workspaces SELECTED WORK

Work that shows how I build.

Selected examples of the same core skill: take messy context, make the workflow visible, and turn it into software people and agents can use. Masthead is the current flagship product build.

Generated Masthead product visual showing Board, Logbook, and MCP retrieval surfaces.
Current flagship product

Masthead

Masthead is a local-first session memory product for AI-agent work. It turns local coding-agent history into a canonical session graph, searchable Logbook, live Board, and bounded read-only MCP context layer.

View Masthead Read build notes
  • check_circleLocal-first session database, Logbook/search, live Board, and read-only MCP retrieval.
  • check_circleTauri desktop shell, local daemon, SQLite session graph, and TypeScript/React surfaces.
  • check_circleProduct judgment around agent context, privacy boundaries, evidence, and reusable work history.
Pip product visual with app screen and character.
AI-native product case study

Pip

Pip is a consumer finance product built around a deliberately narrow idea: before you spend, check one number instead of guessing from a bank balance. The AI layer does not own the financial truth. It routes, explains, and takes setup actions through approved tools while deterministic application code calculates Spendable Cash Today.

View Pip Open app Read build notes
  • check_circleAgent-first product surface with typed actions and deterministic tool boundaries.
  • check_circleSupabase auth/data, Plaid connection flow, OpenAI Agents SDK, and Netlify deployment.
  • check_circleTrust-aware product design: read-only account data, no money movement, and no AI-owned financial math.
Executioner product screenshot
Product build

Executioner

A deployed personal execution-system prototype built around a blunt product belief: people often do not need more planning features, they need fewer decisions between intention and action. It turns messy brain-dump input into structured outcomes, scheduling concepts, and action-oriented review flows.

View landing page Open app
  • check_circleReact/TypeScript product architecture with Supabase-backed concepts.
  • check_circleOnboarding, access-key, billing, email, transcription, and AI-assisted planning workflows.
  • check_circleEvidence of product taste, end-to-end building, and practical execution.
Central Brain
Agent infrastructure

Nova OS / Central Brain

A private AI workflow infrastructure prototype for durable agent memory, project context, task state, and dashboard visibility. The core idea: all agents should know what is going on, what has already been decided, and what needs attention next.

Typed memory

Facts, decisions, fixes, incidents, project statuses, and follow-up tasks.

Agent context

A pattern for agents that need continuity instead of disposable sessions.

Dashboard thinking

Human-readable control surfaces for AI-native work.

Workflow board
Client systems work

Brander Group AI systems work

Ongoing contracted AI systems work in a real client environment: workflow infrastructure, agentic tooling, process support, and practical systems that help operational work move with less manual babysitting.

Workflow infrastructure

Systems around task state, context, and repeatable execution.

Agentic operations

AI-assisted workflows with human judgment preserved where needed.

Practical delivery

Turning ambiguous process needs into tools and workflows people can use.

Julian control panel repository screenshot
AI media workflow

Julian / avatar media control systems

A set of experiments and control surfaces around AI avatar video, generated media, review workflows, and API-driven content production. The useful part is not the novelty of generated video. It is the operational layer around prompts, variants, status, approvals, and repeatable output.

Media APIs

HeyGen, voice, video, and modern generated media workflows.

Review loops

Status, logging, approvals, and iteration instead of one-off generation.

Product thinking

Turning creative AI experiments into usable operator tools.

Rat Detective Online screenshot
Playable multiplayer game

Rat Detective Online

A browser game built with Three.js, TypeScript, socket.io, and physics-based interaction. It is useful proof of real-time, interactive, stateful product work beyond ordinary dashboards.

Play live View GitHub
Workflow handoff tool screenshot
Workflow product experiment

Workflow handoff tools

A small suite of diagnostic tools around lead response, intake, routing, and handoff reliability. The point was not the niche itself. The point was turning a fuzzy workflow problem into simple, interactive product surfaces.

Open tools