Martin | Now

Overview

I am an AI engineer at WP, which makes an AI support engineer for B2B software teams. I have been there since May 2025, working across the whole stack.

The idea I keep working towards: a support team should not have to wait for engineers to get a new workflow, report or tool. They should be able to describe what they want and have it running the same afternoon, built by an AI that tests its own work. Most of what I build serves that.

These pages are a snapshot of work in progress as of October 2026. None of it is finished and it changes every week, so I will keep updating them. Screenshots are from our demo environment with made-up data.

  • Dynamic pipelines – automations anyone can build on a canvas, mixing plain code with AI steps.
  • Agents and tables – AI agents that use the whole platform as their toolbox and keep what they learn in real tables.
  • Chip – the assistant that builds all of the above from a chat. I started it on my own and it has become the centrepiece of the product.
  • Sandboxed coding agents – Claude Code and Codex as a step in any pipeline.
  • Sandboxed apps – real web apps that Chip writes and runs for a team, an internal Lovable inside the product.
  • Error handling and tests – pipelines that handle their own failures, and tests that let an AI prove its work before it goes live.
  • A full example – a daily report for every support owner, written by agents.
  • Policy labels – labels defined in plain English, dashboards on top, and a way to measure how good they are.

The pipeline editor: a Slack trigger feeding a Python node, a table write and an agent node, with the node library on the left

So far

  • 2025: policy labels, their dashboards and evals, and a research assistant over a customer's support history.
  • Early 2026: dynamic pipelines, from the first working editor in February to a chatbot that edits pipelines for you.
  • Mid 2026: records, table views, agent memories, and test layers.
  • Autumn 2026: Chip building whole workspaces, coding agents in sandboxes, and sandboxed apps.
  • Now: most of my time goes into Chip and sandboxed apps.

Tech: Python, Django, PydanticAI, PostgreSQL, Redis, E2B and Docker sandboxes, TypeScript, Next.js and React Flow.