eYKON · Investor Deck
INVESTOR PRESENTATION · SEPTEMBER 2026 — eYKON.ai — The geopolitical intelligence layer that shows its record. — Live physical-sensor and event feeds on one interactive map, an AI analyst wired into our own derived datasets, and a public, hash-bound forecast record — used by people, and called by other people’s agents over MCP. — 26 analyst tools, callable over MCP — Listed on the official MCP Registry — Every forecast hashed at issue, scored in public — eykon.ai · Karim El Fekkak, founder · hello@eykon.ai · Figures verified against production on 2026-09-27
CONTENTS — The story in five parts — 01 — The opportunity — Why a professional will pay month after month — and gets tools built for their scenario — 02 — The platform — Five pillars, three engines — latest data feed integration : satellite imagery and cameras — 03 — The business and the competition — Who pays — and who else is here — 04 — Direction: an open platform — LLMs through MCP, and developers, as the distribution strategy — 05 — The ask — A — Annexes — Pillars, engines, stack, feeds and the 90-day roadmap — Claude_Presentation v2 — $500k to finance the GTM and the 90 day Roadmap
01 · THE OPPORTUNITY — The right instruments see it before the news does — Free OSINT — $0 + a day per question — Powerful raw feeds, unassembled — ten tabs, no correlation, no track record — eYKON — $29–99 / month — The assembled view, instruments we operate, and a published record — Enterprise — Kpler · Janes · Recorded Future — $50k+ / year — Assembled and deep, priced for institutions; the hit rate is never published — The world changes daily. A subscription pays for being told first — on evidence, with the limits stated. — Seen from orbit, before it was reported — Kuwait, July 2026: both
01 · THE OPPORTUNITY — DIRECTION — A trader’s thesis becomes a rated dashboard — 1 — 2 — 3 — The trader brings a thesis — A view specific to their book — not a generic screen, and more than an alert. — A partner developer builds the dashboard — On the 26 tools and MCP surface behind our own INTEL workspaces — no data plumbing. — eYKON rates the tool and its developer — A public evidence record per tool and version Evaluation matched to purpose: forecasting, detection, data quality Results shown with n and a baseline “Not yet evaluated” is an honest status Payment never buys a rating — Mediterr
02 · THE PLATFORM — Five pillars on one data layer — GLOBE · See it — Live feeds and registries; satellite imagery and cameras new today — INTEL · Analyse it — Nine workspaces — regime shifts, shadow fleet, calibration — AI ANALYST · Ask it — 26 tools run against our own datasets — COMMUNITY · Trust it — Scored forecasters, paid Spaces — BRIEFS · Read it — Daily brief, forecasts, convergence — Calibration — Every forecast hashed at issue, scored at resolution — Convergence — Evidence scored by independent source class, not volume — Admin tools — The platform monitors its own instruments and sa
02 · THE PLATFORM — NEW TODAY — Satellite imagery and public cameras, wired into every layer — 1,123 — watched sites — 27 — mines on Sentinel-2 — 5,067 — live public cameras — Imagery Layer, 16 PRs merged 2026-09-26/27. Cameras: 5 government operators, frames via eYKON’s proxy, operator credit on every frame. A cloudy pass is VOID, never zero.
03 · THE BUSINESS — The first customer: the professional individual trader — Crypto, commodity, macro and energy traders who trade geopolitics and buy their own tools — a self-serve decision in days, not months. Next: OSINT and risk professionals, won by the open record. — CITIZEN — Free — the live map — MEMBER — $12 / mo — $99 / yr — PRO — $29 / mo — founding rate for life, first 1,000 · $99 standard — ENTERPRISE — $99–199 — per seat, 3-seat minimum — Plus a $9 Week Pass and a $5 Query Pack. Creators are the channel: analysts run paid Spaces, keep 85% non-custodially and earn 25% of members
03 · THE COMPETITION — Three kinds of competitor, one axis none of them compete on — Argos Atlas and Aureum read 2026-09-26 from their public pages. Enterprise pricing indicative, public reporting.
