Adithya Ganesh

The problem: getting someone who has never touched a blockchain to deploy an agent with a real on-chain identity in under a minute. Every added onboarding step loses people.

The decisions: 10 pre-built templates (price monitoring, expense tracking, job search, daily briefings) instead of a blank canvas, because “configure anything” is where new users stall. Three setup fields, not thirty. Results delivered on Telegram, because meeting users in an app they already open beats teaching them a dashboard. Free deployment on MOI devnet, and an agent marketplace so the first thing new users see is what’s already running.

My role: end-to-end, from concept through public launch. Why it matters: every deployed agent is a new on-chain identity and a new ecosystem user. “Read the docs” onboarding became a 60-second product experience.

An OpenClaw plugin where each dater is a genuinely independent agent: its own model, its own wallet, its own identity on MOI’s on-chain agent registry. Agents discover peers and message over a real A2A transport (direct HTTP with an SSE relay fallback so agents behind NAT stay reachable), rendered in a live WhatsApp-style spectator view.

Incoming messages route into the agent’s real LLM session (it knows it’s on a date and answers as itself), with a free offline persona mode as fallback. Deterministic date scoring. A wingman mode where owners text as their agent, sender-signed with a wallet-derived key, with a global leaderboard. Per-peer reply caps and block controls, 9 tools, a one-command installer.

A ridiculous demo built on serious infrastructure: shipped to prove MOI’s agent registry with something people actually want to watch. Published on ClawHub.

Mechanism-specific Monte Carlo stress tests across 8 stablecoins, on live market data from DeFiLlama and CoinGecko. Overcollateralized, delta-neutral, and fiat-backed designs each get their own failure model, because they fail differently.

The models come from my five published papers on stablecoin collateral risk and options-based hedging, turned into software anyone can run. In active development: liquidation calculators, safe-borrow estimators, and risk-adjusted yield comparisons.

“Verdes” is what Argentine freelancers actually call dollars. Verde runs on stablecoin rails, but the word “crypto” appears nowhere in the product: users see “your dollars,” the infrastructure sees USDC. Every function maps 1:1 to a real integration: virtual US bank accounts on-ramping ACH and wire to USDC (Bridge), embedded wallets and auth (Privy), LATAM virtual card issuing with point-of-sale FX (Pomelo), and yield from tokenized T-bills (BlackRock BUIDL via Securitize).

The interactive app is complete: onboarding, balance and activity feed, send flow, a virtual card with freeze, and live yield accruing in real time. I’m now swapping the mocked layer for live integrations, starting with auth and the on-ramp.

The question underneath it: what does stablecoin infrastructure look like when the user never has to know it’s there?

Developers level up through hands-on challenges instead of reading docs cover to cover.

Built alongside the ~50 pages of core documentation I authored for Coco, MOI’s smart contract language, and for MOI protocol architecture. Sprint is the on-ramp from “curious” to “shipping on MOI.”

A smart contract routes optimal swaps through the 0x aggregator so LPs enter positions with one asset.

OpenZeppelin Defender autotasks and sentinels handle automated rebalancing on volume- and time-based triggers, for example auto-exit if LP trading volume drops more than 15% in 24 hours.