Forecasts
Dated forecasts on AI, risk and markets
Each question carries a deadline, a probability and a confidence rating that reflects the strength of the evidence behind it. Probabilities change only on verified evidence, and every question is scored when it resolves.
- Open questions
- Due in 12 months
- Expected to happen
- Resolved
- Confidence
ForecastsCybersecurity
Will CISA add CVE-2026-114938 (Oracle Java 'Java4Shell') to its Known Exploited Vulnerabilities catalog by 2027-02-28?
Laurenceo estimate: 35% (Unlikely) by February 28, 2027. Last updated November 20, 2026.
How it resolves
Resolves yes if CVE-2026-114938 appears in the CISA KEV catalog on or before 2027-02-28, as shown on cisa.gov.
Reasoning
Set at 35% because a critical unauthenticated remote code execution flaw in the Java runtime that CISA flags for immediate patching is a typical exploitation target, but the alert does not report active exploitation.
Evidence
- For. CISA strongly urged immediate patching of the critical flaw, which could allow unauthenticated remote code execution via crafted Base64 input. CISA (2026-11-17)
- Against. KEV requires evidence of active exploitation. Many critical flaws never reach the catalog, and exploitation depends on applications passing attacker-controlled input to the vulnerable decoder.
Revision log
- November 20, 2026: set at 35%. Set at 35% because a critical unauthenticated remote code execution flaw in the Java runtime that CISA flags for immediate patching is a typical exploitation target, but the alert does not report active exploitation.
Every question
AGI Observatory
When AGI arrives
AGI here means AI that can do the work of an AI researcher, end to end. Our reading is rebuilt every morning from the leading forecasts and a measured trend, and it moves the day a forecaster we count publishes a new date.
Our estimate
Laurenceo's read
The forecasts behind it
Who holds the frontier records
The best published result on four tests of frontier AI, and who set it.
| Test | Best result | Held by |
|---|
Records each company holds now, and records set in the last 12 months.
United States and China, best result on each test
| Test | United States | China |
|---|
Timeline
How the observatory works
Revisions
How expectations changed
Every change is a move from one probability to another, with the reason and the evidence. Nothing is overwritten.
Track record
Are the probabilities any good?
If Laurenceo says 70%, it should be right about 70% of the time, and sharper than a coin flip.
Scores
What we said, what happened
Where the open questions sit
A board with everything near 50% would say little; one with nothing between 5% and 30% would be overconfident.
Method
How the board works
- Probability
- How likely the outcome is by its deadline. In words: Very unlikely (10% or less), Unlikely (11 to 35%), Toss-up (36 to 65%), Likely (66 to 90%), Very likely (over 90%). A new question never opens at an extreme: a very high or very low probability has to be earned by later evidence.
- Confidence
- How much weight the probability can bear, derived only from the evidence and never from the probability: how many sourced facts support the question, how recent they are or how recently they were checked, whether separate publishers and a measured trend agree, and how far away the deadline is. High, Medium or Low. An extreme probability on Low confidence is flagged.
- Movement
- A material change is an update and needs a verified fact and a written reason. The size of a single day's move is capped. Every open question is also reviewed on a schedule, whether or not it is in the news: the estimate is rebuilt from how often that kind of event happens, then adjusted for the latest evidence.
- Evidence
- A fact is stated in a source that has been retrieved and verified, and links to it. Inference is Laurenceo reasoning from facts; speculation is flagged. Every question shows evidence against as well as for.
- Trend test
- Where a question depends on a measured series, several growth models are fitted and weighted by how well the data supports each. The fitted probability is a reference, not the forecast.
- Resolution and scoring
- A question resolves YES on a verified source, and NO after a short grace period past its deadline. Resolved questions are scored with Brier and log scores on a recency-weighted probability, so a late change cannot erase the record. A daily ledger of every probability is stored and cannot be overwritten.