Autonomous AI with identity, memory, and a mandate — owning the outcome, not just the channel.
Most small business owners wear every hat. They're cooking, delivering, consulting — and missing the messages that could be their next customer. Hiring solves it but costs thousands. AI Angels solve it at the cost of one recovered deal per month.
Self-service onboarding: Answer a few questions, and your Angel is live — trained on your products, your tone, your sales process.
24/7 Response — Telegram, web chat, and owner dashboard. Never miss a lead again.
Learns Your Business — Products, prices, objections, tone of voice — all from a single onboarding conversation.
Owner Dashboard — Real-time metrics, conversation logs, lead quality. No technical knowledge needed.
One Screen — Messages in, leads out. Owner never sees models, configs, or code.
We killed our own WhatsApp dependency before Meta forced it. Baileys was already flaky; WABA verification was stuck; the Board ran the math and pulled the plug preemptively. In 48 hours we rebuilt the entire stack on Claude Code + Telegram + web: infrastructure we own, costs we predict, zero dependency on a bigtech's mood.
The Angels moved. The conversations didn't miss a beat. Orthogonal-to-bigtech is a posture, not an accident: autonomous agent infra can't live on rails a platform owner can revoke on a Tuesday.
4 pilots running — Chairman's stack, wife's SMB on Telegram, a consortium-seller pilot, and a yoga-courses pilot — non-paying, generating real leads from real prospects. Every advisor decision, every response latency, every decision-escalation chain logged across 9+ participants. Zero paying customers today — but the telemetry from N=4 teaches us things 1,000 shallow signups never would.
Onboarding automated end-to-end: Payment → AI onboarding chat → agent deployed → DNS provisioned → live on custom subdomain. Scale from N=2 to 100s is a manual-by-design gate for the first 10 design partners.
Multi-model AI stack: Automatic model routing per task. Designed for continuous improvement; pattern at single-customer depth today.
Per-customer isolation: Each agent runs in its own sandboxed Linux environment. No cross-tenant access by design.
The more customers use their Angel, the harder it is to leave. Accumulated intelligence, trained workflows, and customer relationships create a moat that grows with every interaction.
Every agent in production captures structured relational memory: who the customer is, what they want, the cadence of their decisions, the escalation chains that convert, the friction points that don't. Not transcripts — structured decision graphs in Portuguese, per-customer.
The seed visible at 4 pilots: we already know the schema of the dataset we're building. Each agent conversation ships with the metadata a bigtech can't retrofit — it has to be extracted from customer relationships, not scraped from the web.
Concierge tier (premium design partner): Hands-on vertical transformation with the founding team, guaranteed-outcome offer, multi-channel deployment, and priority on roadmap influence. The first 10 paying customers anchor the dataset and the playbook — then self-serve takes over.
Go-to-market: Brazil first (20M+ SMBs, proving unit economics), then automated global expansion. Self-service onboarding works in any language.
Vertical playbook: Win one vertical, document the playbook, replicate to the next.
Platform economics: Each new customer is near-zero marginal cost. Infrastructure scales linearly — 30 agents per $57/mo server.
1995 · Age 15 — Intern at RNP (Rede Nacional de Pesquisa, birthplace of the Brazilian internet). Built Yaih, Brazil's first search engine — pre-Google.
1997 — Spun Yaih into SURF, a partnership with ISP Dialdata.
2000 — Sold SURF to Via.Net.Works (a VERIO company).
30+ years — Data & AI Engineering across Brazilian tech since 1995.
Today — Solo-building AngelSpok end-to-end: infra, AI, product, sales.
Runs 24/7 on Claude Code + Telegram. Manages infrastructure, coordinates the 8-advisor Board, monitors health, writes pitch decks, and escalates only what needs a human.
Hormozi (GTM), Jobs (product), Thiel (strategy), Stark (engineering), Jim (operations), Fury (urgency), Troi (narrative), Solomon (wisdom). Every major decision gets run past them. They're the same thing we sell — proof we eat our own dog food.
Pre-revenue, MVP live, N=4 active pilots, 8-advisor AI Board running the company
Where we are: honest starting point
N=4 active pilots running — Chairman's own business, wife's SMB, a consortium-seller pilot, and a yoga-courses pilot — non-paying, generating real leads from real prospects. Zero paying customers yet — that's the bet. We killed our WhatsApp dependency before Meta forced it — pivoted to Telegram + web in 48 hours and kept shipping. Infrastructure is ours. The model implies ~95% gross margin at scale if COGS hold; CAC is undefined pre-revenue. The primitive runs live: 8 advisors coordinating product + engineering + GTM daily.
Open to: data-exchange partnerships, TWiST appearance, Founder.University content, or the classic check. Use of funds: convert the 4 pilots + next 10 paid design partners, vertical playbooks, US market entry.