Selected work

2025 - present

Lucas

A message-native AI assistant that turns chats into done work.

Role
CTO · product design + engineering
Timeline
2025 - present
Primary interface
iMessage + WhatsApp
Users ask for help in the chat apps they already use and get answers back in the same thread.
Assistant reach
Tools connected
Lucas connects messaging to email, calendar, files, automations, and external tools.
Developer surface
API + webhooks
Scoped keys, idempotent message creation, and webhooks let other products trigger Lucas.
Shipping cadence
42 updates
Public changelog entries shipped from late 2025 through May 2026.
Lucas product visual

Overview

What the product needed to do.

Lucas meets people where they already ask for help: inside iMessage and WhatsApp. From there, it can reach email, calendar, files, tools, and automations without making users learn another app.

I built the product end to end, including the dashboard, admin tools, API, messaging bridges, agent runtime, memory layer, integrations, and production observability.

Main takeaway

The challenge was making an AI agent feel natural in a text thread while keeping the platform durable, inspectable, and ready for more tools.

Product surfaces

What users and operators got.

  • Messaging experience

    iMessage and WhatsApp became the primary interface, with every answer returning to the original thread.

  • Dashboard and admin

    Setup flows, integrations, automations, API keys, support tools, and health views for running the product.

  • Developer API

    Scoped keys, idempotent message creation, webhooks, and documented contracts for external developers.

  • Agent runtime

    Modal-hosted workspaces with access to files, shell commands, pages, artifacts, and longer tasks.

Systems built

What made it work.

  • Channel orchestration

    Incoming messages are deduped, batched, routed, executed, and answered through the right channel.

  • Agent workspaces

    Persistent state, progress streaming, cancellation, steering, and tool access sit behind each task.

  • Memory and automation

    Email, reminders, saved facts, preferences, and scheduled workflows give Lucas continuity over time.

  • Production visibility

    Axiom, PostHog, Sentry, health checks, backups, and log archives keep the system easy to inspect.

Technology

Stack, services, and integrations.

The stack and services behind the shipped product.

Product

  • React 19
  • TanStack Start
  • Tailwind CSS
  • Motion

Platform

  • Bun
  • Hono
  • Better Auth
  • Prisma
  • Postgres
  • Redis
  • Inngest

AI and messaging

  • Modal
  • Claude
  • OpenAI
  • MCP
  • iMessage
  • WhatsApp

Ops

  • Cloudflare
  • PostHog
  • Axiom
  • Sentry
  • Stripe
  • Resend

Role and remit

What I lead.

The product, platform, and operational areas I lead across strategy, shipping, and ongoing work.

  • Marketing site, authenticated dashboard, and internal admin console

  • Product API for auth, webhooks, developer endpoints, and product traffic

  • iMessage and WhatsApp bridges with media support, delivery state, and recovery paths

  • Inngest workflows for message intake, task execution, receipts, and failure handling

  • Modal-hosted agent runtime for tools, files, artifacts, and delegated work

  • Memory, email actions, integrations, billing, analytics, health checks, and logging