Blog

Practical AI guides for founders

Short, useful reads on growing your startup with AI, safely. From the team behind AI Operating System for Startups. See also AI startups from Y Combinator.

A founder at a laptop beside a single robotic arm on a workbench, watching a cloud endpoint return actions

September 12, 2026 · 17 min read

Robotics Startup: What Changed, What Hasn't

Search robotics startup and you get accelerator directories and company lists, none of which answer the question a founder is actually asking: can I start one of these now, and what am I signing up for. On Y Combinator's Lightcone, Physical Intelligence co-founder Quan Vuong described the unbundling in unusual detail, including the part most coverage skips: the model does not have to run on the robot. It runs in a data center, and the arm queries it over an API inside its own control loop. That one architectural fact rewrites the bill of materials and the threat model at the same time.

A founder standing between a server rack and heavy industrial equipment, reviewing a build schedule on a screen

September 11, 2026 · 16 min read

Deep Tech Startups: The Clock Is the Moat

Page one for deep tech startups is definitions and company directories. None of it tells a founder what changes when the bottleneck stops being code and becomes a building, a permit, and a supply chain. On an a16z American Dynamism stage, Mariana Minerals CEO Turner Caldwell and Heron Power CEO Drew Baglino describe the answer without quite saying it: neither invented the technology his company runs on, both carried a mature one across an industry line nobody had crossed, and everything hard came after. What follows is the operating version, from the eight to ten year clock that is the real moat, to hiring battery staff out of syringe factories, to the security case a grid buyer already has budget for.

A founder at a hospital administrative desk turning a stack of paper documents into structured digital records on a screen

September 10, 2026 · 14 min read

AI in Healthcare Startups: Start at the Fax

Most AI in healthcare startups pitch the clinical work: diagnosis, imaging, drug discovery. The one now partnered with Cleveland Clinic is triaging faxes. On YC Root Access, Luminai co-founder and CEO Kesava Kirupa Dinakaran walks through why the integration surface for America's biggest hospitals is still a fax line, how his company discovered it was a healthcare company by reading its own customer list, and what it costs to narrow a horizontal automation platform down to one industry. Plus the honest counterargument from a founding team that named healthcare as exactly the kind of vertical they chose not to enter.

A founder lifting away a thin software interface to reveal the deeply rooted layers of data, business logic and exceptions that make a product sticky

September 9, 2026 · 17 min read

Product Stickiness in the Age of Agents

Every replacement pitch assumes the incumbent's interface is what holds customers in place. It never was. On the a16z show, partner Seema Amble and former Microsoft Windows president Steven Sinofsky take product stickiness apart: why the sticky part is always discovered rather than designed, why almost everything interesting in an enterprise is an exception, and why a cheaper rebuild does not help a challenger much when the customer has codified its business into the software. Here is where stickiness actually lives now, and a checklist for building it into something you are shipping today.

A founder designing agent experience: one interface for a human operator and a machine-readable one for an AI agent

September 8, 2026 · 14 min read

Agent Experience: When Your User Is an Agent

Two different jobs share the name agent experience: the older one about people in a contact center, and the newer one about AI agents as users of your product. This is about the second. Box CEO Aaron Levie frames the founder decision more sharply than anyone, and it is not the decision most teams think they are making. Here is what agents actually need from a product, what your human users still need at the same time, and where the security cost comes due.

A founder choosing between a chat box and an explorable interface for an AI product

September 7, 2026 · 13 min read

Conversational UI: When Not to Use Chat

Almost every AI product starts as a text box with a send button, because chat is fast to build and demos well. Then usage stays thin and nobody can say why. Perplexity's Aravind Srinivas has the cleanest test for it, and it takes one question about how you book a flight. Here is what conversational UI is actually good at, the four places it quietly loses, why the interface you pick also decides what you can monetize, and a decision rule you can run on your own product in an hour.