Day 1000 - The Experiment Begins
My income streams, staring position, and what I am testing
This post is free to read. Starting next week, lab notes are for paid subscribers only. If you’re on the fence, read this first, and then decide.
There’s a version of this post where I tell you I’ve figured something out.
I haven’t. What I have done is looked at where things are heading — for developers, for knowledge workers, for anyone whose income depends on skills that are becoming cheaper by the month — and decided that working through it in public, with real numbers and outcomes, is more useful than pretending I have a map.
I don’t have a map or a crystal ball. I have a set of hypotheses and a limited amount of time to test them.
So, what do you actually get out of reading this? A real-time record of what one person is trying, what it costs, and what happens. Not a framework. Not a success story reverse-engineered from the ending. Something closer to a flight recorder — running whether things go well or badly, with nothing tidied up after the fact. If you’re navigating the same terrain, that’s more useful than most of what’s out there.
That’s what this newsletter is. Not a self-help column. Not a success story told in retrospect. A lab notebook, written in real time.
The situation, plainly stated
I’m a software developer and architect with thirty years of experience. I’ve worked across development, architecture, analysis, CTO roles, consulting, mentoring and coaching. I know this domain well.
I’ve also watched AI quietly pull the rug out from under a business I built. Not because I executed badly, but because the structural conditions that made it viable disappeared. The training market for developer skills. Gone. First slowly, then all at once. All over in a couple of years.
And this isn’t just my problem. Watch what’s happening in any development team right now. Developers who once wrote code are increasingly just reviewing what AI agents produce. That might sound like an upgrade, and in some ways it is, but let’s be clear about what it means economically: you’ve just become a quality control layer. The next generation of agents, the ones that make fewer errors, need fewer human reviewers. The arithmetic is simple. Shouting, for those willing to look at it directly.
I’m in my mid-fifties. This hits differently when you’re not twenty-eight with thirty years of runway ahead of you. But I’d argue it hits harder for the twenty-eight-year-olds — they just don’t know it yet. At the current pace of AI development, the knowledge work landscape in three years will look nothing like it does today. For anyone whose income depends on cognitive tasks, the question isn’t whether this affects you. It’s when, and how much, and what you’re going to do about it.
The experimental capital I have available to work with is real but bounded. Enough to be meaningful. Not enough to be reckless. Every decision I make here is made under the same constraint most of you are working under: genuine consequences for getting it wrong. A thousand days is a very short time horizon. Most of us will need to do some fairly fancy footwork before the clock runs out.
That’s not a disclaimer. It’s what makes this worth reading.
The countdown
Day 1000 is April 8, 2026 — the day this experiment begins. Day 0 is January 3, 2029.
The clock is running.
By Day 0, the employment landscape for knowledge workers will look materially different from today. Compressed rates. Fewer opportunities. More competition from AI-assisted generalists, and from AI itself. Contractors will feel it sooner — there’s no organisational inertia to buffer them, no HR process, no redundancy package, no three-month consultation period. Just fewer renewals. But employees aren’t safe either. The permanent role just has a slightly longer fuse.
Nobody really knows how governments and central banks will respond to job displacement at this scale. It’s new territory. The policy toolkit wasn’t designed for this. I find that either fascinating or terrifying, depending on the day.
If I’m still primarily dependent on contract income when the clock hits zero, the experiment will have failed.
That’s the hard constraint the rest of this is built around.
The four streams: what I’m actually testing
Most income diversification advice is too abstract to be useful. Here’s what I’ll actually be testing, why, and what I honestly think about each one. I’ll go deep on each in future posts.
1. Contracting. The floor, not the future
Still doing it. Right now, it’s my primary income source. It supports my family and funds the experiments. Without it, there are no experiments.
But let me be direct about what it is and isn’t. Contracting income at the senior end is heavily taxed in a progressive regime. You trade time for money, the government takes a large cut, and it scales linearly. More hours, more income, no compounding. If I try to squeeze more out of it after hours, diminishing returns kick in fast: I’m working into my rest periods, my day-job productivity suffers, and the taxman cheerfully harvests much of the gains anyway. I’d just be earning regular income, the hard way.
