Show Notes
Five days. Twelve-hour build sessions. One founder describing himself as delirious.
Jason Katz opens this one by announcing he has AI psychosis, and he is only half joking. Five straight days inside Claude Code, shipping, prototyping, handing off to his dev team, and coming back for more. It is the kind of week that either produces a business or produces a burnout, and the whole episode is really about the difference between those two outcomes.
Ryan is in the same trench from the opposite direction. He shipped Howdi to 36 people after roughly six months of building, and the thing that got him unstuck was not a new tool. It was a hypnotherapist. One session, mostly just saying the problem out loud to another human, and the week broke open.
What makes this one useful is the gap between those stories and the systems underneath them. Jason walks through why a vibe-coded prototype and a production-ready system are separated by a canyon most founders never see coming. He lays out how Kindling decides what to build and what to walk away from. Then he opens up Chief of Staff, an app he has in testing with executives, and describes something genuinely strange: you start with a blank slate, talk to it every morning for ten minutes, and 30 to 45 days later you have a complete operations plan you never sat down to write.
The back half turns into a conversation about avoidance. Not the failure kind. The ordinary kind, where you know exactly what the most important task is and you find four other things to do first. Jason's take is that AI killed one of the biggest reasons for it, because the scope of an unfamiliar project is no longer unknowable. You brain-dump the whole mess by voice and get a map back in minutes.
Then Ryan drops the first item from a list of twelve things he believes supercharge execution, and it is not a tool, an app, or a supplement. It is: build a family. Become a person a partner can thrive with. Learn to relish domestic life as an engine for getting things done. Jason, who says he avoided that exact thing for most of his life, backs it hard.
Two operators, mid-build, saying the quiet stuff out loud. This is what it do.
Named Frameworks
The Prototype-to-Production Canyon
Jason's core point on why vibe coding a working demo is nowhere near a shippable product.
- →The five things that live behind a front end nobody thinks about: permission control, access control, data management, data governance, data modeling.
- →His personal split: he prototypes because he holds the vision, the value hypothesis, and the user flows.
- →His dev team takes it to production, which he describes as significantly steeper and more complicated than most people realize.
Build Governance
Kindling's decision layer for what gets built.
- →The premise is that when you can build anything, the constraint moves from capability to selection, and building for the joy of building quietly produces products with no value in them.
- →Start from a problem statement, never from an idea.
- →Let the problems govern the roadmap instead of chasing one big idea.
- →Separate value in the process from value in the product, because a fun build can have plenty of the first and none of the second.
- →Stay industry agnostic early, then let categorization emerge from volume.
Portfolio of Solutions, Not a Portfolio of Software
Jason's reframe of what Kindling is actually accumulating.
- →As the problem count grows, patterns surface at two levels.
- →Macro: which industry does this problem belong to.
- →Operational: where in the demand fulfillment life cycle does it sit, whether that is customer acquisition, fulfillment, HR, or finance.
- →His phrasing for the strategy: peanut butter spread problem-solving across many solutions, and the big outcome arrives from the spread rather than from the one bet.
The Chief of Staff Pattern
The app Jason has in testing. Structurally it is two agents working a shared surface.
- →The chief of staff agent ingests Slack, email, Teams, meetings, and calendar, then distills what actually matters into a single morning brief.
- →The project manager agent carries a rubric for what constitutes an operational plan, so the founder never has to define one.
- →Interaction is conversational, by voice, roughly ten minutes a morning.
- →The agent schedules meetings, sends email, sets agendas, promotes clusters of tasks into projects, and flags anything unmapped or untouched for 30 to 40 days as a deletion candidate.
- →Starting from a blank slate, a full operations plan assembles itself over 30 to 45 days purely from the conversation, then moves as the work moves.
- →Jason's stated payoff: the anxiety of scanning a list of a million things and guessing at priority disappears, and no time is spent leaving the work to update a tool about the work.
The Scope Dump
Jason's antidote to avoidance-by-unknown-scope.
- →Historically he avoided projects whose size he could not see.
- →Now: open a chat and talk, do not type.
- →Brain-dump the entire situation in plain language, including what you know and what you do not.
- →Ask what you are in for. Read the scope report and start.
- →His framing: it is now easy to find out what you do not know, so the only thing left is actually putting in the time.
The Pressure Line
The diagnostic both of them land on.
- →Pressure is the fuel. Stress and overwhelm mean you have crossed the line and are pushing in the wrong direction.
- →Jason's tell is simple: if the work is depleting you rather than energizing you, the setting is too high.
- →His counterweights are jujitsu in the morning, the gym, near-daily sauna, and deliberately protected time with his wife.
Ryan's First Anti-Avoidance Move
One item from a list of twelve, framed as Ryan's personal conviction rather than a business tactic.
