Build the machine that builds the machine

Bill Doerrfeld | December 19, 2025

A shift back to large end-to-end machine learning models

My latest DirectorPlus column with LeadDev interviews Bedrock Robotics' CTO, Kevin Peterson, on what it takes to develop highly adaptable and safe autonomous machines.


Lately, a big shift he's noticed is a move away from many small, purpose-built models toward one end-to-end system that computes its own best representation.


They're putting this to practice in the field — quite literally — developing models that allow excavators to operate autonomously in large-scale construction projects.


Peterson was formerly leading autonomous system development at Waymo and Caterpillar, so he has some really great experience to share, which other AI/ML or robotics experts will appreciate.


Read the post here, and if you're an engineering leader, definitely consider signing up for DirectorPlus for a monthly check-in with an experienced director, VP, or CTO on a new subject!

Read: Hard-won lessons on building and scaling ML models

Should AI agents scrape or integrate external data?

By Bill Doerrfeld • February 3, 2026
My latest feature for InfoWorld explores when it makes sense to scrape public web sources, and when official API integrations are the better choice for external data.

Going nano with software updates

By Bill Doerrfeld • January 30, 2026
What does it mean to go nano with your software updates — to "carve with a scalpel" instead of swinging a hammer? For my latest DirectorPlus piece, I caught up with Chainguard VP Dustin Kirkland to dig into that idea.

API standards: foundational for agentic AI

By Bill Doerrfeld • January 27, 2026
I recently moderated a webinar that brought together three luminaries in the API community to discuss the importance of API standards in agentic AI development.

Reviewing Port of Context (pctx)

By Bill Doerrfeld • January 26, 2026
The more folks use MCP servers in development, the more they’re realizing it can lead to runaway token usage, unpredictable response sizes, and flooded context windows.

Are all developers equally productive with AI? Not quite.

By Bill Doerrfeld • January 20, 2026
Who really benefits from AI coding tools? New research suggests AI amplifies existing top performers more than average developers. Read my post on LeadDev.

Edge AI becomes more important for inference

By Bill Doerrfeld • January 19, 2026
Many say edge computing will enable the future of AI inference. For InfoWorld, I looked at the tech required, and the roadblocks to overcome to get us there.

Developers are confident on the future of MCP

By Bill Doerrfeld • January 15, 2026
Survey data from Zuplo finds rising MCP adoption, security concerns, and shows how developers are using MCP servers to connect AI with APIs in 2026.

Lessons learned from 6 overhyped tech trends

By Bill Doerrfeld • January 5, 2026
Blockchain for everything, metaverse, big data, NFTs... In hindsight, what were we thinking? Today, I call out some of tech's biggest overhyped trends on InfoWorld.

What is agentic metadata?

By Bill Doerrfeld • January 5, 2026
Like any production software application, AI agents are producing a spectrum of metadata behind the scenes. Some are calling agentic metadata a “gold mine” to direct continual improvements.

What makes a great tech conference abstract?

By Bill Doerrfeld • December 17, 2025
I explore some tips to help speakers craft solid pitches. The Nordic APIs speaker selection committee looks for these sorts of things, but the tips could apply to any tech event.