Necessity, the Mother of Seventeen Inventions

A story about building technology when nobody is coming to save you.

Amended 1 September 2026. Seventeen was true when this was published in July. It is now twenty — SEOboss, FAULTLINE and Strategy as Software have landed since. The timeline figure below has been extended to carry them and the three stories are added as a postscript at the end. Everything between the two is left exactly as it was written, because it was written in July, and quietly editing a past self into looking better informed is its own small dishonesty. The title stays as well: it names the essay rather than a running total, and renaming a piece people have already shared is a worse crime than an out-of-date number.

There is a comforting myth about how technology companies get made. A founder has a single, blinding insight — a bolt of lightning over a whiteboard — and from that one idea, in a straight and well-lit line, a product emerges, then a company, then, if the gods are kind, a fortune. It is a lovely story. It is also, for almost everyone who has actually done the thing, nonsense of the highest order.

The truer story, the one worth telling, is messier and — I’d argue — considerably more interesting. It is a story about constraint. About the strange, almost alchemical way that not having enough money, not having enough time, not having the right visa stamped in the right passport, forces a kind of invention that comfort never would. It’s the story of seventeen products built over five years, not because a business plan demanded them in that order, but because life, in its usual unglamorous way, kept presenting problems that nobody else was going to solve.

It’s worth saying plainly, before going any further, why the reasoning behind each of these decisions matters as much as the products themselves. It would be easy to read a list of seventeen things and conclude that whoever built them was simply restless, or scattered, or chasing whatever shiny problem crossed his path that week. Understanding the thinking — why this tool, in this order, for this reason, at this exact moment of constraint — is what turns that list from a curiosity into a case study. It’s the difference between watching someone build and understanding why they built the way they did, and it’s that second thing, the rationale itself, that’s actually worth a reader’s time.

The idea that arrived a year early

Let’s begin, as any good story should, near the beginning — though not quite at it, because the beginning of this beginning is itself instructive. In early 2021, a year before the world watched tanks cross a border that changed everything, an idea called MiFamilias was taking shape. Its premise was, on reflection, rather ahead of its time: that social media as built — engagement-hungry, ungoverned, indifferent to the people caught inside its mechanics — was not a fixed law of nature but a design choice, and design choices can be unmade. MiFamilias set out to rebuild those mechanics around something unfashionable at the time: the family, with real governance built in rather than offered as a buried settings toggle nobody finds.

It’s worth pausing on the timing, because it tells you something about how ideas actually move through the world. This wasn’t reactive. Nobody in early 2021 was clamouring for family-governed social platforms. And yet five years on, Australia, the EU, and the UK have all passed legislation nudging platforms toward precisely this kind of accountability, with the US and others now openly weighing similar moves. Being early isn’t the same as being wrong. Sometimes it just means waiting for the world to catch up to the problem you already saw.

The trouble was that the development team was based in Ukraine. And then, in February 2022, the world intervened in the way it sometimes does, indifferent to product roadmaps, and the team that was meant to build this careful, humane alternative to the attention economy found itself instead fighting for something immeasurably larger. The project didn’t fail. It simply stopped, the way a conversation stops when someone in the room receives terrible news, and everyone understands, without needing to be told, that nothing else matters right now.

The company that Singapore’s paperwork wouldn’t allow

The same year, 2021, produced a second idea: Bertha, an operating system for marketing agencies, eventually stitched together from nineteen microservices. It was self-funded, with a business partner who had committed to the unglamorous but essential work of paying for bootstrapping. He didn’t. And so began a pattern that would repeat, with variations, for years: promises of capital that evaporated on contact with reality, and a founder left holding a half-built thing and a choice — abandon it, or become, out of necessity rather than ambition, the entire engineering department.

