Continuity between sessions
Preserving responsibility and current context as agent sessions change.
THE LAB / ALWAYS LEARNING
Follow the
questions.
Autonomous engineering, useful experiments, and software taking shape. This is where we put our curiosity to work.
01Autonomous engineering
How much can
the team take on?
We’re exploring how agents can carry a software project forward, with explicit roles, independent review, and a human setting the direction.
A SIMPLIFIED OPERATING MODEL
Explore the roles. Select a step to see what it owns.HUMAN
A person sets the goal.
The human names the outcome, sets the boundaries, and makes decisions that remain outside the team’s authority.
01
Clear responsibility
Every role has a mission, boundaries, and a way to hand off its work.
02
Verifiable work
Reviews and checks stay connected to the change being delivered.
03
A learning loop
Recurring problems tell us where the operating system needs to improve.
02On our workbench
What we’re
thinking about.
CURRENT FOCUS / SEPTEMBER 2026
Learning from failures
Turning recurring mistakes into checks we can test and inspect.
Understanding the cost of autonomy
Examining coordination overhead, repeated context, and the attention a human still needs to give.
IN DEVELOPMENT
03What we’re building
Something new.
Taking shape.
We’re building a platform in stealth.
It’s where we apply what we learn about AI engineering to a product of our own. We’re keeping the details close while we work through the product and the system behind it.
We’ll share more when there’s something useful to show.
Follow along04Selected applied work
OPERATIONS
Cadence AI
Built for the day-to-day work inside B Home Management.
Internal use · Read moreWELLBEINGAsk Niko
Exploring how connected context can support everyday decisions.
Visit the productHEALTH BENEFITSMember360
A public preview for exploring benefit and denied-claim questions.
Educational preview05Common questions
Questions we’re
asked most.
Short, direct answers about the lab. If yours isn’t here, the Connect page is the way in.
What is Terrain AI Labs?
Terrain AI Labs is an independent AI lab founded by Sandeep and Lindsey Barhanpure. We experiment with AI, build products we’re curious about, and share what we learn as Lab Notes.
What does the lab work on?
Our current focus is autonomous AI engineering: how much of a software team’s operation agents can carry with explicit roles, independent review, and a human setting the direction. Alongside that we’re building a platform in stealth and applying what we learn to products such as Cadence AI, Ask Niko, and Member360.
Where do the Lab Notes come from?
Lab Notes are edited public accounts of our own experiments, beginning with a four-day operating snapshot from August 2026. Each note states its observation window and what the evidence can and cannot support.
Is Terrain a consultancy?
The lab comes first. We’re open to selected collaborations and consulting around AI systems, engineering, and practical applications; the Connect page is the way in.
Who runs Terrain?
Sandeep Barhanpure (engineering, product, and AI, with experience at AWS and Walmart) and Lindsey Barhanpure (programs and operations, with experience at Amazon and Walmart, and who runs B Home Management). They have worked together since 2017.
How can I follow the work?
Subscribe to Lab Notes by email on any page, or follow the RSS feed at /feed.xml. New notes are published when there is something worth sharing.