Move over LLMs — it's time for LAMs
Bill Doerrfeld | June 10, 2025
My latest for The New Stack breaks down an emerging AI trend: large action models (LAMs)
Here's a new acronym for you — LAM.
AI researchers are calling the next class of models large action models (LAMs). These models are built to let agents perform increasingly sophisticated tasks — even navigating graphical user interfaces (GUIs).
Think LLMs with bells and whistles. LAMs are poised to make AI more actionable — and outmode RPA.
For The New Stack, I explored what LAMs are, what examples are emerging in the market, and what experts think.
The general consensus? LAMs are a turning point — the brain and eyes that could make AI agents truly actionable. Use cases are still a bit hypothetical, but becoming more real every day.
My latest for InfoWorld breaks down 10 MCP servers powering next-gen devops workflows, from GitHub and Atlassian to AWS and Pulumi.
Excited to share my talk from Nordic APIs' Platform Summit 2025. As the opening talk of the event, I wanted to address the elephant in the room head-on: MCP.
Watch all the great API talks, keynotes, and panel discussions from this year's Nordic APIs Platform Summit, now available on YouTube.
Block's open source initiative is helping to build brand reputation, attract talent, boost partnerships, guide internal open source best practices, and aid long-term system reliability.
Achieving ROI with AI requires a mix of strong leadership stewardship upfront, shifting the talent framework, and embedding proper monitoring and governance throughout the lifecycle.
I recently explored what goes into creating a solid AI agent knowledge base — from the types of data it should contain to the retrieval mechanisms and architecture patterns that support reliable agentic behavior.
I recently spoke with LinkedIn VP of Engineering Prashanthi Padmanabhan about the making of their Hiring Assistant, an agent recruiters are using to optimize the applicant selection process.
How do you fail at platform engineering? Make it 100% UI-first. Don't market it. Survey no one. Measure success by who's onboarded. And just copy what others are doing.
This year's Platform Summit felt reinvigorated, with many new motivations and areas to discuss. Of course, AI agents and MCP stole the show.
For my latest DirectorPlus edition, I interviewed Thomas Johnson, co-founder and CTO of Multiplayer, about lessons learned releasing their MCP server.
