Design
From wireframe to working product.
I start in wireframes, then move into working software quickly, so design ideas get tested with something real instead of a mockup. Pinboard.ai started that way in January.
AI with a purpose
AI that takes action, not just a chat box.
I build AI to do the work, not just talk about it. In Pinboard you tell the assistant what happened, and it moves the items and logs the time for you. Each change shows up as a card first, so you can accept or undo it.
Engineering
Working across the stack.
Pinboard’s interface is Next.js and React, with FastAPI, Postgres and pgvector underneath: 1,700+ API endpoints, and one AI assistant working through 250+ MCP tools. It also ships as desktop and mobile apps.
Shipping
A little better every day.
I ship in small, frequent steps and let what I learn shape the next one. That’s Pinboard’s real commit history this year, one square per day.
Learn and improve
Then I measure, and go again.
After each release I look at both sides: product analytics for what people actually do, like where they drop off and how often they accept the AI’s changes, and feedback for what they tell me. Together they decide the next wireframe.
The loop
Every release feeds the next one.
Design, AI with a purpose, build, ship, learn, then back to the start. No feature is finished after one pass; each one goes around as many times as it takes.
one board per client project
status drives the group bar
AI should do the work, not just chat
source = where the work came from
BOARD VIEW · WIREFRAME
InterfaceNext.js 16 · React 19 · Tailwind 4
Sean Block · Product Engineer
Iconnectproduct,designandcode.
I’m a product engineer and the lead technical product manager at Applied AI Technologies. Scroll to see how I work, with Pinboard.ai as the example.
About me
I’m Sean. I work across product, design and engineering.
01 · Design
Designing the product
I like to sketch the flows first and then design the screens, so the plan and the product stay close together.
02 · Engineering
Writing the code
I work in TypeScript, React and Next.js on the front end and Python, FastAPI and Postgres on the back end, with AI agents and MCP tools across both.
03 · Product leadership
Leading product and delivery
As lead technical product manager, I help set product strategy, lead a team of four and scope new work.
Product · ClerkXpress · Municipal licensing
Turning custom work into configuration.
ClerkXpress is Applied AI’s licensing platform for towns. I led the design and build of its JSON form engine, so a new license can be set up as a config file instead of new code. The form, fee rules, AI document scan and approval flow all render from it.
The day job
My day job: leading product.
I’m the lead technical product manager at Applied AI Technologies, where I work on SaaS for state governments. Being on both sides helps: knowing the code keeps my estimates realistic, and the business side keeps me focused on what clients need.
- 01
Leading a small, AI-assisted team
Applied AI is a six-person company. We moved development from a 15-person team with offshore developers to four direct reports working with AI-assisted workflows, which lowered costs and sped up delivery.
4direct reports · 6-person company - 02
Scoping and pricing new work
I wrote the proposals, RFIs, cost estimates and solution architectures for six new initiatives (four for state agencies and two for commercial products) and presented them to C-level stakeholders.
6initiatives scoped and pitched - 03
Pulling delivery into one system
I consolidated work that lived across notepads, Confluence, spreadsheets, Redmine and Jira into one place for time, tasks, tickets and notes, then set up an internship pipeline alongside it.
5→1tools consolidated - 04
Building for a state regulator ↗
The NYS Cannabis Map is the statewide licensed-dispensary locator for the New York Office of Cannabis Management, live at buylegal.cannabis.ny.gov. I was the senior technical product manager and frontend lead.
19MNew Yorkers it serves
In short
I try to start with the problem rather than the feature list, build the smallest version that proves the idea, and get it in front of real people early. I’ve found design, code and the business case work best when they’re treated as one decision.