A Circuit Named Bandung
Building AI Tools for the South, by the South
If you have been following this column, you may have noticed that the rhythm of updates has slowed in recent weeks. The reason is not laziness. I have been busy with a new project — one that I want to tell you about. At Tricontinental: Institute for Social Research, we have just launched a new column called Bandung Circuit.
The Spirit of Bandung
In April 1955, twenty-nine Asian and African nations — representing more than half the world’s population — gathered in Bandung, Indonesia. They declared the right of formerly colonised peoples to chart their own political and economic course. Non-alignment, self-determination, anti-colonialism: these principles came to be known as the Spirit of Bandung.
Seventy years on, political independence has largely been achieved. Knowledge production is another matter. Researchers across the Global South still rely on databases owned by Northern institutions, analytical tools designed for Northern budgets, and publication systems that gatekeep whose questions count as legitimate. The Spirit of Bandung needs to be rekindled — not in the political arena, where the battle was fought and largely won, but in the infrastructure of knowledge itself.
That is why we named it Bandung Circuit — applying the same principle to knowledge production, building AI research tools that serve scholars and movements in the Global South.
What the Circuit Carries
Bandung Circuit is organised into four sections.
Foundations lays out the theory, architecture, and design principles behind AI-enabled research infrastructure — articles on epistemic sovereignty, declarative multi-agent architectures, and knowledge engineering.
Currents features perspectives from researchers and practitioners on how new computational methods are changing what social science can ask and answer.
Reports presents full-length research produced with AI-assisted methods — where analytical depth meets political questions that matter.
And How-To Guides offers practical tutorials, written so that anyone, regardless of technical background, can follow along.
The point is not to host another academic discussion about AI. It is to build a complete ecosystem — from theory to practice to tools to published research — that puts knowledge production in the hands of the South.
Two stories from the Currents section give a sense of what that means in practice.
The economist John Ross wanted to study the relationship between investment and growth across all 210 of the world’s economies — a systematic comparison nobody had attempted. Using GSI (Global South Insights), a platform built by Tricontinental that unifies 96 international datasets comprising 41,100 indicators and 3.45 billion rows of data, he found a correlation of 0.95 between net fixed capital formation and growth among the ten largest economies — a figure virtually unheard of in real-world economic data. Calculations that once took thirty to forty hours per indicator now execute in seconds.
Then there is the qualitative side: a seventeen-chapter restorationist history of the World Anti-Fascist War, drawing on hundreds of citations in five languages, built on a knowledge base constructed for less than two hundred dollars. When the Global South has sovereign access to the right tools, research that was impossible becomes routine.
The Foundations section provides the theoretical backbone — reclaiming epistemic sovereignty, designing multi-agent research systems, rethinking what it means to do “knowledge engineering” rather than mere prompt engineering. And the How-To Guides bring it all down to earth: step-by-step instructions for setting up an AI workspace, installing and configuring coding assistants, accessible to anyone with a laptop and an internet connection.
The Hidden Wave and the Open Door
There is a deeper reason I have been pouring energy into this project, and it has to do with something I have been observing for years.
In southwestern China, a natural gas station worker lives on-site for eighty-four hours at a stretch. His daily work involves processing information — inspecting valves, photographing equipment, filling safety reports, formatting data for headquarters. The software he uses was not designed for him. It does not match his workflow. But the thought of building something better has never occurred to him. Not because he considered it and rejected it — because it never entered his mind at all. “I don’t have needs,” he once said. “I can’t say what my needs are.”
This silence was not born; it was made. Over four decades, the software industry systematically closed the passage from using information tools to creating them. Six locks, installed by different forces at different times, all on the same door. The deepest lock is not the price of tools or the difficulty of skills — it is the elimination of the very idea from language. The thought “I could build a tool for myself” simply stopped occurring to people.
AI has reopened that door. But there is a pattern I have been tracking for years: every visible wave of democratisation carries a hidden wave of monopolisation. Each technology that made computing more accessible simultaneously made creating computing tools less accessible — not through conspiracy, but as a structural byproduct of progress itself. If that pattern holds, AI will repeat it unless alternative infrastructure is deliberately built.
Bandung Circuit is that deliberate act of building — not just diagnosing the problem, but constructing the alternative.
Think of what it means in concrete terms: a knowledge base that once took months to build can now be assembled in days for less than two hundred dollars. A local researcher in Accra or La Paz can interrogate the same data, at the same scale, as a team at Harvard. The possibility of building one’s own information tools — a thought that had been erased from consciousness — is becoming real again. Bandung Circuit exists to make that possibility systematic, scalable, and sustainable for the Global South.
Not a Farewell
What does this mean for this Substack column? My primary focus will shift to Bandung Circuit, and the update rhythm here will slow from roughly weekly to something less frequent.
But this column is not going anywhere. When I see something interesting in China’s AI landscape — a new application, an unexpected policy move, a grassroots innovation — I will write about it here, just as I have done for the past thirty articles. I live in China, I work in China, and that vantage point does not change.
The two columns are complementary. This one watches what is happening in China. Bandung Circuit builds tools for the Global South to produce knowledge on its own terms. Two sides of the same coin: China’s experience provides reference points; Bandung Circuit provides the infrastructure to act on them.
Bandung Circuit is at thetricontinental.org/bandung-circuit. I will see you there — and I will see you here.