carson.zyli.cz 11 case files

Carson Zylicz: The Plans

Eleven case files. Ten businesses and the person accountable for them. Swipe for the plans. The thesis they came out of is underneath.

1 / 11
CASE #000
Carson Zylicz ZYLI Portfolio Manager Carson Zylicz Freedom fighter
screens all five, eventually
on file 10 positions
what this is the file covers all 10
why it is in the book I run this book. Ten names, no index, no committee, so every line below is a decision I have to defend in public and re-underwrite in private.

The Plan:

The Man With Ten Plans

The problemI own ten companies and nobody has asked me why.

  1. 1Invest in 10 businesses, one at a time
  2. 2Write a plan for each of them, in pen
  3. 3Spend the research budget manufacturing proof the plans worked, then file the proof as the method

The objection“You made that up.”

The answer“I made it true.”

Why it worksA thesis you have written down is a thesis you can be wrong about, in public, on paper, forever. The anecdote is the audit trail, and the pen does not have an undo.

Estimated startup cost$350,000, mostly the anecdote

Accountable, in writing, on paper.

CASE #001
Tony Xu DASH DoorDash Tony Xu Operational Grinder
screens hit
last price $193.62 +0.96%
market cap $83.9B 5Y return -0.6%
why it is in the book Local density is the moat: the more couriers and merchants on a block, the faster and cheaper delivery gets, and the worse the second entrant's unit economics look.

The Plan:

The Burrito Relay

The problemDelivery is expensive because a car has to move, and a car is the most expensive ingredient in a burrito.

  1. 1Hire every resident on the block to stand in their doorway with both arms out, wearing a company jacket
  2. 2Tape a routing number to every doorway and pass the burrito hand to hand like a bucket at a fire
  3. 3Tell the customer it is a car

The objection“You cannot put civilians in a delivery chain.”

The answer“They are not civilians. They are wearing the jacket.”

Why it worksThe more people on the block, the shorter each arm has to reach and the faster the burrito travels. It arrives at 61 degrees, and nobody has ever asked what moved it.

Estimated startup cost$340, almost all of it jackets

Nobody drives. Everybody delivers.

CASE #002
Mike Cannon-Brookes TEAM Atlassian Mike Cannon-Brookes Ecosystem Architect
screens hit
last price $193.98 -1.36%
market cap $49.1B 5Y return -57.7%
why it is in the book Jira and Confluence are the enterprise mainframe of work tracking: cheap per seat, brutally expensive to migrate, with a partner marketplace settled on top.

The Plan:

The Infinite Ticket

The problemNobody knows what work is happening, including the people doing the work.

  1. 1Turn every task into a ticket, including the task of making tickets
  2. 2Have every ticket create two more tickets, then hire an audience of two to watch them
  3. 3Register the company as a performance of work, so the tickets are a script and not a system

The objection“You cannot run a company as a play.”

The answer“We are not running one. We are performing one.”

Why it worksOnce the tickets are the script, migrating means rebuilding the whole performance from memory. The mess becomes the system of record, and nobody can leave without explaining Act One.

Estimated startup cost$2,100, mostly the audience

Cheap per seat. Priceless to escape.

CASE #003
Jack Dorsey XYZ Block Jack Dorsey Ecosystem Architect
screens hit
last price $76.54 +0.41%
market cap $46.0B 5Y return -69.9%
why it is in the book Cash App turned a peer-to-peer wedge into a consumer network, while Square's seller base gives the same company both sides of the payment flow.

The Plan:

The Dollar Handshake

The problemNobody wants a bank account until they owe a friend five dollars.

  1. 1Put a cash machine on the sidewalk that only pays out while two strangers touch the same dollar
  2. 2Photograph the handshake and ask both of them to confirm the other one is their friend
  3. 3Sell them a framed Certificate of Friendship, because a photograph alone will not open an account

The objection“This is just a photo booth.”

The answer“A photo booth does not open a bank account.”

Why it worksOne dollar produces two customers and one relationship, and every argument about who let go first becomes a support ticket you can charge for. Both sides of that counter now belong to the same company.

Estimated startup costOne dollar, reused, plus $80 for the camera

One dollar. Two customers.

CASE #004
Olivier Pomel DDOG Datadog Olivier Pomel Algorithmic Engine
screens hit
last price $278.24 +0.66%
market cap $99.9B 5Y return +66.6%
why it is in the book Observability is the system of record for production. Once every service reports into one pane of glass, replacement means re-instrumenting the entire stack.