Bertha was incorporated in Singapore, which brings us to one of the quieter, more human threads running underneath all of this. A spouse visa — his wife’s, not his own — meant he was legally barred from working there almost wholesale. On top of that sat Singapore’s well-known protectiveness around local employment: a policy stance he’d say carries a certain amount of bias rooted more in how the talent landscape used to look than how it actually looks now. Between the two, there was really only one route left — incorporate his own entity, Agencio Cloud, as the holding company for what would become Bertha, so that at least he could build something as its owner rather than as a worker the system had already ruled out. That workaround wasn’t free either: several thousand additional Singapore dollars in setup and accounting fees, just for the right to legally build his own thing in the country he was living in. It is a strange kind of exile, to be present in a country, embedded in its rhythms, and forced to structure a company around the one thing the law wouldn’t let you do directly. The move to Bangkok solved that particular problem and promptly handed him a new one: a city that runs overwhelmingly in Thai, a language he doesn’t speak, and with it, no existing professional network to lean on. If you have ever tried to build something serious while simultaneously being, in a very literal sense, unable to ask anyone in your immediate vicinity for help, you’ll understand why this detail matters. It isn’t scene-setting. It’s the actual texture of the constraint.

When you can’t hire, you build

Here is the physics of this whole story, if it has any: every time there wasn’t money for a role, a tool got built instead. This is not a virtue so much as a consequence, but it produced, almost as a side effect, an entire portfolio.

There was no budget for a DevOps hire, and the person who’d promised to help became — for reasons that remain, charitably, unclear — uncontactable. So K8 Inspector got built: a full Kubernetes multi-cloud management and AI DevOps tool, born specifically because the commercial alternatives were either priced for companies with venture funding or simply the wrong shape for the job. There was no budget for a security team, so a testing suite got built, and after it, a penetration-testing engine, and eventually SILO — an architecture for watching AI agents for exactly the kind of compromised or rogue behaviour that keeps security engineers awake, sketched for a long time and finally finished to coincide with Bertha’s own launch. There was no appetite for selling Bertha directly into agencies one contract at a time, so a licensing and billing platform got built, deliberately as its own standalone product, because the code could clearly be reused elsewhere — which it was, repeatedly, across the portfolio that followed. Then a friend saw what a LinkedIn Poster build — done roughly 90% with CodeEasy — could actually do, and came round asking what he was up to with crypto — a conversation that surfaced the friend’s own trading tooling was running on bad data from third-party sources. So a proper firehose data engine, memedataengine, got built instead, and turned around to help the friend develop his own trading product, now in progress with early testers. Out of that same frustration with bad data came Posit: tax-grade transaction parsing and wallet analytics for Solana memecoin traders, built in part with CodeEasy, addressing a genuine gap in the market for traders who will, eventually, need to file taxes, however much they’d rather not think about it.

It’s worth putting an actual number on that “no budget,” because vague talk of struggle is easy to wave away and a figure isn’t. Bootstrapping this portfolio over the past four or five years has cost, in total, somewhere in excess of £65,000 — call it $87,000 at today’s exchange rate. That figure isn’t one lump sum for one thing; it’s the accumulation of every ongoing line item a company like this actually carries — hosting, the AI usage that now underwrites so much of the build speed, the early Figma and development help brought in to get MiFamilias off the ground before the war made the question moot, and the less glamorous cost of simply maintaining partner relationships: flights, hotels, the business trips that don’t show up in a demo but are how deals actually get made. None of it is investor money, and none of it was ever treated as disposable. These are not rich people funding a hobby; every one of those pounds was felt on the way out, and the habit of making every penny count is precisely why so much of this got built by hand instead of bought off a shelf. The frugality isn’t incidental to the story. It’s the mechanism.

And it hasn’t happened in isolation from the rest of life, which carries on regardless of how a Series A is going. Living in a country where education isn’t free means school fees for his son are simply another line in the same ledger as hosting and AI credits — not a footnote, a real and recurring cost, carried at the same time as everything above. That load has been shared, not shouldered alone; none of this would have been remotely possible without his wife’s support, financial and otherwise, through years that asked a great deal of both of them. On that account, whatever else has been hard, he’d be the first to call himself lucky. Maybe very lucky.

