Hello!
Community Engineer at Pulumi, host of the CoRecursive podcast, conference speaker, and writer of widely-read guides on AWK, JQ, and infrastructure tools.
I’m Adam. I’m a software engineer at Pulumi, I host the CoRecursive podcast, and I speak at conferences when I have something worth saying. Based in Peterborough, Canada, working remotely since 2011.
The best way to reach me is email — adam@corecursive.com. I read them all, though I’m sometimes slow to respond.
If you’d rather lurk, I send a newsletter every few weeks with what I’m writing, recording, and reading. Signup form is at the bottom of every page.
Elsewhere: GitHub · LinkedIn · Bluesky · Twitter · CoRecursive.
Upcoming
talk I Built an AI Running Coach (That Actually Remembers My Training) (upcoming)
An AI running coach on AWS that reads Strava, remembers training history, and keeps me on track — reliability lessons from a personal agent in production.
talk Let's Build an Infrastructure-as-Code Tool from Scratch (upcoming)
Build a minimal IaC tool from scratch to demystify how Pulumi / Terraform / CDK actually work.
talk When AI Agents Touch Real Infrastructure (upcoming)
War stories from giving AI agents real infrastructure access — what broke, and the patterns that keep them safe.
Recent Podcasts
podcast The Hutter Prize
I texted Don about a side gig: €5,000 for every 1% you shave off a zip file. It's a real contest. For 20 years, Marcus Hutter has offered €500,000 to anyone who can losslessly compress a gigabyte of Wikipedia small enough. Gzip gets you to 322. After that you're on your own. So we tried. Run-length encoding. Pointing back instead of repeating yourself. Huffman coding worked out by hand. Arithmetic coding, which spends less than one bit per character. Then Claude Shannon's 1951 guessing game, and a paper where an LLM beats every record and still wins nothing, because the model counts against you. Don wanted the euros. I had other reasons for bringing him over.
podcast The Bitter Lesson
I've been trying to understand how machine learning actually works. Not use it, but to understand it, down to the ifs and loops. How does a program built out of plain conditionals get better on its own? So late one night I sent Don a paper, reward is enough. The claim is that all of intelligence, the whole thing, comes down to a system maximizing a reward. So we backed up through the history to find out how far "reward is enough" really goes: B.F. Skinner training pigeons, a backgammon program that taught itself, the Go move no human would have played. It's a story about machine learning, and what that leaves for the rest of us who still do it by hand.
podcast The Pre-Training Wall and the Treadmill After It
I've been confusing Don with frontier-lab links late at night for months. Ilya Sutskever told a NeurIPS audience that pre-training as we know it would unquestionably end. There's only one internet, and the data isn't growing. The frontier labs call this the pre-training wall. A leaked Google memo from 2023 argued they had no moat. R1 is on GitHub. Llama is on Hugging Face. OpenAI's secondary-market valuation has climbed past $850 billion. Don kept asking what, exactly, that $850 billion was buying. So he came over and we made an episode about it.
Recent Writing
article Kubernetes Agent Sandbox: What It Is and How to Deploy It with Pulumi
Agent Sandbox gives AI agents kernel-isolated, disposable environments as Kubernetes resources. Here's what it is and how to deploy it on GKE with Pulumi.
( This is the CoRecursive / Adam Gordon Bell newsletter, and you probably signed up from the website, but then Adam stopped sending them for over a year, but now its back. ) It's July and I don't have new CoRecursive episode out this month ( details below) , but I do have a lot…
article Six Live Kubernetes Recommendations: AKS, Cilium, Rate Limiting, and More
Deploy an AKS cluster with Pulumi C#. Six Kubernetes recommendations from a live workshop, including the Docker Hub rate limit that hit mid-demo.
Recent Talks
talk I Built an AI Running Coach (And Gave It Memory)
Adapting the AI running coach to use vector memory (Pinecone) for semantic recall over training history.
talk Workshop: Deploying AI Agents on AWS With Pulumi and Amazon Bedrock AgentCore
Hands-on workshop deploying AI agents on AWS with Pulumi and Amazon Bedrock AgentCore.
talk Building Action-Taking AI Agents: Reliability Lessons from Real Systems
Building AI agents that take real actions, and the reliability patterns that keep them safe.
Recent Videos
video Terraform and HCL in Pulumi: State, Modules, and OpenTofu-Compatible Code
You don't have to throw out your Terraform to start using Pulumi. Three capabilities, demoed live: Pulumi Cloud hosting real Terraform state, publishing your Terraform modules so any Pulumi language can consume them, and running plain HCL on the Pulumi engine. Webinar with…
video Infrastructure as Code on AWS with TypeScript
In this hands-on workshop, learn how to use Pulumi to provision real AWS infrastructure with TypeScript. Adam walks through the building blocks — declarative resources, stacks, outputs — and ties them together by deploying a working multi-tier application: a database, a backend…
video How AI Agents Turned 5 Engineers Into 50
How does a small team ship custom code into customer cloud accounts in a day or two? Adam talks to Ewan Dawson, CTO of Compostable AI, about how they leaned into agentic AI for everything from product to infrastructure — and the rules they learned for building an "AI software…