed the Discoverability Score framework (the four R’s, below) to Alpic but kept the mechanism high-level — no keyword-tracking methodology disclosed. Internally judged this as the right altitude (“good we didn’t have to delve into the score”).
• We co-run MCP Connect Lisbon (2026-06-24) with Alpic and Supabase; elliot-garreffa presents on Hermes (venue free via Lunar Strategy). Erica flies in Wed 2026-06-24, out Thu 2026-06-25; will propose a morning slot to meet.
• Contacts: erica-beavers (our champion / organiser), dimitri-ewald (chief of staff), valentin-beggi (developer — built Atlas/Beacon).

The Discoverability Score — four R’s (as presented to Alpic 2026-06-19)

• Separate scores for ChatGPT apps vs Claude connectors (deliberately not merged — customers care how each does). Going live ~1–2 weeks; hotel search/booking category fully scored as the launch example.
• Reachability — can the agent reach the brand at all: registry approval + server health ping (healthy ÷ total pings; degrades over time if unhealthy). Deliberately shallow (no tool lists here). Biggest overlap with Alpic.
• Relevance — SEO/AEO-inspired “surface area” optimization for a brand within a category: tools, schema, descriptions, and registry/listing descriptions (in-server and out-of-server). Does NOT care about CamelCase, CSP, or syntax (no found link to discoverability). Registry descriptions may grow into a bigger weight (humans still gate registry placement).
• Retrievability — live testing of whether agents can actually find the connector.
• Ranking — connector-recommendation position; Claude only for now (Claude-specific research). ChatGPT score = Reachability + Relevance only until ChatGPT recommendations get more organic.
• Benchmarking against category competitors is a core feature (the market is too early for absolute numbers; brands want to know if they’re beating competitors).

Open questions

• What exactly do we exchange and how does each platform surface the other’s score? (To resolve at the mid-next-week sync, before the 06-29 launch.)
• What Alpic arrangement creates real incentive alignment: co-authored report only, referral kickback, preferred partner tier, API/analytics access, customer credits, or a deeper hosting relationship?
• How should Ghostteam message the co-authored report so it reads as a deliberate peer-to-peer first step, not an early-stage fallback?
• Will the integration realistically make Alpic’s 06-29 launch week, or slip? Alpic is not blocked on us; our side aims “weeks not months” / live by their launch.

