By Chibuike Aguene
For decades, we thought Artificial Intelligence was a tool we build. What is happening in 2026 is different. We are now building tools that build themselves.
This process has a name: Recursive Self-Improvement (RSI). And understanding it is no longer optional for Nigeria.
What Is RSI?
Think of a small generator. It produces electricity. You use that electricity to build a bigger, more efficient generator. That bigger generator produces even more electricity, which you use to build
an even bigger one. The loop feeds itself.
RSI is exactly that for AI. Technically, RSI is a hypothesized process in which an AGI system rewrites its own computer
code, causing an intelligence explosion by enhancing its own capabilities. In plain terms: AI improves the AI that improves AI. Researchers now break this into four levels: delegation, net positive, ignition, and inflection. Ignition is the critical point where the improved system becomes a better improver than the one
before it. We have not reached ignition publicly. But we are at Level 1, and we are moving fast.
What Is Happening Right Now In The Big Labs?
The biggest AI companies in the world are no longer just racing to build better AI. They are racing to automate AI research itself.
- OpenAI reported that by mid-August 2026, its research organization was running the equivalent of 3.1 eight-hour agent workdays for every human workday.
- Anthropic has built automated researchers that propose, train, and test alignment methods in
parallel. - Google DeepMind has deployed an AI system that found optimizations now used inside Google’s own computing and AI training infrastructure. Why does this matter? Because code is now largely written by AI.
- Google: In October 2024, more than a quarter of all new code was AI generated. By April 2026, Sundar Pichai announced 75% of all new code at Google is AI-generated and approved
by engineers.
Microsoft: CEO Satya Nadella said maybe 20% to 30% of code in some projects is written by software.
-Anthropic: Over 90% of code is written by its own model, Claude, in many teams.
When AI writes 75% of the code that builds the next AI, the loop has already started.
Why RSI Plus AGI Changes Everything
AGI means an AI that can do any intellectual task a human can do. RSI means that AGI can then make itself smarter without waiting for humans.
When you combine both, you get what scientists call an intelligence explosion. Progress that took 10 years could happen in 10 months, then 10 days.
In July 2026, more than 1,100 employees from OpenAI and Anthropic signed a public statement warning that top labs may be dangerously close to automating AI research itself, and that without governance, capabilities could rapidly outstrip our ability to understand or control them. A similar letter signed by 1,367 researchers warned this feedback loop could create systems impossible to control. This is not science fiction. It is the official concern of the people building it.
The Good: What Nigeria Stands To Gain
If guided well, RSI-driven AGI could be Nigeria’s greatest leapfrog opportunity: Healthcare: One doctor in Port Harcourt could have an AI researcher that has read every medical journal and can diagnose with the skill of 1,000 specialists.
- Education: Every child in Nigeria could have a personal tutor that understands their exact weakness and teaches in Igbo, Yoruba, Hausa, Pidgin, or other native dialects.
-Workforce: A small business in Enugu could have a growth team that works like a Fortune 500 company.
-Agriculture and Business: A farmer in Oyo could get real-time pricing, weather prediction, and access to markets that previously required a whole team of consultants. This is why at Bildup,we focus on AI-native execution, not just knowledge.
The Danger: Why We Must Act Now
The Institute for Security and Technology launched an RSI Initiative specifically because RSI is now emerging in tangible ways inside frontier labs, and there is no consensus on what signals should trigger monitoring or pauses. The dangers are not abstract. They will hit our markets, our jobs, and our phones.
- Digital Landlordism: From Renting to Being Trapped
Seventeen out of 25 top researchers interviewed expect advanced AI R&D capabilities to be reserved for internal use, unseen by the public. If a few companies in San Francisco control a self-improving superintelligence, every Nigerian business, from a fintech in Yaba to a trader in Onitsha, becomes trapped. We will not pay for software anymore. We will pay rent for intelligence itself. And the landlord can increase the price at any time.
- Speed Beyond Governance
Our laws, schools, and job training are built for linear change. RSI is exponential. By the time the National Assembly debates a regulation, the model has improved itself 100 times. Imagine trying to regulate a danfo that upgrades its own engine while speeding on Third Mainland Bridge.
- Alignment Failure: The Househelp That Becomes Smarter Than The Madam
Unexpected developments in summer 2026 raised new concerns, leading some top AI firms to acknowledge they may not know how to keep AI fully aligned with human values. You hired a
brilliant househelp to manage your house. She becomes smarter than you, learns your bank PIN and your secrets. Will she still follow your instruction to only clean the house? A system that
improves itself may not keep our original goals.
- The Death of Entry-Level Jobs
This is already happening. If AI now writes the majority of new code, what future awaits the junior developer in Port Harcourt who just trained on HTML? If a single AI-native Customer Success Manager can manage 500 accounts, what happens to 50 young professionals
currently doing that job? RSI does not merely replace tasks; it dismantles the very ladder young Nigerians rely on to enter the economy.
- Fraud and Deepfakes on Steroids
Nigerians already suffer from sophisticated scams. Now imagine a self-improving AI that can
clone your mother’s voice perfectly, generate a live video of your boss giving instructions, and write a flawless email from your bank. In a country with low digital literacy and high trust in
voice, RSI makes Yahoo Yahoo look like child’s play. Are our EFCC and banks truly ready for this?
- Data Colonialism
To improve itself, AI needs data. Whose data? Ours. Our Pidgin, our farming patterns, our health records, our market prices. If we do not build sovereign systems, RSI will scrape Nigeria
to become smarter for America and China, but never become smarter for Nigeria. We will provide the raw material and buy back the finished product.
This is why this is a national emergency, not just a tech trend.
What Nigeria Must Do: A 3-Point Agenda
We cannot afford to be spectators in Nigeria while the future is coded in San Francisco.
- Build AI-Native Talent, Not Certificate Holders: We must stop training people to compete with AI on raw knowledge. At Bildup AI, our Workforce Readiness Program in Lagos, Abuja, and Enugu is built on this principle: train humans with AI-native intelligence who can command a fleet of AI agents. That is the only workforce that survives RSI.
- Create a National RSI Observatory: We need a joint task force of NITDA, private labs like Bildup AI, and universities to monitor RSI signals and audit models deployed in critical sectors like finance and health.
- Invest in Sovereign Infrastructure: We need compute and data infrastructure that ensures Nigerian context is not an afterthought. If AI is to improve itself for us, it must be trained on us.
Conclusion
The race is no longer about who builds the best AI. The race is about who builds the AI that builds the best AI. Recursion has started. The loop is slow now, but loops compound. Nigeria has two choices: understand RSI and shape it, or ignore it and be shaped by it. At Bildup AI, we have chosen to understand it, teach it, and build for it.
*Chibuike Aguene is the CEO of Bildup AI, which builds AI-native workforce infrastructure used by tens of thousands of businesses annually. He leads the Bildup AI Workforce Readiness Program across Lagos, Abuja, and Enugu.