The strategic role of contracting in this experiment is to preserve runway while the other things develop. I’ll keep my eyes open for extraordinary opportunities here. But I’m not holding my breath.
Honest assessment: Useful short-term. Structurally limited. Probably shrinking as an opportunity before Day 0.
2. Investing. Real but slow
Investing is on the list, but I want to be straight about where I’m starting from: the capital I have available for this is limited, and a fair amount of it is still in the process of being saved. If you’re in a similar position — income mostly spoken for, not much left over at the end of the month — that’s exactly the situation I’m designing these experiments for. You don’t need a war chest to start, though it obviously helps. You do need to be deliberate about what you do with whatever you have.
Here’s the honest version of what investing can do: doubling your money in three years is a 26% annualised return — exceptional by any measure. It’s just not going to solve a cash flow problem in the near term.
I’ll also confess something. In the past, I’ve been sloppy about this. I spotted two exceptional opportunities in the last two years, knew it at the time, and did nothing with either of them. One would have doubled my money in six months. Another went up 300% over two and a half years. I watched both play out from the sidelines. That kind of thing is clarifying. Good investing turns out to be as much about psychology — what you do when you’re losing, whether you act when you should — as it is about finding the right opportunities. Through general laziness I’ve managed about 10% per annum. Not bad. Also not where we need to be.
I’m being more rigorous now. I’ll document positions and reasoning here as a record you can learn from and adapt — not a system to follow blindly, but a real account of decisions made under real constraints, scored against outcomes.
What investing does particularly well: it runs in parallel. It doesn’t require much of my time to compound. And the analytical skills — reading balance sheets, assessing competitive moats, thinking probabilistically about outcomes — transfer directly to evaluating business opportunities. Which brings us neatly to stream three.
Honest assessment: A genuine parallel track. Best case, I double the capital over three years. Worth running seriously regardless.
3. Building or buying a business. Where the real opportunity is
This is where I think the actual upside lives. It’s also the most complex category, and the one I know least about, beyond what transfers from investing. I’ll be learning this in real time and sharing the process as I go. Feature or bug, depending on your perspective.
The barrier to building software products has collapsed. Any competent developer can now build and deploy in days what would have taken a team months not long ago. That’s simultaneously the threat to contracting income and the opportunity on the other side: if you can identify the right problem, the cost to build a solution is lower than it has ever been in the history of software.
But this stream isn’t just for developers. Bricks-and-mortar businesses are interesting for a different reason, and they’re accessible to anyone willing to do the analytical work. Physical reality still has friction. A good local business — genuine repeat customers, solid margins, owner ready to exit — doesn’t become obsolete because the latest model dropped. And the analytical toolkit from investing transfers directly: you’re still reading numbers, assessing moats, thinking about what makes an economic position durable. The skills compound across categories. Which is convenient, because I need all the compounding I can get.
I’m actively looking at opportunities in both categories. When I find one worth pursuing, you’ll read about it here — including the reasoning, the numbers I can share, and the parts I got wrong.
Honest assessment: Highest potential. Highest complexity. The category where this experiment either works or doesn’t.
4. This newsletter. The meta-experiment
I’ll be straight with you: I have no idea if this will work.
The thesis is a timing bet as much as anything else. Most developers aren’t ready to hear what’s coming. Not because they’re unintelligent, but because the disruption hasn’t hit hard enough yet to feel personal. The AI agents are still making silly mistakes. That will change. The question is whether establishing a clear-eyed voice now, before the urgency peaks, puts this newsletter in a position to matter when it does.
If the timing is right, the readers who show up early themselves gain the most value and, in turn, become the most valuable subscribers of the newsletter. If I’m too early, I’m writing into a relative void for a year or two while the market catches up to the thesis.
Either way, this is a business experiment in its own right, not a side project. The metrics I care about: paid subscribers, churn, revenue per subscriber, and whether the writing holds up when I re-read it in two years.