And The Possible carries that second, personal thread too — the visa, the language barrier, the isolation of starting a network from absolute zero in a new country. Having spent three or four years moving through exactly this category of founder problem, the idea widened from “a product for traders” into something closer to a paradigm: a not-for-profit build-tank, co-founded with his wife Veronica, connecting founders, builders, and investors who hit the same walls, in whatever city they happen to have washed up in. Whether it also happens to bring the right people into his own orbit is, he’d say, almost beside the point. It’s a story he believes is worth telling because it might spare someone else the five years it took him to learn it the hard way.

Even the products that look, on paper, like indulgences turn out to have the same DNA. SpeakEasy — a macOS speech-to-text tool — exists because he is dyslexic, does a great deal of note-taking, and found that the market’s offerings either demanded a constant internet connection he didn’t reliably have, or, in Apple’s case, simply weren’t accurate enough to trust with anything important. So he built the tool he actually needed, and it now runs offline, on his terms, precise where the alternatives were approximate. LinkedIn Poster started life as a smaller thing again: a test of CodeEasy as a build tool, nothing more — roughly 90% of it built by CodeEasy itself, start to finish. It’s worth sitting with that figure for a second, because it’s the whole argument in miniature — a working product, not a prototype, with the machine doing nearly all of the actual building. It was at that exact point, watching what CodeEasy alone could produce, that The Possible stopped being a nice idea he might get to one day and became something he knew, with real certainty, was fully achievable. SignalFabric — an autonomous trading system, now quietly running in paper mode with promising early results — draws on years spent around fintech and institutions like Credit Suisse, built partly because the honest truth about bootstrapped founders is that money makes money, and he didn’t have any. Sidekick, the newest of the seventeen, wasn’t born from a gap in the market at all, but from a conversation with a friend working through a problem, and the mid-conversation realisation that what would actually help was an assistant that remembers everything and can be asked about any of it later — built first to run his own life, with half an eye on what it becomes if the advisory work he’s already doing informally is ever made formal.

The strange gift of not having enough

Here is where the Brian Cox in me wants to zoom out, because there’s something almost cosmological about this pattern, if you squint. Complexity in nature rarely arises from abundance. Diamonds are carbon under unbearable pressure. Stars only fuse heavier elements once the easy fuel runs out. Evolution doesn’t design organisms in comfort; it selects for what survives constraint. There is a real sense in which the seventeen products in this portfolio are the technological equivalent — not engineered from a boardroom’s abundance, but pressed into existence by an absence of money, an absence of the right visa, an absence of a partner who kept his word, an absence, even, of a functioning speech-to-text app that worked without Wi-Fi.

And here’s the Fry in me, unable to resist the observation that there is something wonderfully absurd about the whole enterprise too — a man in Bangkok, unable to order a coffee in the local language, building the security architecture that monitors artificial intelligence for signs of going rogue. If you’d pitched this as fiction, an editor would have sent it back for being implausible. And yet. Here we are.

The most recent chapter accelerates the whole pattern, almost dizzyingly. The last few products — memedataengine, The Possible, Sidekick — came together in a fraction of the time the earlier ones took, once Fable entered the toolchain. That speed carries its own cost, a genuine and non-trivial one, but that’s a ledger for another article. The point here isn’t the invoice; it’s the acceleration itself, and what it says about where this is all heading. If five years of necessity could produce seventeen products built more or less by one person, in the gaps between visa applications and language classes and a business partner’s broken promises, it’s worth asking, soberly, what the next five years look like now that the tools for building have themselves become extraordinary.

There’s a sharper edge to that acceleration worth naming honestly, too. Building on frontier AI tooling means racing a clock that isn’t yours — pricing, access, and capability all move on the hyperscalers’ schedule, not the founder’s, and there have been more eighteen-hour days than he’d like spent simply trying to stay ahead of that. It stopped being an abstract risk in June 2026, when Anthropic briefly suspended access to Fable and its sibling model Mythos to comply with US export controls, restoring it a few weeks later on the first of July. Nobody building on top of someone else’s platform gets a vote on decisions like that. It’s part of why the toolchain behind all of this has been kept deliberately varied rather than locked to a single vendor — so that a decision made somewhere else, for reasons that have nothing to do with the work itself, can’t simply stop the work.