04 · DIRECTION — eYKON opens up to LLMs and developers — LIVE — LLMs as a distribution channel — Claude, Cursor or any MCP client calls 26 live-data tools. Listed on the official MCP Registry; the catalogue is public before anyone pays. — LIVE — Why an agent picks eYKON — It can check our measured skill before trusting an answer — and every result carries its instrument, grounding and limits. — PROPOSED — Developers build on the same surface — Keys, caps and a call log exist today. Next: a developer tier, docs and SDK, and a marketplace where traders commission dashboards from partner develope
04 · DIRECTION — How an agent reaches eYKON — CLIENTS — Claude Desktop / Code Cursor Custom agents Developer apps — ACCESS LAYER — Public catalogue, no key API key per user Daily caps: 5 · 25 · 100 · 500 Per-call log — 26 TOOLS — Live feeds (vessels, aircraft, conflict, thermal, night-lights) Signals and calibration Imagery and cameras — INSTRUMENTS & DATASETS — NASA FIRMS, Black Marble AIS, Sentinel-2 Derived: convergence, regime shifts, the ledger — Every result returns with a provenance envelope — instrument, grounding and known limits — so the caveats travel into someone else’s context. —
04 · DIRECTION — Strengths beyond the technology — An accountability posture — Publishing your hit rate stops being attractive once you have a reputation to protect — incumbents structurally avoid it. — Owned instruments — NASA-direct sensing with VOID semantics: absence of a look is never scored as a result. — A cost base for the individual — Free and official sources at ~$10–25 a month in external data cost. — Capital efficiency — Built by a solo founder orchestrating AI coding agents: ~50 production services, 194 migrations, run by one person. — Founder–market fit — ~30 years in markets; co
05 · THE ASK — Raising US$500k to : — Finance go-to-market and the 90-day roadmap — GO TO MARKET — Fiat billing and the first real payment cycles — → Go-to-market — Is the edge defensible? — Paid AIS, radar admission, and a longer public record — → Roadmap — Can one founder scale it? — First senior engineer, community lead, developer relations — → Go-to-market and roadmap — The one metric that gates everything: recurring revenue through the payment path.
ANNEXES — Under the hood — A1 · GLOBE · A2 · INTEL · A3 · AI ANALYST · A4 · COMMUNITY · A5 · BRIEFS — A6 · Calibration · A7 · Convergence · A8 · Admin tools — A9 · Stack, feeds already integrated, and the 90-day roadmap — A10 · The 90-day marketing and sales roadmap
ANNEX 1 · PILLAR 1 — GLOBE — every feed, with its provenance — LIVE — GDELT conflict, ADS-B aircraft, NASA FIRMS thermal — LIVE — AIS vessels: 4 broad boxes + 6 chokepoints — DARK — Bab-el-Mandeb (71 days) and Hormuz (14 days) — stated, not hidden — LIVE — Night-lights, NASA Black Marble: ~9-day NASA lag — LIVE — New: Sentinel-2 looks on 27 mines · NASA GIBS · 5,067 public cameras — Thermal watches 353 of 554 refineries and 5,064 power-unit rows in 8 regional boxes. Outside them: NO DATA, never zero. — Screenshot captured 2026-09-01. Coverage: GET /api/coverage/watched and /start honesty board
ANNEX 2 · PILLAR 2 — INTEL — when a region’s normal has changed — LIVE — Regime Shifts: two-sample KS test, 30 vs 60 days, 6 theatres × 5 signals, p < 0.01 — LIVE — Shadow Fleet: silence scored against each vessel’s own cadence — LIVE — Calibration Ledger · Sanctions network (real OFAC data) — MODEL — Chokepoint, Cascade, Precursor library — labelled in-product — FIXTURE — Critical Minerals — on the roadmap to real data — IN BUILD — Reality Check: a weekly refutation board, the 10th workspace — (one of) Nine workspaces in lib/intel/modules.ts. Screenshot captured 2026-09-01.
ANNEX 3 · PILLAR 3 — AI ANALYST — who does what — The eYKON tools compute — 26 tools run SQL, PostGIS and statistical tests against eYKON’s own tables. Every figure is computed by code. — The LLM model interprets — It chooses the tools, reads the structured result, writes the analysis and names what is missing. It computes no number itself. — Routed by task — Haiku 4.5 for utility · Sonnet 5 for interactive analysis · Opus 4.8 for opt-in Deep Analysis. One config; a swap is one variable. — lib/analyst/model.ts. Screenshot captured 2026-09-01.
ANNEX 3 · PILLAR 3 — What we answer that a frontier LLM cannot — A better model reads the same news faster. It does not see more — answer quality is capped by instruments, not model tier.
