Superintelligence Is Now A Networking Product

An article about the path to artificial superintelligence turns out to be sponsored content in which a vendor names its own software as the path.

Sign up for my Substack daily AI newsletter here.

See my AI Training course portfolio for corporate Business Leaders here.

Article link: https://johanosteyn.substack.com/p/superintelligence-is-now-a-networking?r=73gqa&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/

An article appeared recently under the headline “The path to artificial superintelligence.” It carries the masthead of a publication founded at the Massachusetts Institute of Technology in 1899. It is also an advertisement, and the distance between those two facts is where a great deal of South African boardroom thinking about artificial intelligence is currently being formed.

CONTEXT AND BACKGROUND

The piece argues that superintelligence will not arrive through ever-larger single models, but through a horizontal architecture allowing many specialised agents to share intent, context and reasoning. It opens with a healthcare scenario in which agents handling symptom assessment, scheduling, insurance and pharmacy coordinate patient care without a human making the decisions.

The argument is coherent and the writing is good. Every quotation in it comes from Vijoy Pandey, senior vice president and general manager of Outshift by Cisco. Every technology named as the solution is Cisco’s. The page is labelled as sponsored, states that it was produced in partnership with Outshift, carries a closing note that it was produced by the custom content arm rather than by editorial staff, and directs readers to the vendor’s website. Nothing here is concealed. The disclosures are present and accurate. They simply sit below everything most readers will ever see.

INSIGHT AND ANALYSIS

Notice what the framing accomplishes. Superintelligence has been a term for a hypothetical threshold in machine capability, argued over by philosophers and safety researchers for three decades. In this article, it becomes a destination reachable through a coordination layer, which is a category of enterprise networking software which the author’s employer sells. The most consequential idea in the field has been converted into a procurement rationale. The move is not dishonest. It is simply very effective, because a board that accepts superintelligence as the frame has already accepted that the obstacle is architectural rather than, for instance, evidentiary. The evidence points elsewhere.

The one independent study the article cites is a Berkeley paper accepted at NeurIPS 2025, built on more than 1,600 annotated traces across seven multi-agent frameworks, which opens by observing that despite enthusiasm for multi-agent systems, their performance gains on popular benchmarks are often minimal, and which identifies fourteen distinct failure modes clustered into system design issues, inter-agent misalignment and task verification. Read on its own terms, that paper is a caution. Quoted inside a sales argument, it becomes proof that the product is needed.

I have previously written about this in an article on the rise of the digital oligarchy, where I argued that private commercial decisions are increasingly performing a function that looks very much like public policy, and that countries in Africa and much of the Global South have limited influence over how those decisions are made.

IMPLICATIONS

The commercial pressure behind this is not incidental. MIT FutureTech’s expert panel identified competitive dynamics as a risk in its own right, noting that highly competitive development leaves organisations with limited incentive to slow down or invest sufficiently in safety. Urgency is the product being sold alongside the software. The results so far do not support the urgency. EY predicts that more than 40 percent of agentic AI projects will be cancelled by the end of 2027 because of escalating costs, unclear business value or inadequate risk controls. For a South African organisation with constrained capital, the practical discipline is simple and unglamorous. Before the technical evaluation, establish who wrote the document, who paid for it, and what they sell. Then ask whether the independent research cited actually supports the conclusion drawn from it, because in this instance it does not.

CLOSING TAKEAWAY

None of this means the underlying thesis is wrong. Multi-agent coordination may well be the next real constraint, and Cisco may well be building something valuable. The point is narrower. When the word describing the most significant possible development in machine intelligence becomes a reason to buy networking infrastructure, the vocabulary of the debate has been captured, and captured vocabulary is difficult to reason within. Reading the byline before the argument is now a governance competence rather than a scholarly nicety. Somebody paid for that article to reach your desk, and knowing who is the beginning of evaluating it.

Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net