- →Create a family, be a person a partner can thrive alongside, and treat domestic life as a tool that removes avoidance rather than a drag on ambition.
- →His argument is that when you have a family, moral and cultural relativism stops feeling slippery and you know exactly what needs doing.
- →Jason's response is that purpose flips pressure into fuel, and every moment spent avoiding is purpose left unfulfilled.
Founder Experiment
Run the blank slate ops plan, 30 days. Do not build anything. Do not buy a project management tool. Open a voice conversation with your assistant of choice tomorrow morning and give it ten minutes. Week one: every morning, talk through what happened yesterday, what is stuck, what is looming. Do not organize it. Just dump. Week two: start asking it what it is seeing. Which things keep coming up. Which tasks have quietly become a project. Which ones you have mentioned four times and never moved. Week three: let it name your lanes of work for you. Do not name them yourself. See if its version of your operation matches the one in your head. Week four: ask it what you have been avoiding. Then ask it to scope the biggest avoided item in plain language. At day 30 you either have an operations plan you never sat down to write, or you have proof that your work does not run through your communications, which is its own useful finding. Total cost: five hours across a month, all of it talking.
Glossary
Vibe coding
Building software conversationally with an AI model, guided by intent and feel rather than by specification. Produces working prototypes fast, and rarely produces production-grade systems.
Prototype vs production
A prototype proves the idea works. Production means it works reliably for many users under real conditions with real data, and the distance between them is where most AI builds stall.
Access control
Rules determining which users can reach which parts of a system.
Permission control
The finer-grained layer of what an authorized user is allowed to do once inside.
Data governance
The policies covering how data is stored, retained, protected, and audited.
Data modeling
How information is structured and related so a system can query and trust it.
Build governance
Jason's term for the decision framework that determines whether a given idea should be built at all.
Demand fulfillment life cycle
The full path from acquiring a customer through delivering the thing they bought, plus the internal functions supporting it.
Operational plan
The living map of an organization's lanes of work, projects, milestones, and tasks. Notoriously undefined and inconsistent from one company to the next, which is why Jason's agent supplies the rubric.
Agent
Software with an AI model at the center that takes actions on your behalf rather than only answering questions.
Equivalency ratio
A uniform unit of measurement that lets you compare cost across different product form factors built from the same input.
COGS
Cost of goods sold. The direct cost of producing what you sell.
MRR
Monthly recurring revenue.
AI psychosis
Jason's own joking self-diagnosis after five consecutive twelve-hour build days. Not a clinical statement about anyone.
Lead magnet
A free asset offered in exchange for contact information.
Purple belt
An intermediate rank in Brazilian jiu jitsu, typically several years of consistent training.
Q&A: What Founders Ask After This Episode
Why do AI prototypes fail to reach production?
Because the front end is the easy part. Permission control, access control, data management, data governance, and data modeling all sit behind it and none of them show up in a demo. Jason Katz splits the work deliberately: he prototypes to prove the vision and the user flows, his dev team hardens it for production.
How should a founder decide what to automate first?
Start from problem statements rather than ideas. Kindling Solutions runs a build governance framework specifically because the ability to build anything makes selection the real constraint. A build can have enormous value in the process and none in the product.
Can AI actually build an operations plan for a company?
Jason's Chief of Staff app suggests yes, indirectly. Rather than asking a founder to write a plan, it ingests communications, briefs the founder each morning, and lets the plan assemble itself over 30 to 45 days of conversation. In his account it then updates as the work moves, without anyone leaving their work to maintain it.
How do I stop avoiding the most important task in my business?
Two moves come out of this episode. First, name the avoidance out loud to another person, which is what unstuck Ryan Estes after roughly six months on Howdi. Second, use AI to remove unknown scope, because much of founder avoidance is fear of a project whose size you cannot see. Talk the whole problem through by voice and ask what you are in for.
How many hours a day should a founder actually build?
Jason walks back his own 12-hour claim to a realistic six or seven, and calls that the healthy number when it sits alongside training, recovery, and protected time with his wife. His diagnostic is whether the work energizes or depletes.
What is the difference between healthy pressure and burnout?
Pressure is motivating and creates urgency. Stress and overwhelm mean you have pushed past the line. If you can feel the pressure as pressure, Jason argues you are probably already over it.
What does Kindling Solutions actually do?
It builds AI-run operations for businesses so output scales without adding headcount. The entry point is a 21-day Blueprint at $5K that maps how work moves and identifies the constraint, followed by either FullStak, a full custom AI team installed across the business, or In-House Agent, a single AI role solving one defined problem.
Who is Jason Katz?
Co-founder of Kindling Solutions in Littleton, Colorado, and the recurring co-host of the What It Do series on AI for Founders. His background runs through cannabis industry operations at CULTA, Story Cannabis, Harvest Health and Recreation, and EVOLAB.
