Page 3 / 3

The Observatory

Lesson · Foundations
Notice the Problem First, Then Let Agents Build It
Paul Graham's essay "How to Get Startup Ideas" argues that the best ideas are not invented at a whiteboard but noticed in the world — specifically by founders who experience a problem themselves, recognize it as real, and are positioned at the edge of what has just become possible. The translation for founders building AI-native companies today is sharper than it might appear: the arrival of language-directed software changes what is buildable in a weekend, but it changes nothing about how worthy problems are found.
Read the lesson
The Wire
Fundraising Is a Parallel Process — and So Is Every Agent Task You Delegate
How to Raise Money · Paul Graham
Why it matters
If you've handed real decisions to agents while simultaneously trying to close a round, you already know attention is the actual scarce resource — Graham's fundraising essay teaches you why that's a structural problem, not a willpower problem.
Read the dispatch
Lesson · Human Judgment
Fundraising Is a Mode, Not a Mood
Fundraising deserves a clean start, a clean exit, and no confusion about what it actually proves — lessons that matter twice as much when your company runs on agents and your credibility is built on trust rather than headcount.
Read the lesson
Lesson · Foundations
When the Handyman Business Runs on Agents
An AI-native handyman company is plausible only if you accept one hard constraint: the intelligence layer must be load-bearing, not decorative, which means the entire operation — scheduling, scoping, quoting, follow-up, quality review — has to be designed around the loop from the start.
Read the lesson
The Wire
Leaky Buckets Kill Agent Products Before the Demo Wears Off
Retention Is King · Andrew Chen
Why it matters
If your agents are impressive at onboarding and invisible by week three, you don't have a growth problem — you have a retention problem wearing a growth problem's coat.
Read the dispatch
Lesson · Distribution
Retention Is the Bucket Your Agents Fill
Acquisition buys you a second chance; retention decides whether that chance compounds or drains away. Andrew Chen's "Retention Is King" makes this concrete: a small persistent edge in keeping users alive on your curve is worth more than any burst of new signups, because the math of churn is unforgiving and the math of compounding is generous.
Read the lesson
Lesson · Distribution
Your First 20 Customer Conversations: A Working Template
When software is cheap to build, the scarce thing is genuine customer understanding — and the only way to earn it is to have the conversations yourself, before you delegate anything.
Read the lesson
The Wire
Stop Validating What You Hope Is True — Kill the Riskiest Bet First
Running Lean · Ash Maurya
Why it matters
If your agents are executing a flawed plan faster than any human team could, you are not winning — you are accelerating toward the wrong cliff. Ash Maurya's method tells you exactly how to find the cliff before you sprint.
Read the dispatch
Lesson · Building & Agents
De-Risk the Agent Before You Scale It
Ash Maurya's Running Lean is a discipline for finding what is broken in your plan before the market finds it for you. For founders building AI-native companies today, that discipline applies not just to the business model but to the agentic systems doing the work — and the two must be stress-tested together, in sequence, starting with whatever is most likely to kill you.
Read the lesson
Lesson · Foundations
Staffing an AI-Native Company: Who You Actually Need
In an AI-native company, the operating system is the loop — task, action, eval, review, learning — which means staffing decisions follow a different logic than in a traditional software org. You are not hiring to fill seats in a factory; you are hiring people who can own judgment, set the boundaries the loop runs inside, and continuously improve the system that does the first draft of almost everything.
Read the lesson
The Wire
The Messy Start Was Always the Feature, Not the Bug
Founders at Work · Jessica Livingston
Why it matters
If you're handing agents real work and watching them fumble toward something useful, you're not doing it wrong — you're doing it exactly as the founders in Livingston's book did it.
Read the dispatch
Lesson · Foundations
What Embarrassing Beginnings Teach the Agent-Era Founder
The founders Jessica Livingston interviewed for Founders at Work did not execute clean visions — they stumbled into their real products, survived on stubbornness, and made consequential decisions with almost no information. That pattern has not aged out. It has simply moved to a new layer of the stack, where the bottleneck is no longer whether you can build but whether you know what is worth building at all.
Read the lesson
The Wire
Term Sheets Haven't Changed. Your Leverage Equation Has.
Venture Deals · Brad Feld & Jason Mendelson
Why it matters
When your company's core labor is done by agents, investors will price the cap table differently — and you need to know exactly which clauses let them reach in and adjust the controls.
Read the dispatch
Lesson · Foundations
Designing the Core Loop for an AI‑Native Startup
AI‑native firms embed a self‑reinforcing loop of task definition, tool selection, context provision, action, evaluation, human review, and learning, making the loop itself the product’s operating system. By treating that loop as immutable infrastructure, founders can scale intelligence without retrofitting after the fact.
Read the lesson
Lesson · Building & Agents
Week‑One Bounded Autonomy in a New Venture
In the first week, founders should grant agents just enough freedom to act on clearly defined tasks while keeping human oversight tight. This balance lets early experiments move quickly yet remains anchored to the company’s core goals and trust structures.
Read the lesson
Lesson · Foundations
The AI-Native Company
AI-native companies are designed around agents, workflows, evals, human judgment, and continuous learning from the very beginning — not bolted on after.
Read the lesson
Lesson · Foundations
Software 3.0 & the Builder-Editor Founder
As software becomes language-directed, founders need taste, judgment, and orchestration as much as raw coding ability.
Read the lesson
Lesson · Building & Agents
Agents, Workflows, Evals & Harnesses
Reliable agentic work comes from clear tasks, good tools, context, permissions, tests, and feedback loops — not from vague autonomy.
Read the lesson
Lesson · Distribution
Distribution in the Age of Infinite Software
When software gets cheap to create, distribution, customer access, trust, and taste become the scarce, valuable things.
Read the lesson
Lesson · Human Judgment
Human Judgment in Autonomous Companies
Autonomous companies still need humans to set goals, make tradeoffs, handle trust, define boundaries, and own accountability.
Read the lesson

The Observatory

The observation log

Notice the Problem First, Then Let Agents Build It

Fundraising Is a Parallel Process — and So Is Every Agent Task You Delegate

Fundraising Is a Mode, Not a Mood

When the Handyman Business Runs on Agents

Leaky Buckets Kill Agent Products Before the Demo Wears Off

Retention Is the Bucket Your Agents Fill

Your First 20 Customer Conversations: A Working Template

Stop Validating What You Hope Is True — Kill the Riskiest Bet First

De-Risk the Agent Before You Scale It

Staffing an AI-Native Company: Who You Actually Need

The Messy Start Was Always the Feature, Not the Bug

What Embarrassing Beginnings Teach the Agent-Era Founder

Term Sheets Haven't Changed. Your Leverage Equation Has.

Designing the Core Loop for an AI‑Native Startup

Week‑One Bounded Autonomy in a New Venture

The AI-Native Company

Software 3.0 & the Builder-Editor Founder

Agents, Workflows, Evals & Harnesses

Distribution in the Age of Infinite Software

Human Judgment in Autonomous Companies