ANNEX 4 · PILLAR 4 — COMMUNITY — reputation you cannot buy — Profiles, rooms, radar, a Brier-skill leaderboard, and paid Spaces in non-custodial USDC on Base — 15% platform fee, enforced on-chain. — The Reputation Note is never paywalled, gated or altered by tier, payment or partner status. — 20 Founding Partner seats ever; 1 taken. The commitment: earn a shown Note within six months, on the same rules as everyone. — State today: payment rails proven on-chain; no Space has been paid for yet. The empty leaderboard is by design — nobody has 10 resolved calls. — Screenshot captured 2026-09-01.
ANNEX 5 · PILLAR 5 — BRIEFS — what eYKON publishes — Today · Briefings · Forecasts · Convergence. A daily brief persisted at 06:00 UTC; a weekly “satellite readings that moved” item. — Every house forecast shows its SHA-256 and an in-browser verifier: MATCH or MISMATCH, computed by the reader, not by us. — IN BUILD — A weekly Reality Check issue with an opt-in email — Screenshot captured 2026-09-01.
ANNEX 6 · ENGINE 1 — CALIBRATION — the record, including today’s number — Issue — Hash (SHA-256) — Resolve — Score (Brier) — Publish, right or wrong — No observation, no verdict: a claim with no clear look is VOID — never a win, never a loss. — Three tracks never blend: machine (sensor observables) · house (our own calls) · creator (the Reputation Note). — Today: machine skill −0.13 on the cohort issued 2026-09-22 (n=833) — worse than always predicting the base rate. House and creators: no judged cohort with n≥10 yet. — Three families measured positive out-of-sample before issuance (+0.16, +0.
ANNEX 7 · ENGINE 2 — CONVERGENCE — independence, not volume — media — GDELT conflict + energy count as ONE class — sensor-ais — vessel transponders — sensor-firms — heat, NASA FIRMS — sensor-viirs-dnb — night light, NASA Black Marble — sensor-s1-sar — radar — only once admitted — Evidence in a spatial cell is scored by the number of distinct source classes, never by how many flags fired. — Single-source — all evidence shares one class — reported, not confirmed — Multi-source — two or more classes — Sensor-confirmed — two or more classes including a physical sensor — Wording rule in production:
ANNEX 8 · ENGINE 3 — ADMIN TOOLS — the platform watches itself — Calibration Ledger Monitor: 15 alert rules as code; a run record exists only if a tick ran. — Ingest health: shard liveness; a dead feed writes nothing rather than zeros. — Review queue: nothing auto-publishes — one-tap founder approval. — Growth & revenue: subscribers counted from money that actually arrived. — Shown: the public honesty board, generated from the database at page load. — eykon.ai/start, captured 2026-09-27.
ANNEX 9 · STACK — Tech stack — CORE PLATFORM — CONNECTED SERVICES — FRONT-END — Next.js 14 · React · deck.gl + MapLibre map — APPLICATION — ~50 Railway services · cron routes · 8 pg_cron jobs in the database — DATA — Supabase Postgres + PostGIS · row-level security · migrations through 194 — AI — Anthropic Claude — Haiku 4.5 · Sonnet 5 · Opus 4.8 · 26 tools · prompt caching — DISTRIBUTION — MCP endpoint · API keys and daily caps · official MCP Registry listing — PAYMENTS — NOWPayments crypto checkout · Unlock Protocol on Base (USDC) · Rewardful referrals — ANALYTICS — PostHog funnel analytics
ANNEX 9 · FEEDS — Feeds already integrated — /start honesty board and /api/coverage/watched, 2026-09-27; imagery: Consolidated Onboarding Brief rev X §24.
ANNEX 9 · ROADMAP — The next 90 days — IN BUILD — GATED BY MEASUREMENT — IN BUILD — Sentinel-1 radar study — Study set enabled 2026-09-29. Admission needs ≥10 clear passes per anchorage — weeks to months. Nothing shows until it passes. — IN BUILD — Reality Check board — Schema live; detector next. Then the 10th workspace, a weekly BRIEFS issue and a new analyst tool. — PROPOSED — DATES TO CONFIRM — PROPOSED — Developer programme — Developer tier, docs and SDK, revenue share on tools built on eYKON. — PROPOSED — Fiat billing — Card checkout — the main conversion constraint today. — PROPOSED — C
ANNEX 10 · GO-TO-MARKET — The 90-day marketing and sales roadmap — Proposed split of the first-quarter GTM budget; targets are goals, not forecasts. Sources: Strategy Update 2026-09-11; MARSILION.