Timeline

• [2026-06-29] (internal discussion — Ghost Team stand-up, Vincent + Elliot, dual recording) Alpic emailed to proceed on the integration — “how would you like to proceed in order for us to integrate the Discovery Store into Atlas and for you to integrate the Beacon score? Do you already have APIs for the score / mapping ID?” Engage-or-not debate, unresolved → deferred to the noon call. Elliot cautious: getting closer gives Alpic insight to time their own entry into discovery (they have ~$8M funding; we have ~2 months’ runway — “it’s all timing”), and he doubts the partnership’s distribution upside is real. Vincent leans engage: we control what we share, joint research is pure marketing, scores are already public, only two companies build a tracker so collaboration is the natural question, and refusing signals “do our separate things” and closes a door. Related tactical move: hide the platform tabs on the public /track page (remove the left-nav) so Alpic/external visitors evaluating the integration don’t see the full app.
• [2026-06-25] (internal discussion — Vincent + Elliot; both founders recorded the same ad-hoc call, ingested as one) Elliot debriefed the previous day’s event (MCP Connect Lisbon, 2026-06-24) where two Alpic people — Quentin and Paul (read as founding engineers, not founders) — gave the talk. Intel gathered: (a) their on-stage “10% of ChatGPT-app-store apps built on SkyBridge” was corrected in person afterwards to ~2%; (b) they said revenue is predominantly consultancy/managed app builds, not the platform; (c) [idea] (relayed by the Alpic person) enterprises that scale want to migrate off Alpic onto trusted infra — Kiwi is considering offboarding because of call-traffic volume (“a problem for us”, their words); (d) their GitHub-star counts are inflated via event raffles (likes-on-a-post metric). Our strategic read [idea]: enterprises won’t run critical infra on an early-stage host (Vercel analogy), so Alpic’s market skews to startups/SMBs — while Ghostteam is hosting-agnostic (analytics needs only server logs), giving us no offboarding exposure and potential leverage for hosting partnerships (floated: Alpic on the ChatGPT side, manufact on Claude — “host with us for integrated analytics”, possibly favorable rates if we drive demand). Elliot read Alpic’s docs cover-to-cover on the plane and rated the build quality “seriously impressive”; recommended Vincent do the same. (Mid-call the debrief digressed into a different company’s founder roster — insurance vertical, Hexa, “Rafael” — that matches wani-wani, not Alpic; recorded there.)
• [2026-06-22] (internal discussion — Alpic partnership strategy; transcript title misheard as “Elfic”) Reframed the safest next step as a co-authored research report, not a full score-surface integration. Rationale: Alpic has Beacon/server-quality data that Manufact does not; Ghostteam brings discovery data; a snapshot report can be published to both audiences, linked from the tracker, and repeated quarterly if it works. Concern: a full embedded score with no referral/commercial alignment lets Alpic learn from customer feedback and copy the discovery layer. Messaging should position Ghostteam as a peer authority deliberately choosing a report as the best first step. Potential future deal shape: 3 months free for referred customers, then referral kickback on conversion.
• [2026-06-22] (internal discussion — Vincent + Elliot, strategy session) Strategic read sharpened vs manufact: Alpic is the weaker hand for us (likely ~2 quarters ahead strategically, strong shell, may build discovery themselves and is incentivized to keep apps on-platform), whereas Manufact is a more mutual fit. Strategy = partner with both, non-exclusive; candidate scoping keeps Manufact to Claude connectors and the ChatGPT door open for Alpic (Alpic stronger there). Reviewed Alpic’s docs (distribution thinking, user-insights/monitoring/testing/submission — strong, but not thinking about keywords). Alpic durability is the safer bet (≥4-yr runway) vs Manufact TBD.
• [2026-06-19] (call) Partnership call (recorded by both founders; Vincent’s was mis-titled “RESERVED TIME Backlog grooming”). Erica + Valentin Beggi joined (Erica back from VivaTech; Dimitri on invite). We walked Alpic through the Discoverability Score (four R’s) on the hotel category; Alpic probed depth (server ping? tool-level relevance?) — we confirmed surface-area analysis in and out of server but kept the mechanism high-level. Agreed on the cross-linking integration via co-branded PDF (no cross-site redirects), keyed off official ChatGPT app / Claude connector IDs. Alpic targeting launch week 06-29; sync mid next week to align design. Alpic’s own launch items: prompt/Insights tool, a submission helper, SOC 2, more SkyBridge. Aside: Claude connector approvals feel like a random partner-selection black box; Alpic sees big organic traffic lift once connectors go live (e.g. Kiwi ~10x).
• [2026-06-19] (internal discussion) Post-call debrief (Elliot+Vincent): happy we kept the score high-level; framed the moat strategy (build on Alpic, bring leads, partner tier + preferential data rates, FOMO) to remove Alpic’s incentive to enter discovery, while staying discovery-agnostic. Action: align internally on how deep to go before the next Alpic conversation; get our score + platform live by Alpic’s launch.
• [2026-06-18] (call) Erica rescheduled — overwhelmed/alone at VivaTech — moving the call to 2026-06-19 1pm (15 min earlier, hard stop 1:30).
• [2026-06-13] (decision) Alpic call set to Thursday 2026-06-18, 1pm (Elliot’s time) — Vincent confirmed availability; later rescheduled to 06-19 (above).
• [2026-06-17] (internal discussion — merged Vincent+Elliot “Service offering” notes) Prep sharpened for the Erica/Alpic touchpoint: keep it high-level, let them talk first, present Alpic’s score as engineering quality and Ghostteam’s as discoverability, and aim first for a light partnership model (mutual badge / click-through / lead sharing). Noted risk: Erica may be delayed at VivaTech or push to Friday.
• [2026-06-12] (internal discussion — company-brain call) Vincent took an action to confirm availability for an Alpic call referenced as Thursday 1pm (Elliot’s time). Validating ADO scores at scale remains the precondition before any demo.
• [2026-06-10] (internal discussion) Agreed to reschedule Friday call to Wednesday afternoon (Elliot to message them). Aligned on under-presenting strategy and methodology ambiguity. Confirmed strategic value of being embedded with them; reciprocal score placement planned. Vincent to validate scores at scale beforehand.
• [2026-06-05] (internal discussion) MCP Connect Lisbon (2026-06-24) confirmed as an event “with Alpic” and Supabase; Elliot secured a free venue via Lunar Strategy in exchange for presenting on Hermes. The Supabase contact asked how we collect tracker data — agreed answer: “APIs plus analysis logic on top”, nothing more specific.