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 small business owner at a kitchen-table desk orchestrating AI agents that clear a queue of small unfinished jobs and match customers to chefs

August 1, 2026 · 12 min read

AI for Small Business: What to Automate First

Most advice about AI for small business is a list of tools, which is the least useful part. What an owner needs is an order: which job goes to an agent first, which one stays human for now, and what breaks when you move too fast. Siddhi Mittal, co-founder of the private chef marketplace yhangry, gave a short talk at Y Combinator's Root Access about going all in on agents across engineering, product, and growth. Her sequence is copyable by any small business, and the part that broke is the most useful part of it.

A founder watching a user work around a missing feature by hand, then turning that workaround into a product

July 31, 2026 · 13 min read

Latent Demand: What It Is and How to Find It

Latent demand is usually taught as a market gap: a want the market has not met yet. That definition is true but not actionable, because it does not tell you where to look. Boris Cherny, who created Claude Code at Anthropic, has an operational version he calls the single biggest idea in product: people will only do a thing they already do, so find the thing they are doing by hand and make it easier. CLAUDE.md and plan mode both came out of that rule. Here is what latent demand is, how it differs from stated and effective demand, and how to run the rule on your own product.

A founder at a workbench studying one deep channel cut through a single industry workflow next to a row of shallow horizontal lanes

July 30, 2026 · 15 min read

Vertical SaaS: The Legora $100M ARR Playbook

Vertical SaaS used to be the conservative choice: smaller market, lower ceiling. AI inverted that. When the cost of building generic software collapses, the part that does not collapse is knowing exactly how one industry works and having permission to touch its data. Legora went from $1M to $100M in ARR in 18 months selling into law firms, one of the least fashionable buyers in software. Here is the business model underneath that run: why bundling beat the point solutions, what actually defends a vertical against a foundation model lab, why one workflow travels across countries, and where the model still breaks.

A founder at a workbench watching a simulated room on a monitor resolve into a real robotic arm picking up an object

July 29, 2026 · 13 min read

Physical AI: What It Means for Founders

Robotics spent a decade as the field where nothing generalized. Then it started copying the language-model playbook: pretrain on the world, fine-tune on actions, let reinforcement learning close the gap. Nvidia's Jim Fan calls it the great parallel, and it changes what a founder is actually looking at. Here is what physical AI means in practice, why the constraint moved from models to data, what breaks when software can act on atoms, and the vertical playbook a small team can run without owning a fab.

A single founder tracing a live multi-tier chip supply chain map on a workbench while spreadsheets and paper purchase orders fall away behind them

July 28, 2026 · 13 min read

AI for Supply Chain: The Founder Opportunity

A single advanced AI chip crosses a dozen countries and takes months to build, and the people responsible for it are working from spreadsheets, SAP, and phone calls. Y Combinator put out a call for founders to build supply chain 2.0 for chips, and the reason it is a startup rather than a feature is specific: the buyer's real pain lives in the second and third supplier tiers, where the incumbent system of record has no data at all. Here is what AI for supply chain actually has to do, why the bottleneck moved to packaging and memory while the tooling stayed still, and the test for whether your version is a company.

A single founder writing a specification and acceptance criteria while AI agents implement, test, and iterate on the code behind them

July 26, 2026 · 13 min read

Spec-Driven Development for Startups

In spec-driven development the specification, plus the tests or scenarios that define success, is what humans write. Agents write the implementation and keep iterating until the spec is satisfied. A YC partner calls the result an AI software factory, and StrongDM has published how its AI team runs one under two rules: code must not be written by humans, and code must not be reviewed by humans. Here is what that actually takes, what it changes about your team, and where it still breaks.