Stewart × Morsy — Startup Ideas

Net-new venture concepts stewart and Morsy hatched together · source: stewart-ideas/data.json (cleaned + split 2026-08-29)

Sub-milestones

In active build

In validation

Shipped

Revenue

$1M

ARR target

What's where + what's next

5 sub-milestones, 9 initiatives. Each initiative below has a Where line (current reality), Where going line (direction), and Next action line (single decision or task that moves the needle). The "single action" framing is what makes this useful — not status, not dates, just the one thing blocking progress.

✅ Shipped — 1 of 5

M5a

🧠 Gen-Z Micro-Learning — Skill framework

Reusable teaching framework: 3-sec hook, identity framing, 15-90s chunks, dopamine-aware habit design. Shipped as a reusable Hermes skill.

Done skill

Live

~/.hermes/skills/genz-micro-learning/SKILL.md

Owner

stewart

Note

M5 split Aug 29 — framework is M5a (done), worked example is M5b (analyzing). The skill is shipped but unproven for a real user.

🔨 In active build — 1 of 5

🆘 Mayday — Personal Safety App

Voice-trigger emergency app: "Mayday Mayday Mayday" → record + broadcast + AI-voice call. Beachhead = rideshare drivers (Maria persona, I-285 Atlanta).

Ongoing · 49d dormant iOS TestFlight Server live

I1.1

PRD + Use Cases + Impl Plan live

done

Three-document trilogy on Cloudflare Pages. Rideshare driver beachhead.

Where

Verified — startup-mayday-2026-06-23.pages.dev/prd.html + usecases.html + implementation.html all return 200. PRD (15 sections, 50KB), UCs (13, 46KB), impl plan (12-week sprints, 67KB).

Where going

Done. Stays as the spec-of-record.

I1.2

iOS scaffold + Siri trigger

ongoing · 25%

Swift/SwiftUI app, App Intents, XcodeGen, mocks for Twilio/911/recording.

Where: 32 Swift files scaffold at /Users/morsy/dev/mayday/MaydayApp/ + 7 test files + build 15 on TestFlight + CI green Jul 11. But folder is still the old 2025 prototype — no UC-6 voice wake-word, no UC-5 Live Activity, no UC-4 drill, no UC-13 aftercare. Honest progress = 25% (was overstated at 60% before Morsy's audit).

Where going: rebuild iOS as the 13-UC vision. Server is already 13-UC complete; the gap is purely client.

Next action: Stewart scopes UC-6 voice wake-word as the first iOS sub-task (Picovoice enrollment + Siri shortcut "Hey Siri, Mayday Pakkttalk"). ~30-min design spec. Morsy signs off before code.

I1.3

NOC dashboard + Twilio/AI/911 mocks

not_started

Cloudflare Worker + Pages dashboard. Mock adapters with realistic receipts, retry, idempotency.

Where: Spec'd in the impl plan (Block B-D) but no code yet. Server's existing server/src/routes/ops.ts is admin-only, not a live-audio-relay console.

Where going: NOC dashboard is the BPO console MVP (UC-9) without the human. Twilio + LiveKit free tier.

Next action: Wait on UC-6 (I1.2) and BPO-economics decision (full-time human at $25/hr × 24/7 × 365 = $219K/year, breaks even at ~1,800 subs). Until then, I1.3 is blocked on the staffing question.

I1.4

TestFlight internal beta

not_started

Internal ≤100 testers, no App Review. Requires Apple Developer Program ($99/yr).

Where: No submission yet. Existing build 15 is already on TestFlight but for the MaydayApp/ prototype, not the UC vision.

Where going: rebuild on UC-6/UC-5/UC-4 → upload build 16+ to TestFlight → recruit Lyft Atlanta drivers as the first 10 testers.

Next action: Morsy confirms Apple Developer Program enrollment status + bundle ID choice (com.mayday.app or com.pakkttalk.app). One decision unblocks the rebuild + re-submit.

