404_NOT_FOUND

Following my initial analysis, I wanted to check whether, conditional on occurring before the end of July, a hurricane has a lower than the baseline 43% chance to be classified C3 or higher.

Copying the approach of @alter_hugo here, I checked the list of C3, C4, and C5 events in the past 50 years of the historical record, and reported below every time they happened before the end of July.

Hurricane	Category	Date
Bertha	C3	July 1996
Bertha	C3	July 2008
Dennis	C4	July 2005
Emily	C5	July 2005
Beryl	C5	July 2024

It only happened in 5 of the past 50 years, meaning that the average Jun/Jul hurricane is significantly weaker than the average hurricane in a season. This is likely due to lower ocean temperatures at the beginning of the season.

I'm still conservatively forecasting higher than the empirically measured baseline, as the available data is very sparse.

As correctly highlighted by @ctsats here, 2 early major hurricanes happened in 2005. These two events were likely correlated in some way, so the base rate is actually even lower than 1/10.

Historical data from the past 20+ years (obtained from Wikipedia).

Month	Avg # hurricanes
Jun	0.25
Jul	0.9

Approximately 43% of hurricanes are classified as C3 or higher, so the expected number of C3+ hurricanes is 0.49

The forecast can be modeled using a Poisson distribution.

Judgmental adjustments can typically be performed when information about ENSO is available, but current predictions are that we are shifting to a neutral phase of the oscillation, implying an average number of hurricanes is to be expected. [1]

[1] ENSO: Recent Evolution, Current Status and Predictions

You can download the data from here.

I've random-walked the most recent reading (Jan 29th = 13.47) all the way through the end of March. For each step, I naively used the average and spread of the daily change for that day in the ice extent from the past 25 years and iterated 5k walks. 

The result is that the most likely outcome is an average ice extent of 14.17 in March 2026, with only a 32% chance that it will set a new record minimum. 

Danger zone! I've also vibe-coded a SARIMA model that outputted a more optimistic 14.4 forecast, with just a 15% chance for a new March minimum. It's becoming increasingly convenient to use AI for the heavy-lifting on questions like these, so I'm experimenting a little bit. 

I hypothesize that the path of the 2025 curve is an outlier, with significant ice loss much earlier than the spring equinox, which is why modeling with historical data suggests a repeat is unlikely.

Below is 2025 vs the 2010-2020 average

As I understand, the surge in funding that happened in January is typical.  At these events, startups pitch and deals are announced. Expect higher than normal activity through March.

I don't think the crowd reacted vigorously enough at the $88 million seeded in this first month. The total so far is $253 M, and the average monthly funding in 2025 was $13.7 M, making it significantly less likely that the outcome will fall in the first bin.

A recent paper from the European Council on Foreign Relations concludes that:

It will take Russia some 5-10 years after the end of the war in Ukraine to refit and rearm for such an attack. To ensure Estonia’s security, the country and its allies need to continue developing their defence capability now and in the coming years—even if they have minimal assistance from the US. [1]

The paper examines two scenarios: outright invasion or hybrid in-and-out campaign, with the latter being more likely.

Far more plausible than Russia’s direct invasion of Estonia is a hybrid attack: a rapid, deniable operation blending local proxies, cyber sabotage and limited Russian incursions. The goal would be political shock, not territorial gain: to create confusion, delay NATO’s decision-making and demonstrate Western impotence and disunity.

One critical piece of information is that the paper claims that European NATO countries can maintain a credible deterrence even with minimal US assistance, and despite the perceived unreliability of the American partner

Two interesting forecast-related items:

• A 2016 RAND wargame concluded that Russia could seize Tallinn within 60 hours of an invasion, which the Ukraine conflict has obviously demonstrated is not the case. Right at the beginning of the invasion of Ukraine, there were claims that Putin expected to capture Kyiv within 48 hours. It's plausible that Russian expectations were off not purely because of bad intelligence or a yes-man culture in the military, but because the Western perception of Russian capabilities was also exaggerated. This could suggest that Russia will be more careful than it has been with Ukraine.
• There's a linked article containing aggregated forecasts for: 1) When Russia will have the capability to attack; and 2) When Russia might attack. [2]

2027 and 2028 are highlighted as the two years in which Russia would be most likely to attack a NATO country. Yet, many of those forecasts date back to 2024 (and even earlier), so there might have been some expectation that the ongoing conflict would have already been halted by now. 

[1] The bear in the Baltics: Reassessing the Russian threat in Estonia

[2] Would Russia Attack NATO and, If So, When?

Initial forecast based on the fact that the number of successful launches each month has been remarkably consistent through 2025 and SpaceX appears to be committed to continuing to launch at a similar pace.

I've used the normal distribution of the average/spread of monthly launches to build my distribution through the bins.

Here is a PDF with easily accessible cumulative monthly snowfalls. Data should be similar enough.

I'm starting with an empirical distribution that used the past 20 years of data, and I'm factoring in a high likelihood of about 1 inch of snow forecasted during the first few days of February.