Six domains renew on my card every year. padiso.co, searchfit.ai, d30.io, capitaly.vc, brightlume.ai, d23.io. Written out like that it reads as a diagnosis rather than a list. The man cannot finish anything.
I have stopped defending the list and started defending the arithmetic under it, because it is not six businesses. It is two businesses, one community, and three assets that exist because something else needed them first. Getting that distinction right is most of the work. Getting it wrong is how operators end up with six half-built things and no surplus anywhere.
The thesis
Running several ventures at once is not diversification. Every one of them draws on the same account, and that account is my judgement, not my capital. An asset you cannot split does not diversify anything.
What multiple ventures actually buy is optionality. Options only pay if you exercise them, and the exercise here is the exit, not the launch. So the allocation question is never what to fund. It is what you can get out of cheaply, and how fast.
The scarce resource is not money
Cash stopped being the binding constraint on a small software venture some time around 2024. A managed Postgres, an Apache Superset instance, a model API and a domain is a few hundred dollars a month. I can stand up the infrastructure for a new product in an afternoon. That is precisely the problem. When the cost of starting collapses, the cost of starting badly does not show up on the invoice. It shows up in attention, six weeks later, when the thing needs a decision and you are the only person who can make it.
Judgement hours are the real currency. They do not scale, they do not compound across contexts, and they are the one input every venture on my list competes for on the same day.
Which is why the allocation rule I use has nothing to do with expected value. It has to do with redeployability: if this venture stops working, how much of what I built inside it survives the decision to stop.
The study that changed how I think about this
Simone Santamaria published a paper in Management Science in 2022, volume 68, issue 1, pages 333 to 354, tracking more than 5,700 entrepreneurs longitudinally. The finding is uncomfortable and it is the reason I keep the list.
Portfolio entrepreneurs, people running several businesses concurrently, launch more successful ventures than single-business owners. Their individual ventures are also less likely to survive. Both things are true at once.
The mechanism is not what most people assume. Santamaria found no evidence that portfolio operators pick better opportunities at entry. Performance at launch is indistinguishable. The divergence appears later, and it comes from two things: exiting the losers faster, and redeploying people and capital out of them into the survivors. His phrasing is that redeployment reduces the sunkenness of their investments in new projects.
Read that backwards and it is an instruction. The advantage is not selection. It is the low cost of being wrong. Which means the only structural decision that matters, made at the start, is whether a new venture is built out of parts you can pull back.
What the ABS numbers are actually saying
The Australian Bureau of Statistics released its business counts on 18 August 2026. At 30 June 2026 there were 2,814,778 actively trading businesses. Over the year there were 460,461 entries against 375,331 exits, an entry rate of 16.9% and an exit rate of 13.8%. Net increase: 85,130.
Here is the line almost nobody pulled out of that release. Non-employing businesses grew by 83,105. Employing businesses grew by 2,025. So 97.6% of Australia’s net new businesses last year employ nobody at all.
That is not a small-business story. It is a portfolio story. Hundreds of thousands of people are adding entities to a personal stack rather than building firms, and the exit rate of 13.8% tells you how routinely those entities get shut. The behaviour is already normal. What is missing is the allocation discipline, because nobody teaches it to operators. They teach it to investors.
The allocation logic
Three questions, in order, for every entity on the list.
First: does it feed another venture. If yes, it is a shared asset, and the thing to fund is the asset, not the venture.
Second: if it feeds nothing, can I pull the people and the code out of it inside 30 days. If yes, it is an option. Small allocation, dated kill.
Third: if it feeds nothing and I cannot pull the parts out, it is a trap. Exit now, while exiting is still cheap.
Then once a quarter, the surplus goes to exactly one name. Not spread across the list. One.
How this reads against my own list
D23 is the clearest case. It is a managed Apache Superset platform, and it exists because the equity screening work in D30 needed a data layer that I was otherwise going to rebuild for every engagement. It is not a separate bet. It is a shared asset with a bill attached, and the correct question about it is never whether D23 is growing, it is what breaks in D30 if I stop paying for it. Fund the asset, not the venture.
SearchFIT sits on the second branch. It tracks whether AI answer engines mention a brand when buyers ask, and it is a real product with real customers, but it was also the instrument that told me what was happening to inbound at padiso.co. If it stopped tomorrow, the engineering, the crawlers and the evaluation harness would land back inside PADISO within a fortnight. Redeployable. That is what earns it a line.
Capitaly is the honest exception. I co-run it with Houman, which means it is not mine to allocate unilaterally, and a shared venture is a different instrument entirely. The moment a second person’s time is inside a venture, the 30 day pull-out test stops being a test I can run alone. I keep it in a separate column for that reason.
Three entities, three different answers, and none of them came from a revenue forecast.
Where this breaks
This logic falls apart the moment you take outside money. A priced round, a board and a vesting schedule are all devices designed specifically to make your commitment un-redeployable, and they work. If you have raised on one venture, the option value of the other five is not yours to harvest, and behaving as though it is will end badly for you and for your investors. The portfolio frame is for operators funding themselves. It is not a governance model.
It also breaks when the ventures share nothing. Six unrelated businesses is not a portfolio, it is six jobs, and the redeployment advantage Santamaria measured does not appear because there is nothing to redeploy. The shared asset is the precondition, not a bonus.
And the most honest objection is inside the study itself. Portfolio ventures survive less often. If your goal is one durable business that outlives you, the evidence says run one thing. I am optimising for a different outcome and I should say that plainly rather than pretend the tradeoff is not real.
What I would do on Monday
Write down every entity you pay for. Domains, subscriptions, companies, repos. Most operators cannot produce this list from memory, and the ones they forget are always the traps.
For each one, answer two questions only: what does it feed, and could I pull the people and code out in 30 days. Not revenue. Not potential.
Put a date on every option. An option with no expiry is not an option, it is a liability you have agreed to keep paying for.
Pick one name to receive the surplus this quarter. One. The failure mode is not backing the wrong venture, it is spreading the surplus so thin that nothing ever crosses a threshold.
Kill the cheapest thing on the list this week, even if it is small. The muscle you need is exiting, and the study says that is the muscle separating the two groups.
The six domains
The six domains still renew. What changed is that I stopped reading that renewal as six commitments and started reading it as one commitment and five positions, each with a condition attached and a date on it.
The list is not the flex. The kill rule is.
Most of what I write about here started as a client problem. If you have one, bring it to me.