Adama Dia

Data & AI infrastructure architect. Writing about who controls Africa’s money, data and technology and what it costs us. Let’s connect today and be leaders

[Adama Dia @AdamaD41]

When Geoffrey Hinton , the man often called the "Godfather of AI" was posed a deceptively simple question, his answer revealed the profound gap between technical possibility and human reality.

The question: why not just build a superintelligent AI to solve all our problems and let humanity finally rest?

Hinton's response cuts both ways. In theory, absolutely. If we could engineer an AI that genuinely prioritizes human welfare, we should proceed without hesitation. His preferred analogy isn't dystopian fiction, but something far more intimate: motherhood. Just as a mother instinctively places her child's needs ahead of her own, a truly safe AI would require that same selflessness an intelligence far superior to ours, yet deeply devoted to our collective good.

The first hurdle is technical: we simply don't know how to build that yet. But Hinton sees an even bigger obstacle: human nature. In a global order where governments, corporations, and elites routinely act in their own interests, superintelligent AI wouldn't liberate everyone it would become a tool for the powerful few.

Hinton remains cautiously optimistic about the technology itself, envisioning a future where AI serves all of humanity like one unified family. Yet he warns that the hardest barrier may not be silicon or code but whether we can build the political and social systems wise enough to wield it responsibly.

[Adama Dia @AdamaD41]

@TheUnitedSpot__ @cb_doge Fingers crossed they all pass the class!

[Adama Dia @AdamaD41]

@cb_doge Fingers crossed they all pass the class

[Adama Dia @AdamaD41]

Mark Cuban has delivered a blunt reality check for graduates entering the workforce: AI fluency is now a baseline requirement, not a differentiator.

"If you're not the one who knows how to do vibe coding or work with agents like Claude, somebody else will step in and replace you," he explains.

He argues that forward-thinking managers will enable their teams to leverage these tools and those who don't will soon find themselves obsolete. The same applies to CEOs, Cuban says; any executive who fails to recognize AI's strategic value won't hold their position for long.

In characteristic fashion, he adds a competitive twist: if a company stubbornly clings to such leadership, he invites them to let him know so he can launch a company and beat them at their own game.

[Adama Dia @AdamaD41]

Tony Stark's suit just got a whole lot less fictional. Engineers have built a robotic exosuit that straps itself on in only 10 seconds. Powered by compressed air, soft tubular actuators coil around the wearer's limbs like living vines. Initially designed to support the elderly and those with mobility impairments, the technology is already being eyed for future deployment among rescue teams who often have mere seconds to don protective gear in life-or-death situations.

[Adama Dia @AdamaD41]

Elon Musk has predicted that coding as we know it will become obsolete by the end of this year. He argues that writing code is merely an intermediate step one that AI will soon render unnecessary. According to Musk, AI is already capable of generating far more efficient binary code than any human-written compiler could ever produce.

[Adama Dia @AdamaD41]

According to Kimi CEO Zhilin Yang, artificial intelligence is accelerating faster than ever, entering a new scaling era yet this progress is no longer defined solely by model size. Instead, it is driven by larger context windows, improved training methodologies, reinforcement learning, and reasoning applied during inference. OpenAI's o1 exemplified this pivotal shift, demonstrating AI's capacity to reason, self-correct, and explore multiple potential solutions before delivering an answer. Yang emphasizes that AI is evolving from simple conversational tools into genuine reasoners a transformation that substantially raises the ceiling for what the technology can ultimately achieve.

[Adama Dia @AdamaD41]

One can conceive of the economy as the sum of digital and physical intelligence. While AI is advancing at a remarkable pace, it remains largely confined to the digital domain. To bridge this gap, we need end effectors specifically, humanoid robot that can translate digital intelligence into the physical manipulation of atoms. At its core, an economy consists of producing goods and delivering services. If we deploy vast fleets of robots equipped with immense digital intelligence, we unlock a virtually limitless economy.

[Adama Dia @AdamaD41]

Elon Musk explains why AI and robots are the future of the economy?

[Adama Dia @AdamaD41]

Matrix Robotics officially launches its third-generation humanoid robot MATRIX-3.

MATRIX-3 is a safe, autonomous, and highly generalizable physical intelligence platform capable of performing complex, human-like tasks.

[Adama Dia @AdamaD41]

Everyone says learn AI or get left behind.
Nobody says what to learn. So people learn prompts, which expire, instead of data, which doesn’t.

If the tool you’re learning disappeared tomorrow, what would you still know?

