Paul Graham argues brand-building is what companies do once real differentiation runs out. He is right about the failure mode and wrong about the category, and Jasmine Bina supplies the test that separates the two: what did this cost you to make, and can anyone tell.
A weekly intelligence brief from Base X Studio. Pan-African tech, global brand strategy.
Startups raised $1.44bn in H1 2026, roughly flat year on year, and the flat number hides the change. Q2 equity fell 40% to $260m across 38 deals while M&A nearly doubled. Flutterwave bought Mono in a roughly $35m all-stock deal, Paystack absorbed a struggling Brass, and the buying ran outward too: Spiro into the UK, Nomba into Canada, Yassir into France. When equity dries up and acquisitions double, the market is not slowing, it is changing the currency it settles in.
If you were planning a 2026 raise, price in a 40% smaller room and decide which side of the consolidation you are on. The companies getting bought are bought for one legible asset: a licence, a corridor, a customer base, a product an acquirer would otherwise build. Name yours in a sentence this quarter. If you cannot, you are not a target, you are a competitor to somebody who already has one.
One job, done well, in an app. Switching costs near zero, so a customer who uses only a lending app leaves for whoever is cheaper next month, and rising acquisition cost eats the difference.
MTN MoMo across 17 countries with 60m+ customers, Airtel Money across 14 with 35m+, both on agent networks venture money cannot replicate quickly. In Kenya, M-Pesa already holds the position four Nigerian categories are still racing for.
If your product is one job and your users could leave on a Tuesday, your moat is not your product and this quarter is when to admit it. You have three moves: attach to a distribution network you do not own, go so specific inside one segment that a super-app will not bother, or build the thing a platform will want to buy rather than copy. Watch the regulators before you commit capital. If your market rules for open-banking separation, the super-apps lose their data advantage and the specialist play gets substantially better.
Duplicating physical infrastructure has stopped paying, and the operators have said so out loud. The consequence is not cheaper data, it is that the physical asset stops being a competitive asset. When two rivals run on the same fibre, everything that distinguishes them sits above the pipe, in brand, service, pricing and product.
If you build on telco rails, your cost base and rollout timeline in these three markets are about to change, so re-run your unit economics before you sign off a 2027 market-entry budget. If you are a telco or an ISP, the harder implication: coverage, the thing you sold against for a decade, is converging with your competitor's. Every rand you were going to spend proving your network is better is now better spent above the pipe. Decide this quarter what you are, because "we have the best network" has an expiry date on it.
Corporate-strategic and crypto-native money together, not development finance and not generalist venture. Money like that buys a place in a payment chain, not a return on a consumer app. Alongside South Africa's Moment raising a $22m Series A for payment infrastructure, the direction is consistent: African stablecoins are graduating into plumbing a treasury department can be asked to approve. Whether regulatory permission arrives at the same speed as the capital is the unanswered part.
If your cross-border settlement is still a workaround somebody in finance manages by hand, you now have an institutional-grade option to evaluate, and that is a this-quarter task. Ask for the licence position in each corridor you actually move money through, not the company-wide claim. And take the wider lesson: the money that changes a category is strategic money, not venture money, because it arrives attached to a distribution relationship. Court the investor whose customers you want.
Cross-border buying ran the other way too. Spiro bought UK-based Coexlion, Nomba bought a Canadian payments company, and Algeria's Yassir bought a French adtech startup. African companies are no longer only the thing being acquired.
Branding is centrifugal, design is centripetal. Brand demands difference while good design converges on the same right answers, and Graham's case is the Swiss quartz crisis: once quartz made mechanical watches functionally obsolete, survivors like Patek Philippe stopped competing on accuracy and became luxury brands, enlarging logos, shifting the message from precision to price, eventually buying their own watches back off the secondary market to manufacture scarcity. He reads the AI moment the same way and tells founders to follow the problems instead. He is right about the failure mode and wrong about the category. Everything he describes is brand deployed to disguise an absence of substance, which is a real disease and not the only reason a business builds a brand.
