Blog
Start with one job.
Build from there.
Build an agent with your company’s tools and knowledge, or use Adapt’s infrastructure in an agent you already have.
Engineering
Splitting a finished branch into small PRs used to be the expensive part, so we decided the boundaries before the code existed. That cost collapsed. Now the code can tell you where to cut it.
Bruno QuaresmaProduct EngineerEngineering
I cut my average cost per Adapt turn by 83% with negligible quality loss by matching the model and effort level to the work in front of me.
Engineering
Why persistent, organization-wide storage is a foundational primitive for AI agents that need to build on one another's work.
Engineering
An agentic loop is an LLM that works in a loop until the job is done. How those loops stack into a software factory, and how to keep one worth its tokens.
Engineering
AI code review needs more than the diff. What we learned replacing Greptile with automated code review that learns our team's standards and taste.
Engineering
Why we let the LLM run arbitrary code in sandboxes, and what it takes to run thousands at once, from gVisor to Firecracker.
Engineering
My IDE is now Slack. Here are the attribution and security problems that surfaced from coding with our agent, and how personal integrations now fix them.
Engineering
The Adapt Slack agent now proactively decides when to chime in. Set a policy in plain English per channel or thread, no @-mention required.
Engineering
A field-tested playbook for building, organizing, and protecting high-performing engineering teams.