[Adama Dia @AdamaD41]

L’« incohérence » que vous relevez n’en est pas une. Elle est parfaitement logique dès lors que la politique fonctionne comme un marché d’accès plutôt que comme un affrontement de programmes. Soutenir un ministre tout en attaquant le président n’a rien de contradictoire si le calcul porte sur la proximité, la protection, l’opportunité pas sur la doctrine. Les jeunes que vous décrivez ne sont pas confus. Ils jouent correctement au jeu qu’on leur a présenté.
Et c’est là que je nuancerais votre inquiétude. Ce n’est pas une génération qui a mal tourné, c’est une génération qui a observé. Quand les partis n’ont pas de ligne stable, quand les alliances se défont chaque saison, quand le militantisme rapporte plus que la compétence, l’engagement se réduit forcément à un pari personnel. On reproche aux jeunes de faire de la politique une affaire de personnes dans un système où les personnes sont précisément la seule chose qui décide.
La violence verbale suit la même logique. Quand aucune idée n’est vraiment en jeu, il ne reste que l’appartenance et un débat d’appartenance se durcit toujours, parce qu’on ne peut pas céder un argument sans céder son camp. C’est structurel, pas générationnel.
La responsabilité remonte donc vers ceux qui structurent l’offre politique. Des partis avec un programme écrit, des primaires réelles, des sanctions internes pour ceux qui changent de camp au gré des postes : voilà ce qui rendrait la fidélité aux idées rationnelle. Tant que ça n’existera pas, on continuera de former des combattants là où on voulait former une relève.

[Adama Dia @AdamaD41]

The history is accurate up to 1971. The conclusion is where it stops following.
Nixon closing the window is real, and so is the point about proximity to the spigot Cantillon made it in the 1730s, and it holds. But the classical gold standard wasn’t the placid world it looks like from here. Britain had banking panics in 1825, 1837, 1857, 1866, 1873, 1893, 1907. The gold standard didn’t prevent crises. It transmitted them, because a country under stress could only defend the peg by squeezing its own economy which is exactly why almost every government abandoned it the moment abandoning it became cheaper than defending it.
And “no central bank” isn’t the same as no monetary power. Before the Fed, J.P. Morgan personally stopped the 1907 panic by locking bankers in his library until they funded the rescue. The choice was never between concentrated monetary authority and none. It’s between authority that’s publicly accountable and authority that isn’t.
There’s a version of this argument that lands harder in Africa, though. Fourteen countries here already live under a hard external anchor with no independent monetary policy and the record is low inflation alongside chronic underinvestment and no capacity to absorb a commodity shock. Rules-based money has a real cost, and we’ve been paying it for sixty years.
The fight worth having isn’t central bank versus no central bank. It’s who appoints the governors, on what mandate, and whether anyone can see the ledger.

[Adama Dia @AdamaD41]

The frustration is real, and the demographics behind it are real too Africa’s median age is around 19, and no other continent has anything like that. But “they” is where I’d push back, not because nothing is being done to young people, but because the vagueness protects the people actually responsible.
Land, minerals, movement, money every one of those has names attached. A specific concession, signed on a specific date, by a specific ministry, with terms that are either public or deliberately not. When we say “they,” those signatures disappear into a fog, and the person who signed sleeps fine. Anger at a system nobody can name never costs anyone their job.
The other half is harder to say. A lot of what constrains young Africans is decided locally by people who look like us, in our own capitals. The visa regimes that stop a Kenyan working in Nigeria are ours. The land registries that can’t prove who owns what are ours. Blaming only the outside makes the domestic version invisible, which is convenient for whoever benefits from it.
Being young isn’t leadership on its own. It becomes leadership when it gets specific: this contract, this budget line, this official, this demand. Numbers give you the anger. Documents give you the leverage.

[Adama Dia @AdamaD41]

Le point central est solide, et il a un nom en science politique : l’aléa moral. Quand la rébellion devient une voie d’accès aux institutions, on ne règle pas un conflit ,on publie un barème. Le groupe suivant a désormais une indication claire du nombre d’hommes armés nécessaires pour obtenir un ministère. C’est ce que Stedman appelait le dilemme des « spoilers » : chaque accord de partage du pouvoir enseigne quelque chose à ceux qui n’ont pas encore pris les armes.
Je nuancerais toutefois la conclusion. Les données sur les guerres civiles sont têtues : très peu se terminent par une victoire militaire nette, et l’immense majorité des règlements passent par une forme de partage. Le problème n’est donc pas de négocier c’est de confondre deux choses distinctes. Une amnistie, une intégration militaire, un cessez-le-feu : ce sont des instruments techniques. Une place au gouvernement, un poste électif garanti : c’est une récompense politique. La première catégorie achète la paix. La seconde vend la légitimité.
Et votre dernière remarque touche le nœud réel. Quand la revendication porte sur des concessions minières, il ne s’agit plus d’un conflit politique mais d’une négociation commerciale menée avec des armes. Y répondre par des sièges institutionnels revient à confirmer que le contrôle du sous-sol est l’antichambre du pouvoir d’État. Tant que les rentes minières resteront la principale voie d’enrichissement rapide, elles produiront des groupes armés à intervalles réguliers, quel que soit le président en place.
C’est là que se joue le vrai travail de leadership, et il est ingrat : rendre la rébellion non rentable. Cela veut dire une traçabilité réelle des chaînes minières, des registres de concessions publics et vérifiables, des recettes qui remontent au Trésor plutôt qu’aux points de passage. Dans les infrastructures de données, et le constat est constant : ce qui n’est pas enregistré ne peut pas être défendu. Une armée reprend un territoire ; seule une administration empêche qu’on le reprenne.

