Heavybit's DevToolsDigest: agentic engineering and agent infrastructure
Special Edition
Announcing Crucible, Heavybit's MCP Server
What if your agent could learn from a decade of Heavybit talks? That's the idea behind Crucible, our new MCP server, and here's why we built it. Add one endpoint to Claude, Cursor, Codex, or ChatGPT, sign in free, and put those lessons to work in your own agents. Get started→
Crucible's AI Products skills cover how to evaluate, ship, and govern systems whose output changes from run to run. The two newest, Agentic Engineering and Agent Infrastructure, draw on the conversations below.
01
6 Months of Agentic Engineering with Simon Willison
Datasette founder Simon Willison looks at how coding agents got good enough to take on real programming work. He argues that clear goals, the right tools, and context make the difference, and that steering agents well is a craft of its own.
02
The Limits of Lights-Out Coding with Dexter Horthy
HumanLayer founder Dexter Horthy examines why models handle bounded programming tasks well but struggle with the long-term effects of architectural decisions.
03
The System Is More Than the Code with Charity Majors
Honeycomb co-founder and CTO Charity Majors joins Third Loop to talk observability, AI, and the changing economics of software. A central thread: what happens when code becomes increasingly disposable.
04
Designing for Agents: Lessons from Sentry
Sentry co-founder David Cramer joins Brian Douglas to explore how design changes when agents continuously reshape a product. Agents draft specs, revise plans, and iterate through design patterns and frameworks.
05
The Mythical Agent-Month with Wes McKinney
Wes McKinney, creator of pandas and Apache Arrow, revisits The Mythical Man-Month for agent-driven development. He argues that coordination is part of the work, and that judgment and taste matter more than ever.
06
Why Agents Need Computers with David Crawshaw
David Crawshaw makes the case for virtual machines as the foundation for AI coding agents. He also covers how agent-driven development affects code review, security, deployment, and the economics of personal software.
07
Building Sandboxes for AI Agents with Ivan Burazin
Daytona co-founder and CEO Ivan Burazin explores the rise of sandbox environments for AI agents on The Kubelist Podcast. He explains why conventional infrastructure falls short for agents.
08
Long-Horizon Agents: Memory Is More than Recall
Heavybit's Amir Zohrenejad argues that agent memory is more than saving and retrieving past interactions. He frames memory as part of the harness, deciding what to write, how to retrieve it, and how it shapes future behavior.
Heavybit, 523 Octavia St, San Francisco, CA 94102, USA