Honest assessment: Complete unknown. The experiment is live whether I like it or not.
The prediction ledger: one call to open the account
This section is what separates lab notes from commentary. I make specific, falsifiable predictions. I assign confidence levels. I score them when they resolve.
The scoring system is the Brier score — a proper scoring rule that rewards calibration, not just correct picks. The short version: you’re penalised for being confidently wrong, and rewarded for being accurately uncertain. If you want the full technical explanation, ask, and I’ll write it up. It keeps me honest in a way that vague prognostication doesn’t.
Some issues will have a prediction. Some won't. Here's the first one.
Prediction #1: The OECD unemployment rate, currently sitting at 5.0% (January 2026), will rise by at least 2 percentage points to above 7.0% by end of 2026 — approximately 1 percentage point attributable to macroeconomic headwinds (trade disruption, geopolitical friction, slowing growth), and at least 1 percentage point attributable to AI-driven displacement of knowledge work.
Confidence: 75%. Scores on Day 650 (approximately March/April 2027, allowing a quarter for the December 2026 figures to be released). Baseline and scoring data: OECD.org monthly unemployment releases.
The macroeconomic 1% feels relatively uncontroversial given where we are with wars, tariffs, trade fragmentation, and general global uncertainty. The AI 1% is the more interesting call. Organisational inertia will buffer wholesale displacement for a while — companies don’t restructure overnight, and the transition has retraining and liability dimensions that slow things down. But the trickle has started. By end of 2026, I expect it to be visible in the numbers.
One caveat worth making explicit: if AI investment crashes between now and then, that doesn’t invalidate the thesis. The dotcom bust in 2000 didn’t mean the internet wasn’t going to matter. It just meant the market got ahead of itself. The underlying capability keeps developing regardless of what the Nasdaq does.
What comes next
Paid posts go out weekly. I’m committing to that cadence partly because there’s enough happening to write about, and partly because without it I’ll let things slide. You’ll get positions taken, decisions made, things read and found useful, outcomes scored. Easy calls and harder ones. Occasionally, I turn out to be wrong about something in public.
Next week goes deeper on investing — specifically, the approach that led me to those two opportunities I mentioned, and why you don’t need to spend all day buried in annual reports to identify a candidate worth looking at. I also built a Python script that does much of the intelligence-gathering for me. More on that next time.
If you have questions about the setup, the predictions, or anything I’ve glossed over here — comment or reply. I read everything.
The clock is running. Tomorrow is Day 999.
— Olaf
Before you go: one practical tip
Each issue will end with something immediately useful. Not a framework. Not a mindset. Something you can act on.
This one is about debt, and it matters more right now than most people realise.
If your income outlook is uncertain, reduce your debt.
The logic is simple. Debt is a fixed obligation in a world where your income is becoming variable. It doesn’t care whether you’re between contracts, whether your hours got cut, or whether the renewal didn’t come through. The repayment is due regardless.
Most people think about debt in terms of interest rates. The more important variable right now is fragility. A mortgage extension, a car loan, a credit card balance — each one narrows the gap between an income disruption and a genuine crisis. The bigger the debt load, the less runway you have when things get bumpy. And things are going to get bumpy.
This isn’t about becoming debt-free overnight. It’s about not adding to the pile while your income is still reliable. The time to shore up the foundations is before you need them, not after.
If you’re currently thinking about taking on new debt — extending the mortgage, upgrading the car, whatever it is — ask yourself one question first: how does this look if my income drops 30% in twelve months? If the answer is uncomfortable, wait.
That’s the tip. Simple. Not exciting. Worth more than most advice you’ll read this week.
The Next 1000 Days documents one person’s attempt to build financial resilience in the face of AI-driven disruption to knowledge work — in real time, with real numbers, scored against outcomes. If someone you know is thinking seriously about this, forward it to them.
Nothing here is financial advice. Any capital you lose acting on my reasoning is your problem. Any gains are obviously due to my outstanding insights.