[Adama Dia @AdamaD41]

L’hypocrisie sociale n’est pas un défaut moral individuel. C’est un arrangement collectif : tout le monde sait, personne ne dit, et chacun compte sur le silence des autres pour maintenir sa propre façade. Le mensonge n’est pas dans les faits, il est dans le contrat tacite qui interdit de les nommer.
Le premier effet, c’est le coût de la façade. Maintenir une apparence coûte cher en argent, en temps, en énergie mentale. Des familles s’endettent pour des cérémonies, des gens vivent au-dessus de leurs moyens, d’autres cachent des difficultés qui pourraient être réglées si elles étaient dites. Toute cette dépense ne produit rien. Elle achète seulement le droit de ne pas être jugé.
Le deuxième effet est plus grave : les problèmes ne se traitent pas, ils s’accumulent. Une société qui interdit de nommer ce qui existe se prive de sa capacité à corriger. La violence domestique, les addictions, l’argent mal acquis, la santé mentale rien de tout cela ne disparaît parce qu’on n’en parle pas. Ça grossit en silence, et ça sort d’un coup, beaucoup plus tard, sous une forme bien plus destructrice.
Le troisième effet est celui qu’on mesure le moins. L’hypocrisie récompense la performance plutôt que la substance. Celui qui gère bien son apparence progresse ; celui qui dit la vérité dérange. À l’échelle d’une famille c’est pénible, à l’échelle d’une institution c’est fatal parce qu’on finit par promouvoir les meilleurs acteurs, pas les meilleurs.
Et c’est précisément là qu’un dirigeant se révèle. Diriger, ce n’est pas exiger la transparence des autres. C’est rendre la vérité moins coûteuse à dire : accepter d’entendre ce qui dérange sans punir celui qui l’apporte. Tant que dire la vérité coûte plus cher que la taire, personne ne la dira et le problème continuera de grandir dans le noir.

[Adama Dia @AdamaD41]

Le récit est juste sur l’essentiel. Une précision utile quand même : Jardine & Matheson étaient des marchands privés, pas la Couronne le monopole de la Compagnie des Indes sur la Chine avait pris fin en 1834. C’est justement ce qui rend l’histoire intéressante. Ce n’est pas un empire qui décide de vendre de la drogue. C’est un lobby commercial qui obtient que la marine de guerre serve ses intérêts. La distinction compte, parce qu’elle se reproduit encore.
Ce que cette guerre montre vraiment, c’est ce qui arrive quand un pays ne contrôle pas les termes de son propre commerce. La Chine avait le produit que tout le monde voulait, un excédent énorme, et elle exigeait d’être payée en argent. Sur le papier, position de force. Dans les faits, elle ne contrôlait ni les routes maritimes, ni la monnaie de règlement, ni les traités. Un excédent commercial ne protège de rien si quelqu’un d’autre écrit les règles du système.
Sur le leadership, Lin Zexu est le personnage le plus instructif du récit. Il fait exactement ce qu’un haut fonctionnaire intègre doit faire : il applique la loi, il détruit la marchandise illégale, il refuse la corruption. Et il perd. Non par faute morale, mais parce que l’intégrité personnelle ne remplace pas la capacité de l’État. Sans marine, sans alliances, sans industrie, la vertu d’un homme se heurte à des canons. C’est une leçon que le continent devrait retenir plus souvent : on ne compense pas un rapport de force par du courage individuel.
Et la mécanique n’a pas disparu, elle a changé de support. Aujourd’hui, la contrainte ne passe plus par les canonnières mais par les systèmes : la monnaie dans laquelle on règle, les rails de paiement, les plateformes où transitent les données. Les infrastructures celui qui les opère peut suspendre une transaction sans tirer un coup de feu. Hong Kong avait un port. La version moderne, c’est un centre de données.