A record you’d play with the lights off

There’s a particular strand of nineties garage that understood something about this before any of us did — the sort of 12” you’d find at the bottom of a crate, worn from being played out at the wrong hour on too many nights. T2 featuring Robin S released “You Got the Love” in 1997 on Champion, and the Curtis & Moore Vocal Club Mix is the version that sticks: nearly seven minutes built around a promise to keep holding on, to keep giving everything there is to give even when the going gets rough, because turning back was never really on the table. It isn’t a song about having already made it. It’s a song about the holding-on itself — the part nobody puts in the highlight reel.

That’s the register this whole story wants to be read in, if I’m honest. Not a triumphant founder narrative with the sad bits edited out for the pitch deck, but something closer to that record: unglamorous, built for the small hours, carrying real pressure underneath the energy, and stubbornly, defiantly upright anyway. The Kubernetes tool wasn’t built to impress anyone. The speech-to-text app wasn’t built to look good in a portfolio. They were built the way that track keeps its promise for seven straight minutes — because the alternative was giving in, and giving in wasn’t an option. Under enough pressure, misery doesn’t just sit there being miserable. It gets put to work. It becomes the tool, the product, the thing that outlives the circumstance that forced it into being.

The thesis, if there has to be one

If there’s a single thread holding seventeen apparently unrelated products together, it isn’t a market thesis, or a growth strategy, or even artificial intelligence itself, though AI runs through nearly every one of them now. It’s simpler and older than that: when you cannot buy the thing you need, and walking away isn’t in your nature, you build it — and you keep the whole of it, IP and lesson and scar tissue together, when you’re done.

It has been, by his own account, financially brutal and emotionally exhausting in roughly equal measure. Nobody sets out to lose a development team to a war, or to be barred by paperwork from working in the country they live in, or to watch a business partner’s promises quietly dissolve for the second or third time. And yet it would be dishonest to call this a sob story, because it manifestly isn’t one — it’s a portfolio of working software, built by someone who simply refused to let the absence of resources be the end of the sentence. Adversity didn’t derail this. Looked at squarely, adversity is the architecture. Every wall he hit is, quite literally, load-bearing in what got built next.

That is, I think, the real story here — not seventeen scattered inventions from a restless mind, but one continuous, entirely rational response to circumstance, repeated for five years, until the pattern itself became the point. Or rather — until it stopped needing to prove anything at all, and became, simply, how he survived.

The method inside the madness

It would be easy, reading this so far, to mistake the whole thing for improvisation — one crisis after another, answered on instinct, with the portfolio as an accidental by-product. That’s not quite right, and it’s worth being precise about the difference, because it’s the difference between a lucky escape and a strategy.

Look again at the order things arrived in and a pattern holds that isn’t accidental at all. Nearly every tool was built to be standalone and reusable from the outset, not bolted permanently to whichever product first demanded it — License Server was deliberately separated from Bertha specifically so its billing and licensing logic could be redeployed elsewhere, and it was; SpeakEasy’s underlying tooling started life inside MiFamilias and later resurfaced, rebuilt, as its own product; the code inside Humanify carried straight across into Bertha’s own architecture. That’s not a founder throwing things at the wall. That’s a conscious decision, repeated deliberately across five years, to treat every piece of engineering as an asset that could outlive the reason it was first written — and, just as importantly, to own the whole of that IP outright rather than license fragments of it away to a client or a partner who wouldn’t still be there in two years.

