Writing.

The most constrained environments will build the most complete AI workflows first — not because of technological advantage, but because in those environments the process-elimination question is forced rather than optional. The gap is not the barrier. It is the brief.

Glow was never a product — it's the look of healthy skin, and healthy skin is personal. For deep-skinned people, the market sold products built for someone else and left them to expensive, marking trial and error. The new glow is personalization, and Zahra is how it finally scales.

Kenya's civil society compliance gap is not a story about organisations evading registration. It is a story about a registration system designed for organisations with lawyers — and what changes when the infrastructure to navigate it finally exists.

For well-specified, high-volume, rule-based processes, human involvement does not produce accountability. It produces a bottleneck. The instinct to keep humans in the loop is the institutionalisation of distrust in systems that were never designed well enough to deserve trust. That is no longer the only option.

Most of what is being called AI transformation is acceleration of existing processes, not elimination of them. The pattern is identical to enterprise software in the 1990s. The organisations that will actually transform are the ones asking a different question.

The expense report is one of the most universally resented administrative instruments in corporate life. Understanding why it exists — what it is actually for — is the prerequisite for understanding why agentic payment infrastructure makes it obsolete.

A manifesto for workflow-native AI. The test is binary: when the task was finished, did the AI do it, or did it help a person do it? Most of what is being called AI transformation fails this test. Here is why the distinction matters, and what it means to build software that passes it.

The PBO Act 2013 is better legislation than what it replaced in almost every respect. Registration rates did not meaningfully improve after it passed. The failure mode in African regulatory reform is almost never the quality of the law. It is the absence of the translation layer that makes the law navigable.

International donors' due diligence requirements are producing compliance pressure the Kenyan state has never consistently applied. The organisations that cannot meet them are increasingly the ones doing the most essential ground-level work. Nobody designed this outcome. Understanding it is the first step to addressing it.

The AI and law conversation is fixated on the hard problem — can AI reason about law? It has missed the tractable problem that is already solved: can AI read a document against a specific checklist and identify gaps? The answer is yes. This is what PBOMaster does, and what it finds.

The PBO Act's requirements are not unreasonable. They are inaccessible. This essay walks through what compliance actually demands — clause by clause — and explains why the gap between what the law requires and what a typical CBO has is structural, not intentional.

The hyperlink is a navigation primitive. It connects documents. What it cannot do is bring the explanation to you — inline, at the point of confusion, without a context switch. The web gave us extraordinary access to knowledge. It never built the layer that helps you understand any of it.

The SaaS pricing page is not where a company tells you what its product costs. It is where a company attempts to alter your decision-making. An AI procurement agent has no psychological vulnerabilities. It cannot be anchored, socially pressured, or confused by artificial urgency. This is going to restructure B2B software pricing — structurally, not incrementally.

Every procurement solution since 1950 — charge cards, corporate cards, P-cards, virtual cards — has been a partial answer to the same trust problem: how do you authorize a specific transaction without manually validating every purchase after the fact? Stripe's agent wallet, launched April 2026, is the first complete answer.

Location data records where your body was. Purchase data records what your wallet did. Epistemic data — what you looked up, what confused you, where your understanding ran out — records what your mind was doing. It is more intimate than any other category of behavioral data. And the tools being built to help people read are being built right at its center.

Kenya's PPADA 2015 treats software like office furniture — specifiable in advance, comparable by price, deliverable on a fixed date. The result is a predictable cycle of failed systems, vendor capture, and fraud-enabling infrastructure. This essay argues for reconceptualising software as a distinct category of public acquisition.

Kenya has 100,000 Community Health Promoters visiting every household monthly. The data they collect mostly disappears. Here is what a well-instrumented community health system would look like.

Most software deployments in Africa fail for the same reason: they were built assuming connectivity that doesn't exist. Building for offline-first isn't optional — it's the prerequisite for data integrity.

We talk about democracy as values and ideals. But in practice, democracy is what runs on the infrastructure. When the infrastructure is a WhatsApp group, the democracy is whatever fits in one.

The debate over CBDCs is a distraction. What Africa needs, now, is stablecoin-native payment rails that route around the 7% tax the legacy system levies on every transfer.

A utility that can see its own operations in real time. A constituency office that knows the status of every case. What changes when Africa's institutions can finally see themselves?

Africa's public institutions aren't failing from a lack of resources. They are failing because they have never had the right tools. The technology exists — it just hasn't been built for here.

Africa's institutions are not short of mandates, people, or data. They are short of the operational layer that lets them see themselves. This is the thesis that started Flux.