Progress update: a daily inflation index, and the NGX in real terms
Since the launch a few days ago, three things have shipped: a daily inflation-pressure index, a USD- and inflation-adjusted view of the Nigerian stock market, and the first training corpus for the open economic model.
A daily reading on inflation
Nigeria's official inflation number arrives once a month, from the National Bureau of Statistics. Between releases, anyone tracking the economy is working with a figure that can be five weeks old.
The Asotele Inflation Index fills that gap. Every day it combines the cost signals that move faster than the official print — food prices, the official and parallel naira rates, fuel and transport costs, and the inflation tone of financial news — into a single pressure score. It also publishes a provisional CPI nowcast, clearly labelled as provisional: the mapping from these signals to the official rate will be calibrated properly once enough NBS releases are available.
It is a pressure gauge, not a replacement for the official statistic — but it means the daily briefing no longer goes quiet on inflation for weeks at a time.
What the NGX really did
The Nigerian Exchange All-Share Index is trading near 250,000 points — a record, and the kind of number that makes headlines.
Adjust for the naira and the picture changes. Over the past five years the All-Share Index is up around 550% in naira terms. Measured in US dollars, it is up about 94%. Adjusted for inflation, over the decade to the end of 2024 the index lost roughly 44% of its real value — even as the nominal figure nearly tripled.
None of that means the market is failing. It means a naira figure, in an economy with this much inflation and currency depreciation, mixes real performance with currency debasement. Asotele now reports the All-Share Index three ways — nominal, in dollars, and inflation-adjusted — so the headline number is read in context. This is the purpose of the project in miniature: the same data, seen clearly.
Building the model's corpus
Asotele's longer-term goal is an open, fine-tuned language model specialised in Nigerian and African emerging-market economics. That model needs a training corpus, and this week the corpus took shape.
The system now turns its own daily briefings into structured training examples, and — using four decades of backfilled history, with Brent crude back to 1987 and exchange rates and inflation back to the 1960s — generates more than 600 grounded historical analyses covering real episodes: the 1995 inflation peak, the 2016 recession, the 2020 oil crash, the 2023 naira float.
One design choice matters here: every training example is generated deterministically from real data. No language model writes the corpus — each figure is computed directly from the source series. A model is only as trustworthy as the data it learns from, and a corpus that fabricates nothing is the first line of defence against a model that hallucinates.
Next
The corpus grows daily. The remaining pieces — real institutional documents, multilingual coverage, and the first fine-tuning run — come next. Updates will land here.