The Signal Ledger: 2026, Read Backwards

In one minute:

By January 2026 the main signals were already on the table: cheap oil that could reverse, a stalled easing cycle, and health systems losing funders.

Brent rose from about $61 at the start of the year to $118 by the end of March, and Kenyan pump prices followed at about 0.43 of the percentage move.

We registered twelve dated forecasts today, 16 October 2026, so the Brier score can judge them in January 2027.

12 January, and a ledger with no entries yet

On 12 January 2026 the year had not decided what it would be. Brent sat in the low $60s, the kind of price that makes a fill-up feel like a small mercy. Central banks still talked about cuts. Nobody in a Nairobi matatu was thinking about a strait.

This is a backcast. We stand on 12 January, ask what was already visible, and then show what happened. We did not publish these readings in January. Anything we say about the future points to our registered forecasts, F1 to F12, which we published today and will score.

What was visible

The year rested on five forces. Each one was already moving before the year turned.

• Energy. Oil was rebuilding a risk premium over Iran. Brent closed at $70.69 on 30 January, up from $63.76 on 15 January.
• Rates. Disinflation had stalled. US headline CPI was 2.7% in December, and the Federal Reserve held on 28 January with two votes for a cut. BLS Federal Reserve Cuts were being pushed out, not pulled forward.
• Capital. Frontier AI money was gathering in a few labs, and that concentration kept growing through the spring.
• Health. The United States formally left the World Health Organization on 22 January, and aid budgets were shrinking. Funding gaps tend to show up in outcomes later, not on the day they are cut.
• Climate. El Niño was building, and the risk to East African rain and disease rose with it.

Read together, the forces make one sentence. An energy shock turned a year of expected rate cuts into a year of hikes, while health defences thinned and climate stress rose in the same region. The people who felt it first were those with the least slack in their household budgets.

What happened

The table lists the readings that carry the story. Each number links to its source.

Table sources: Anthropic, EIA, TechCrunch, WHO, CNBC, WHO, FRED.

The Brent monthly average moved from $70.89 in February to $103.84 in May (FRED).

The math

The Brier score measures a probability forecast against what happened. For each question, set the outcome to 1 if the event happens and 0 if it does not. The score for one forecast is (forecast − outcome)². Average it over all your questions.

Take a forecast of 0.7, that is, 70% confidence.

• If the event happens: (0.7 − 1)² = (−0.3)² = 0.09.
• If it does not happen: (0.7 − 0)² = 0.7² = 0.49.
• A forecaster who always says 50% scores (0.5 − 1)² = 0.25 on a yes and (0.5 − 0)² = 0.25 on a no. The score is 0.25 either way.

Lower is better. The 0.25 coin-flip baseline is the bar. A forecaster claims skill only by scoring clearly below it. The lesson is simple: a 70% call that goes wrong costs you 0.49, and a 70% call that goes right earns only 0.09. The score is what keeps a forecaster honest.

The Pawanax lens

For a clinic, the signal that matters most is distribution. In July, WHO said 80% of eastern DRC Ebola cases came from unknown transmission chains, which means the disease spread where nobody was counting. A clinic that reports and triages through a phone, in a language its patients already use, is public-health infrastructure, not a feature. Pawanax is building toward that layer.

For families, the cost squeeze is the story. Kenya's CPI reached 6.8% in September (KNBS), and transport prices were up 15.6% on the year. When a household has less slack, it chooses which visit to skip. For health systems, debt service took 77.6% of tax revenue in the first quarter of the 2026/27 fiscal year (US News), so clinics will buy what saves money: fewer missed referrals, faster claims, less wasted stock.

The next 90 days

Three registered forecasts fall due first, and each will be scored with the Brier rule above:

• F5: 55%. Kenya's December 2026 CPI, year on year, is at least 7.0%. Resolves 5 January 2027.
• F8: 60%. The FAO Food Price Index for December 2026 is above 136.0. Resolves 8 January 2027.
• F1: 65%. Brent settles above $90 on 15 January 2027.

Our longer bet is F6: 80%, that the WHO Bundibugyo Ebola emergency is still in force on 31 January 2027. Every forecast in the full table below resolves to 1 or 0, and each score will be published whether it looks good or not.

The registered forecasts, F1–F12 (published 2026-10-16):

Scoring: Brier = mean((p − outcome)²). A forecaster who says 50% on everything scores 0.25, so we only claim skill if our score is clearly below that. The results get published either way.

The ten chapters

• 01: The Quiet Before the Price. January's signals, and how a 10% prior became 25% after one good signal.
• 02: Cheap Oil Is a Loaded Spring. Why a calm February was not a safe February, and what fat tails mean.
• 03: When a Strait Becomes a Price. March's shock, and the 0.43 pass-through from crude to Kenyan pumps.
• 04: Markets Trade Headlines. April's 9–13% daily swings, and what a headline is worth in expectation.
• 05: The Emergency Nobody Priced. Ebola's growth in exponential terms, and how doubling time changes.
• 06: The Sixty-Day Peace. A time-limited deal, and the hazard rate that prices its collapse.
• 07: Unknown Chains. Why 80% unknown transmission changes the odds of detecting an outbreak.
• 08: The Trillion-Dollar Circle. AI financing that suppliers fund, and the concentration risk it hides.
• 09: The Turn. Real rates, bond duration, and why Kenya keeps KSh 22.40 of every KSh 100 in tax.
• 10: Where the Signals Point Now. Turning stories into calibrated probabilities, and the forecasts that test them.

Part 0 of The Signal Ledger. Read the series: [The Signal Ledger](/blog/signal-ledger-2026-read-backwards).

Cover image: The Blue Marble 4463x4163.jpg — NASA. Photo taken by either Harrison Schmitt or Ron Evans (of the Apollo 17 crew). · Public domain. Photographed by the Apollo 17 crew on December 7, 1972, from about 29,400 km away while traveling to the Moon; this is an archival photograph, not a current view of Earth's surface or weather. Resized for web display.