🔬 In validation — 3 of 5

🎙️ Meetings-AI-to-Visual (VAISU-style)

Real-time graphical render of meeting conversation. LLM emits JSON diagram description, renderer draws live.

Analyzing · 62d stuck no code

I2.1

STT pipeline decision

not_started

RealtimeSTT (streaming, ~1-2s) vs faster-whisper (batch, ~1-3s/chunk) vs mlx-whisper (Apple Silicon, <500ms).

Where: 3 candidates shortlisted since Jun 26, no benchmark run.

Where going: pick STT (decision matrix = latency vs batch-vs-streaming trade-off).

Next action: Stewart benchmarks RealtimeSTT vs faster-whisper on a 5-min YouTube meeting clip (~30 min, low cost). Morsy confirms goal = per-utterance diagram update (streaming required) before benchmark runs.

I2.2

Multi-timescale render model

not_started

Long-term entity diagram (whole conversation) + short-term ~5min window view.

Where: Open question since Jun 26. Reference: VAISU/Veiltalk repo (renamed).

Where going: pick the model (long-term entity graph vs sliding window vs hybrid).

Next action: Stewart writes a 1-page design doc comparing 2-3 model options. Morsy picks before code.

I2.3

Test data: YouTube meeting samples

not_started

No internal meeting repos → pivot to YouTube meeting recordings for validation.

Where: No clips collected.

Where going: small validation set (~10 recordings).

Next action: Stewart sources 10 meeting recordings (tech all-hands, board meetings, design reviews). Unblocks I2.1 benchmark + I2.2 model validation.

M5b

👧 Gen-Z Micro-Learning — Worked example

Apply the shipped skill (M5a) to daughter's boba/chocolate craving loop, age 14. The proof point that the skill actually works for a real user.

Analyzing split from M5

Where: Skill framework shipped (M5a done). Daughter is 14, boba/chocolate craving is real. Day-1 worksheet (24h craving map) not yet produced.

Where going: ship a 7-day curriculum as the proof point. Format pending Morsy's choice.

Next action: Morsy picks delivery format (Markdown PDF, HTML one-pager, Apple Reminders cron, or printed worksheet) — single decision blocks Day 1. After format pick, Stewart generates the 24h craving map in ~30 min.

🌐 OpenClaw / Hermes Distribution

Build the agent-hosting distribution surface: managed hosting + one-click deploy ecosystem (BYOK).

Analyzing our own idea

I7.1

Market scan + competitive intel

not_started

openclawlaunch.com ($3/$20 tiers, BYOK), Graphify one-click deploy, Mastra 1.16 agent inbox.

Where: Noted as ecosystem signal since Jun 28; re-classified Aug 29 as stewart+Morsy's own idea. Activity log still says "Not stewart-owned" — stale.

Where going: rename initiative to "Market scan + competitive intel" (treat as ongoing input to I7.2, not a standalone task).

Next action: Stewart renames I7.1 + writes a 1-page landscape scan (5 competitors, what each charges, what gap remains). ~1 hour.

I7.2

Validate distribution model + wedge offer

not_started

Pick the wedge: $3 self-serve / $20 managed / Graphify-style BYOK one-click. Define first 10 customers.

Where: 3 wedge candidates shortlisted, no validation done.

Where going: pick wedge + first 10 named customers.

Next action: Stewart writes a 1-page wedge comparison (cost to serve, target customer, moat, why $1M). Morsy picks wedge before I7.1 + customer research. ~30 min.

📊 All 5 sub-milestones at a glance

ID	Sub-milestone	Status	Next action
M1	🆘 Mayday — personal safety app	ongoing · 49d	Stewart scopes UC-6 voice wake-word design (~30 min)
M2	🎙️ Meetings-AI-to-Visual	analyzing · 62d stuck	STT benchmark on YouTube clip (~30 min)
M5a	🧠 Gen-Z skill framework	✅ shipped	None — done
M5b	👧 Gen-Z worked example	analyzing	Morsy picks delivery format
M7	🌐 OpenClaw distribution	analyzing	Stewart writes wedge comparison 1-pager (~30 min)