The Plan:

The One Window

The problemEngineers cannot see what is broken until a customer phones to tell them.

  1. 1Put every server in one room behind one very large window
  2. 2Make the window load-bearing, then sell the engineers binoculars
  3. 3Sell them a second window to watch the first window, and hold the only permit that covers load-bearing glass

The objection“That window is structural.”

The answer“That is why it will never be removed.”

Why it worksOnce every machine reports into one pane of glass, taking out the glass means rewiring the building. Nobody has ever rewired a building to save money, and nobody is going to start with yours.

Estimated startup cost$4,700, almost all of it glass

One window. Load-bearing.

CASE #005
Brian Chesky ABNB Airbnb Brian Chesky Ruthless Aggregator
screens hit
last price $160.39 -2.24%
market cap $94.6B 5Y return -6.0%
why it is in the book A two-sided network where the supply is homes and the switching cost is reputation: nobody gets to re-earn a host's review history somewhere else.

The Plan:

The Reputation Hotel

The problemHotels own buildings, and buildings are expensive.

  1. 1Do not buy a single room
  2. 2Rent out strangers' bedrooms and let them keep the furniture
  3. 3Argue that a drawer of guest reviews is not a hotel, so the inspections do not apply

The objection“This is an unlicensed hotel.”

The answer“This is a drawer. The drawer is licensed.”

Why it worksThe reviews live in a locked drawer rather than in the rooms, so a host can leave but four hundred ratings cannot. The reputation is the hotel, and the drawer is what keeps it standing.

Estimated startup cost$610, mostly the drawer

No rooms. All reviews.

CASE #006
Matthew Prince NET Cloudflare Matthew Prince Ecosystem Architect
screens hit
last price $355.01 -1.25%
market cap $126.4B 5Y return +82.3%
why it is in the book Sits in front of a huge share of the internet, so it sees traffic nobody else sees, and Workers plus Zero Trust turn ripping it out into a re-architecture, not a vendor swap.

The Plan:

The Traffic Cop

The problemEvery website on earth is getting attacked and none of them can see it coming.

  1. 1Stand at the front door of the internet and offer free parking
  2. 2Ask every car for its license plate and a small non-refundable deposit
  3. 3Laminate your own badge and let the city assume it hired you

The objection“You are not a police officer.”

The answer“I am a traffic officer.”

Why it worksOnce you have seen every license plate, you know who is speeding before they speed, and the notebook is worth more than the road. No city can remove you without removing the road you are standing on.

Estimated startup cost$340, mostly the badge

Free parking. Permanent records.

CASE #007
Luis von Ahn DUOL Duolingo Luis von Ahn Algorithmic Engine
screens hit
last price $148.13 +0.70%
market cap $6.9B 5Y return -14.7%
why it is in the book A daily habit loop that manufactures its own training data: difficulty, content and retention all get tuned by the same flywheel that keeps users coming back.

The Plan:

The Owl That Never Sleeps

The problemPeople quit learning languages, and people handle guilt very poorly.

  1. 1Assign every user a personal owl and teach it which notification hurts that particular person most
  2. 2Put the owl outside the house in a costume, so the lesson feels attended
  3. 3Have every owl collect an apology, then hire a night shift to read them, because the apologies are the product

The objection“That is a man in a costume.”

The answer“He is a man in a costume who reports to me.”

Why it worksEvery apology is labelled training data, so the product gets sharper the guiltier the user feels. The night shift is not overhead, it is the factory, and it runs when the owls work best.

Estimated startup cost$9,400, mostly the owls

Five hundred million owls. All watching.

CASE #008
Dylan Field FIG Figma Dylan Field Ecosystem Architect
screens hit
last price $21.93 -0.36%
market cap $11.7B Since IPO return -68.8%
why it is in the book The design file is where the work actually lives. Multiplayer editing plus a shared component library means switching cost is measured in team-wide rework.

The Plan:

The Shared Cursor

The problemDesigners keep leaving, and when they leave they take the files with them.

  1. 1Put the entire company inside one file
  2. 2Require two people to be present before anything can be changed
  3. 3Strap a live activity monitor to one of them and publish the readout, so the client can see the work is real

The objection“That is surveillance.”