The sequencing itself follows a logic too, once you look for it rather than at it. Governance and security tooling (the testing suite, the Siege Engine, SILO) was consistently built in the run-up to Bertha’s own launch, not before it was needed and not after — a deliberate hardening pass timed to the product it existed to protect. The infrastructure tools (K8 Inspector, License Server) arrived exactly when the underlying systems they managed reached the complexity that made them necessary, not before. And the recent cluster — memedataengine, The Possible, Sidekick — leaned on everything built before it, from the coding tools to the reusable code libraries, which is precisely why that cluster could move so much faster than the years that preceded it. None of that is coincidence. It’s a rational allocation of scarce time and scarcer money toward whatever the moment’s biggest constraint actually was, executed in the same order a well-run engineering organisation would have chosen if it had been given the choice — except here, there was no organisation. There was one person, doing the prioritisation an entire product function would normally do, because the alternative to having a strategy was not having a company at all.

If there’s a lesson worth taking away from all of this, beyond the war and the visas and the language you don’t speak, it’s a quieter and more useful one: software innovation has, in real ways, become cheap. Not free — it comes at a genuine cost, financial and otherwise, and anyone who tells you differently is selling something — but cheap enough that a single, determined person can now do what would once have needed a funded team and a boardroom’s worth of permission. A one-person company with a genuine portfolio of working products isn’t a thought experiment or a LinkedIn aspiration. It’s demonstrably possible. It’s been done, here, seventeen times over, in the gaps between everything else that was going wrong.

If that’s actualised into something bigger, all the better. But even set against the years of financial and emotional strain it took to get here, the products stand on their own regardless of what comes next.

Postscript — three more, September 2026

Two months after this was published, three more had landed. They belong here rather than only on the portfolio page, because they run on exactly the physics described above.

SEOboss is the least romantic of the three and probably the most instructive. Search work — the audit, the drafting, the tidy list of everything wrong with your title tags — became a commodity somewhere around 2025, and commodity is a polite word for “a machine now does it for pennies at three in the morning”. The usual response to that is grief. The more useful response is to ask which half the machine actually took. It took the finding. It did not take the applying, and it certainly did not take the proving. So SEOboss drafts the fix, applies it to the live site, and then checks on the next scan whether the number moved — which is the part clients still pay for once the novelty of a score has worn off. It also, in passing, produced a good deal of the plumbing the third product on this list is built from. The reuse habit of the previous five years, arriving on schedule.

FAULTLINE points the same instinct at public companies: read what a company reports, what its management says and what the numbers show, and find the places where those three have quietly stopped agreeing. The short side first, because a negative thesis is the cleanest thing in the world to falsify. Its first real run, on 29 August, across fifty US retailers and 5,701 filings, found precisely nothing — top-decile contradiction pressure at 0.94 times the base event rate, which is to say very slightly worse than guessing. That is published in full, along with the bugs the real data exposed, because a result you only publish when it flatters you is not a result. It is also, in its way, the most honest thing in this whole essay.

Strategy as Software is the one that matters, and it turned up last, disguised as something duller. Reading a company’s board pack some months after finishing SEOboss, it became embarrassingly obvious that a tool which finds where a website contradicts its own claims is the small version of a tool that finds where a company contradicts its own strategy. That is the entire product: the plan read as testable claims, the budgets and hiring plans and contracts read as evidence, and the gaps reported as candidates with the innocent explanation already attached — for an incoming chief executive with ninety days and eleven thousand documents, or an investor whose only instrument is a quarterly pack written by the people it judges. It is in beta with a written kill criterion, which is the single most useful thing five years of this teaches you to attach to anything. The longer argument is in On Track, According to the People Being Judged.

Twenty, then. The portfolio page counts a little higher, because it also lists working builds this essay never told the story of and one kept purely for teaching — and I would rather explain the discrepancy than quietly round it away.

The arithmetic underneath is unchanged. When you cannot buy the thing you need, you build it, and you keep the whole of it afterwards. The only thing that has moved is how quickly the sentence finishes.

If any of this resonates — if you’re building something, backing something, or simply want to talk about what happens when necessity meets the right tools — get in touch. The investors page on this site has the details.

---

Amendment history — 1 September 2026: extended from seventeen products to twenty with the addition of SEOboss, FAULTLINE and Strategy as Software. The timeline figure and its caption were updated and the postscript added; the original July text between the banner and the postscript is unchanged.