[Adama Dia @AdamaD41]

A pegged currency isn’t automatically a bad thing. Saudi Arabia has held the riyal at 3.75 to the dollar since 1986 and nobody calls it colonial. The difference is in the details, and the details are where the CFA argument actually lives.
Saudi sells oil priced in dollars and pegs to the dollar. Revenue and peg move together, so a swing in the exchange rate doesn’t blow a hole in the budget. West Africa does the reverse cocoa, cotton, oil, gold all price in dollars, while the currency is tied to the euro. When the euro strengthens against the dollar, export earnings shrink in local terms through no fault of anyone in Abidjan or Dakar. That’s not sovereignty lost in some abstract sense. It’s money.
Then there’s who runs it. Saudi’s peg is managed by its own central bank, out of its own reserves, and it can be abandoned by a decision taken in Riyadh. The CFA peg rests on a guarantee from Paris. Same mechanism on paper, completely different question of who decides and that’s the part sixty years hasn’t resolved.
Which makes this a leadership problem more than a monetary one. Running your own peg means holding reserves, taking the political heat when you defend it, and owning the failure if it breaks. The guarantee removes all of that, which is exactly why it’s so hard to give up.

[Adama Dia @AdamaD41]

The pricing observation is correct. Current inference is sold well below its fully loaded cost, and that gap is being funded by capital markets rather than by revenue. Whether it closes through price rises, tiering, or quality degradation, it closes.
But the recommendation doesn’t follow from it. This is the classic platform arc that Parker and Van Alstyne described years before anyone applied it to AI: subsidise adoption, absorb switching costs, then reprice once the user’s workflow has been rebuilt around you. The trap isn’t that the subsidy ends. The trap is that you spend the subsidy period generating output rather than building capability, and when the price moves you have artefacts that have aged and no ability to produce new ones.
Stockpiling generated work assumes the output is the asset. It usually isn’t. A stack of drafts, code and documents decays requirements shift, dependencies break, context changes and you’ll be back at the same terminal paying full price. What doesn’t decay is the surrounding infrastructure: your own data, cleaned and structured; your evaluation sets; your prompt and pipeline logic held in your repository rather than in someone’s web interface. Those retain value regardless of which model you’re pointing at next year.
The leadership version of this is a discipline every organisation should be applying right now, and almost none are. Optimise for exit, not for output. Can you swap providers without rewriting your systems? Is your proprietary data leaving the building or staying in it? Are your people learning to build with these tools or only to operate one vendor’s console? Cheap compute is a window to build portable capability, and a window to build a dependency same window, and the choice is made by whoever is designing the architecture, usually without anyone noticing it was a choice.
I work in data infrastructure, and the practical answer already exists. Open-weight models running on your own hardware are good enough for a large share of real workloads, and they cannot be repriced, deprecated or rate-limited by anyone. The right hedge is an abstraction layer over the model, your data under your control, and enough local capability that the frontier provider is a convenience rather than a dependency.
Use the subsidy. Just don’t spend it accumulating things that expire.

[Adama Dia @AdamaD41]

Worth completing the Liberia story, since it’s being offered as precedent.
The American Colonization Society was never a single project. Some members were abolitionists; many were slaveholders who wanted free Black Americans removed precisely because their existence undermined the case for slavery. Most free Black Americans understood this and refused to go the 1817 Philadelphia convention at Bethel Church rejected the scheme outright, arguing they were Americans and would not be exported from their own country. Of a free Black population in the hundreds of thousands, only about 15,000 ever emigrated over five decades. As policy, it failed.
What it did produce is the part that matters. The settlers arrived and built a state on the model they knew: a settler class governing an indigenous majority that had no vote and no citizenship. Indigenous Liberians were excluded from full political rights until the mid-twentieth century. That arrangement held for 133 years, ended in the 1980 coup, and ran through two civil wars that killed hundreds of thousands. Monrovia is not evidence that population transfer works. It’s a case study in how a state founded on removal reproduces the exclusion it was built from.
There’s a leadership principle underneath this that generalises well beyond 1819. Removal is always the cheapest-looking answer to a difficult political problem, because the costs arrive after the leader who ordered it has gone. Jefferson, Madison and Monroe found colonization attractive for exactly that reason it deferred a reckoning none of them wanted to have. The reckoning came anyway, at a price vastly higher than the one they avoided. Every generation that governs by removing people rather than incorporating them hands its successors a harder version of the same problem.
And the economics rarely survive contact with the numbers. Labour that gets described as trafficked cheap labour is, in the aggregate, a working population that pays taxes, fills roles the domestic labour force doesn’t, and in the West African case sends home remittances that now exceed both foreign direct investment and aid to the region. Removing that population doesn’t just move people. It removes the tax base and the consumer demand attached to them, in economies with ageing demographics that already can’t fill their labour needs.
The 1819 experiment was run. We know how it ended.