In the race to build larger AI models, one weakness remains stubbornly unresolved: memory. Dubai & London-based SeKondBrain is betting that the next frontier in AI will be systems that preserve context, meaning and knowledge over time.
That matters because most skilled work is cumulative. An electrician diagnoses a fault faster because she has seen a similar pattern before. A writer improves through successive drafts. A chef sharpens instinct through repetition. Knowledge has value on its own, but real expertise comes from what is retained, refined and carried forward.
Today’s AI systems are still poor at that kind of continuity. Many leading systems try to compensate by processing ever-larger volumes of raw text, known as tokens, in pursuit of better outputs. By expanding their context windows, these models can read more, summarize more and generate more than ever before. But when it comes to preserving understanding over time, they still falter.
McKinsey estimates that employees spend one-fifth of their time searching for, interpreting, or recreating information. Meanwhile, research cited by Stanford’s Institute for Human-Centered AI suggests that as context windows expand, consistency and reasoning can degrade.
AI was supposed to reduce these frictions, not reinforce them. Yet in many cases, it still accelerates fragmented work rather than fixing it.
SeKondBrain AI is tackling this problem directly, with a focus on preserving meaning and memory. “Tokens are sand. Concepts are Lego,” says Duggal, by which he means that without conceptual structure, raw information slips through our fingers.
That thesis places SeKondBrain within a broader shift toward neurosymbolic AI. While purely generative systems are powerful at producing text, they often struggle with explainability, structured reasoning, and durable memory. Neurosymbolic systems address that gap by combining neural pattern recognition with symbolic reasoning. The field is rapidly gaining momentum, with the global neurosymbolic AI market projected to reach $4.87 billion by 2034 and foundation capital pitting the context layer as a 1 trillion-dollar market.
Founded less than a year ago by entrepreneur Sachin Duggal, SeKondBrain is building what it calls a “cognitive operating system” – a shared intelligence layer designed to sit beneath people, enterprise software, and AI agents.
Modern AI systems, in Duggal’s view, are increasingly powerful engines with an obvious missing layer – “autonomous cars without a map.” The car may be fast and its sensors impressive. Butwithout a stable map of the world – what matters, what’s changing, what’s allowed, and so on – it’s still can’t get out of the driveway.
In contrast, SeKondBrain’s platform expresses itself through interconnected surfaces. Kora functions as a conversational intelligence layer, preserving context across conversations, teams, and calendars. Kanvas serves as a workspace where documents, ideas, and digital artifacts are connected by concept rather than location. And the Kemory allows AI to have a shared contextual memory so you don’t need to copy and paste between your chats.
For businesses there is also KompanyBrain; a single pane of glass to bring all your organisational & customer intelligence and knowledge together with early design customers in Consultative Industries like Legal and Consulting.
Early test cases were intentionally practical. One is an “Experience Finder” – a system that can answer questions like, “Who inside the organisation has solved this problem before?” or “Which team has worked on this client, sector or regulatory issue?” and the other was for law firms to simply have a single pane of glass across client workflow, files and communications.
SeKondBrain sees these systems evolving into organisation-specific knowledge layers for sectors including law, consulting, finance and government, where expertise is often buried across documents, meetings, emails and internal conversations. This would also support faster software development, since new tools no longer need to worry about the messy data layer causing information dementia in modern day companies.
That ambition also helps explain why Duggal is building from Dubai. A system designed to connect fragmented knowledge across teams, sectors and markets needs more than technical talent. It needs proximity to global businesses, government buyers, capital, and commercial corridors where ideas can move quickly from pilot to deployment.
In one sense, the choice extends Dubai’s historic role as an entrepôt: a place where goods, capital and talent move through, are refined, and then exported outward. SeKondBrain applies that logic to knowledge infrastructure. Built from the Dubai corridor to London, refined with a team that convenes in the emirate each month, and designed to ship globally, it treats Dubai less as a market to sell into than as a platform from which to build.
But it also reflects a deeper shift. For decades, advanced technology was assumed to be built elsewhere – in Silicon Valley, London or Beijing – and then sold into the Gulf. That assumption is now being challenged. The UAE is no longer just a destination market for global technology; it is increasingly a place where globally relevant companies are built.
The economic momentum is significant. The UAE’s AI market is projected to reach $46 billion by 2030, with AI expected to contribute close to 14 percent of national GDP. But beyond the economics lies a more consequential question: who will control the underlying architecture of intelligence itself?
The last era of the internet was shaped by those who owned the servers, platforms and cloud infrastructure. The next may be shaped by those who own the systems that preserve memory, meaning and shared cognition.
SeKondBrain’s wager is that part of that foundational layer can be built in Arabic and English, developed from Dubai, and exported to the world.
It’s a bold bet. But in a crowded AI market, the ability to make intelligence persist – to remember, adapt and compound over time – may prove to be one of the most valuable opportunities of all; because as Duggal says “models commodities, knowledge compounds”