The answer“It is a receipt.”

Why it worksWhen the file is the workspace and the work is witnessed, leaving means redoing everything in front of an audience. The monitor turns collaboration into evidence, and evidence is the switching cost.

Estimated startup cost$15,000 a month, all of it witnesses

One file to rule the design team.

CASE #009
Vlad Tenev HOOD Robinhood Vlad Tenev Ruthless Aggregator
screens hit
last price $112.00 -1.85%
market cap $100.7B 5Y return +220%
why it is in the book Distribution first. Zero-commission trading pulled a generation onto one app, and the roadmap is bundling every adjacent financial product into it at zero marginal price.

The Plan:

The Free Sample Share

The problemBrokerages charge money, and people hate paying money.

  1. 1Hand out one free share on the sidewalk to anyone who makes eye contact
  2. 2Make selling it possible only by phone, inside a two minute window on a Tuesday
  3. 3Send forty notifications a day, plus one at 3am about a stock they do not own

The objection“This is not free, they pay in attention.”

The answer“Attention is not money.”

Why it worksNinety percent of people who take a free sample buy something, and the rest become customers anyway when the 3am notification wakes them up. Acquisition is the product, and everything else gets bundled into it later.

Estimated startup cost$0.03, and forty notifications a day

Free to start. Impossible to stop checking.

CASE #010
Brian Armstrong COIN Coinbase Brian Armstrong Operational Grinder
screens hit
last price $185.74 -1.32%
market cap $49.0B 5Y return -41.9%
why it is in the book Wins the unglamorous half: licences, custody, compliance and cold storage. That regulatory surface is the barrier, and the assets custodied are the switching cost.

The Plan:

The Vault With Paperwork

The problemCrypto is unregulated, and customers are, reasonably, nervous.

  1. 1Read the entire rulebook out loud to a regulator, then collect a licence from every government that has one
  2. 2Collect three more licences from governments that do not exist
  3. 3Register the vault as a museum, so the assets are exhibits, the public pays admission, and nothing can be seized

The objection“This is not a museum.”

The answer“It is a museum. The exhibits are just very liquid.”

Why it worksDoing the boring, expensive, regulated part is the barrier that keeps everybody else out. The assets on display are the switching cost, and so is the paperwork required to view them.

Estimated startup cost$40,000, mostly licences from countries that do not exist

Boring is the moat.

Swipe, scroll, or use the arrow keys.

The five screens

Every position has to clear the left side of at least two of these. The right side is what disqualifies. Click a screen to see which files hit it.

When two screens stack

Four founder archetypes

Structural moats show up as a personality before they show up in a spreadsheet.

  • The Ecosystem Architect

    The Steve Jobs

    Rejects fragmentation. Obsesses over vertical integration by controlling both the physical hardware and the proprietary software, to create a frictionless, closed-loop user experience that is incredibly difficult for competitors to penetrate.

  • The Ruthless Aggregator

    The Bill Gates / Zuckerberg

    Prioritises distribution above all else. Builds a massive user network with a wedge product, then aggressively clones, bundles, and drops the price of adjacent tools to zero, turning stand-alone competitors into mere platform features.

  • The Operational Grinder

    The Jeff Bezos

    Attacks brutal, low-margin, real-world logistics or regulatory problems. Wins through sheer density and high operating leverage, building a moat out of capital-intensive physical infrastructure that pure software companies are afraid to touch.

  • The Algorithmic Engine

    The Larry & Sergey

    Solves massive, messy human problems through pure engineering and proprietary data loops. Originates from deep technical backgrounds and relies on algorithmic superiority, where the product inherently gets smarter with every interaction, rather than on subjective design.

  • Spend ruleManagement must spend on product over marketing.
  • MissionOwn the product-led builders, avoid the story-led spenders.

Post-mortem rules

Written before the position, so the exit is not a negotiation with myself.

  1. 1Don't invest in a CEO that doesn't have good experience.
  2. 2Marketing masks product friction.
  3. 3Trust is a short-term moat.
  4. 4Distribution beats features.
  5. 5The middleman gets squeezed.
  6. 6Populist hate leads to regulation.
  7. 7Early founder exit is an early warning.
  8. 8Freedom fighters, mercenaries, conscripts. Know which one you backed.
Warren Buffett

“Freak the fuck out and panic sell everything right now. It's fucking over.”