Before you spend on brand this year, answer his question honestly, because he is describing a real trap and you might be in it. Is your brand covering for a product that has not earned its position, or catching up to a business that has? If the product is not proven, he is right and the money belongs in the product. If it is proven, your revenue is real, and the constraint is that nobody outside the room can tell why you are better, then his advice is the expensive one. Following the problems does not fix a business whose problem is that growth depends on the founder being in every conversation.
Dunford's admission is the useful part. The five-component framework from Obviously Awesome is now widely understood and teams still fail at it, so the problem was never the framework. Three of the four roadblocks she names are organisational: departments anchored on different competitors, teams underselling real strengths while fixating on gaps, and value stated so broadly it goes generic. Positioning does not fail because nobody read the book. It fails because agreement was never reached and the document papered over it.
Put that question to your head of sales, your head of product and yourself this week. If you get three different answers, no amount of copy will fix it, because your teams are selling against three different competitors. Fix the agreement before you touch the words. And if you have added a second or third product line since you last wrote your positioning, be explicit about what you are positioning, the company or the product, because leaving that undefined is why the messaging keeps drifting back.
Her other half is the one to keep: anything that costs the giver something real is appreciating. AI is exposing the implicit contracts underneath every relationship and forcing them to be renegotiated in the open, and the test buyers will apply is what did this actually cost you to make, and can I tell. That is a better definition of the decoration line than most of the industry has, because it is falsifiable. Decoration is cheap and reversible by construction. Infrastructure is credible precisely because it was expensive and hard to undo.
Audit your own signals this month against one question: which of these could a competitor have generated in an afternoon. Your positioning statement, your thought leadership, your case studies, most of your site. What survives is what you should be spending on, and what fails is costing you credibility rather than building it, because buyers are learning to spot the cheap version faster than you are producing it. The things that read as expensive now are the irreversible ones: named clients who will say it out loud, a public position you cannot walk back, a number you published before you knew it would be good.
The operative question moves from did users visit my site to did the model use my content when somebody asked it a question about your category.
Organization, Product and Offer schema, content in a format a retrieval system can parse, and a machine-readable statement of what you sell and who you sell it to.
This is aggregate industry commentary rather than a single named study, and the traffic-at-risk numbers circulating with it are second-hand. Treat the direction as real and the magnitude as unproven.
A retrieval layer will answer "what does this company do" from whatever it can parse. Vague positioning now costs you twice: once with the buyer, and once with the system deciding whether the buyer ever sees you.
Ask an AI assistant what your company does and who it is for, then compare the answer to your positioning. Ten minutes, and it will tell you more than a brand survey. If the answer is wrong or vague, the fix is not more content, it is structured data on your site and a clear, parseable statement of what you sell and to whom, which your developer can ship in a week. Do it before your competitors do, because the retrieval layer rewards whoever is easiest to cite with confidence, and almost nobody in your category has bothered.
Paul Graham published "The Brand Age" and argued that brand-building is what companies do once real technical differentiation runs out, and that the honest move is to follow the problems instead. He is right about that failure mode. Brand used to cover for a hollow product is a real and common disease, and it is worth naming because so much of the industry sells exactly that. But he has described one use of brand and named the whole category after it. Jasmine Bina, reading the same commoditisation the same month, supplies the distinction he skips: what appreciates is what cost the giver something real, and what was reversible and cheap to produce is discounted on arrival. That is the line between decoration and infrastructure, and it is a test rather than an assertion. Decoration is cheap by construction. Infrastructure is credible because it was expensive and hard to undo. So answer the question honestly before you spend, because both sides of it are real. If your product has not yet proven that customers will pay for it, Graham is right and the money belongs in the product. If your product has proven it, your revenue is real, and your growth is capped because you are the only person who can explain why the company matters, then brand is the constraint and following the problems is the expensive advice. The test is Bina's. Look at what your company has put into the world this year and ask which of it cost you something a competitor could not fake in an afternoon. Whatever passes is your infrastructure. Everything else is decoration, and Graham is talking about you.
Acquisitions nearly doubled while equity fell 40%. Telco rivals agreed to stop competing on fibre. Single-product fintechs found their moat sitting in somebody else's agent network. In each case a thing that used to be a defensible asset became a shared cost, and what remained defensible sat above it, in the customer relationship and the reason to choose you. That is the same question Graham and Bina are arguing about, arriving from the infrastructure side.