Ben Crow — Full Stack Engineer

Full Stack Engineer at Culture Amp (Nov 2024–present), building features end-to-end in Next.js, TypeScript and Ruby on Rails at scale. I integrate LLMs into development workflows — from AI-powered tooling to shipping AI features in production. I'm focused on how agentic coding patterns change the way teams build software. A strong communicator, I work well with people from a wide range of professional backgrounds and perspectives.

Technologies

Frontend

Next.js
React
TypeScript
HTML / CSS
Tailwind CSS

Backend

Ruby on Rails
Node.js
Express
PostgreSQL
REST / GraphQL

AI & Agentic

MCP Servers
OpenAI API
LLM Integration
Harness Engineering
Context Engineering

Infrastructure

Docker
CI / CD
AWS / Cloud
Git
Testing (Vitest, RSpec)

Deployed tools and active work on agent runtimes, personal agents, and agent-native interfaces.

Reno Layouts

A browser-based floor-plan studio for renovation planning, forked from Alza. Draw or trace a plan in 2D, then raise it to 3D. An agent works the same live model through WebMCP tools, covering surveyed geometry, wall build-ups, fixture rough-in, tile set-out and construction-stage sheets. Runs client-side with no backend.

TypeScript React three.js WebMCP Agent Tools

GitHub Deployment

Weave

An experimental goal-directed runtime for software that can build the capabilities it needs and improve how it works. Weave combines bounded model judgment with explicit authority, replayable evidence, capability crystallisation through TDD, and measured optimisation.

TypeScript Agent Runtime Jev Evaluation Capability Crystallisation

GitHub

ScrolLess

A device-first feed aggregator where a personal agent uses your logged-in browser sessions to collect content, while the server stays out of the platform-integration layer and relays encrypted payloads to the PWA.

Preact / TypeScript Fastify Personal Agents MCP Web Push

GitHub Article

This Site's Agent

Ask about Ben chat panel with a sample conversation using a mocked on-device model

The portfolio doubles as an agent-interface experiment: an on-device browser model can answer questions about my work through explicit site functions, with those same capabilities exposed as WebMCP tools and connected to page-level interactions.

On-device LLM WebMCP Tool Use Web APIs

GitHub

Writing on agentic AI and the systems it's quietly reshaping.

Agents Get Better by Becoming More Human

May 2026 Agentic AI — Evaluation — Knowledge Work

As agents become more capable, we're turning proven human patterns for thinking and working into software.

AI Agents Make Management More Important, Not Less

May 2026 Agentic work — Management — Customer Zero

AI agents may make execution cheaper, but they make context, trust, autonomy, and development more important.

Headless Agents Make HR a Work-Architecture Problem

May 2026 Agentic work — HR Tech — People Science

Headless agents change how work is delegated, measured, and governed — and that makes HR a work-architecture problem.

Personal Agents Make Experience More Important Than Content Acquisition

April 2026 Agentic Content Apps — Part 1 of 3 — Start here

Personal agents can move the integration layer out of content apps, leaving apps to compete on experience instead.

The Hard Part of Letting Agents Gather Content

April 2026 Agentic Content Apps — Part 2 of 3

The pattern is technically visible, but access, trust, reliability, setup, and cost still stand between it and mainstream adoption.

Trustworthy Agentic Apps Need Device-First Architecture

April 2026 Agentic Content Apps — Part 3 of 3

A walkthrough of the ScrolLess architecture: the agent fetches, the relay stays blind, the device owns the data, and trust is structural.

Note on process: every piece is AI-drafted by Claude and ChatGPT. I direct, source, and edit — extending agentic patterns from code into writing.

View all writing →

Sydney-based developer — open to a chat!

My resumé can be found here.

I'm Ben Crow

Full Stack Engineer

About Me