Predictions

A set of stochastic predictions about security, tech, and society

This is where I maintain a list of my predictions about technology and societyβ€”along with their accuracy over time.

I do this because I'm obsessed with improving my model for how the world works, and there's no better test than making clear predictions and watching them fall apart over time (or not).

The goal is to find systemic errors in my thinking that I can then correct.


Predictions Quick View ​

Here's a short view of the predictions in table format.

PredictionConf %Date PredictedStatus
Recession-like shock caused by AI job loss by 2027Chances About EvenJuly 2025πŸ”„
Russia will significantly return to normal trading status by 2027Chances About EvenMarch 2023πŸ”„
We'll have AGI betewen 2025 and 2028Almost CertainMarch 2023πŸ”„
Open-source models will largely catch up to closed-source modelsProbableFebruary 2023πŸ”„
Apple is about to go from the worst AI implementation to the best.ProbableJanuary 2025πŸ”„
Everyone, including people, will have an APIAlmost CertainDecember 2016πŸ”„
Personal daemons will broadcast preferencesAlmost CertainMarch 2014πŸ”„
Venues personalize based on customer preferencesAlmost CertainMarch 2014πŸ”„
People will lack meaning and look for it in gamesAlmost CertainJuly 2006πŸ”„
Masive jumps in AI will be made through slack-in-the-rope advancements, i.e., "tricks"Almost CertainAugust 2024βœ…
Extreme liberals will get Trump re-elected in 2024Almost CertainMid 2020βœ…
Trump will officially end the war in Ukraine by April 1, 2025LikelyMarch 2023❌
People will return to Twitter within 6 monthsAlmost CertainDecember 2022❌

Correct Predictions ​

PredictionConf %Date PredictedStatus
Extreme liberals will get Trump re-elected in 2024Almost CertainMid 2020βœ…
Masive jumps in AI will be made through slack-in-the-rope advancements, i.e., "tricks"Almost CertainAugust 2024βœ…

Missed Predictions ​

PredictionConf %Date PredictedStatus
Trump will officially end the war in Ukraine by April 1, 2025LikelyMarch 2023❌
People will return to Twitter within 6 monthsAlmost CertainDecember 2022❌

Detailed Predictions ​

Recession-like shock caused by AI job loss by 2027

Status: In Progress πŸ”„
Confidence: Chances About Even
Date Made: July 2025
Estimated Date: Before 2027
Reference: Post
Where Said: Blog post
Quote: "AI will cause significant job displacement leading to economic shock"

We'll have AGI between 2025 and 2028

Status: In Progress πŸ”„
Confidence: Almost Certain
Date Made: March 2023
Estimated Date: 2025-2028
Reference: Blog Post
Where Said: Blog
Quote: "We're not waiting for a single model with the general flexibility/capability of an average worker. We're waiting for a single AGI system that can do that."

Open-source models will largely catch up to closed-source models

Status: In Progress πŸ”„
Confidence: Probable
Date Made: February 2023
Estimated Date: 2025
Reference: Verbal
Where Said: Conversation/Interview
Quote: "The gap between open and closed models will significantly narrow"

Apple is about to go from the worst AI implementation to the best

Status: In Progress πŸ”„
Confidence: Probable
Date Made: January 2025
Estimated Date: April 2025
Reference: Newsletter
Where Said: Newsletter
Quote: "Apple will leapfrog from worst to best AI implementation"

Russia will significantly return to normal trading status by 2027

Status: In Progress πŸ”„
Confidence: Chances About Even
Date Made: March 2023
Estimated Date: 2027
Reference: Post
Where Said: Blog post
Quote: "Russia will normalize trade relations within 4 years"

Everyone, including people, will have an API

Status: In Progress πŸ”„
Confidence: Almost Certain
Date Made: December 2016
Estimated Date: Open
Reference: Book
Where Said: Book publication
Quote: "Personal APIs will become as common as social media profiles"

Personal daemons will broadcast preferences

Status: In Progress πŸ”„
Confidence: Almost Certain
Date Made: March 2014
Estimated Date: Open
Reference: Blog Post
Where Said: Blog
Quote: "These are sets of preferences, attributes, and interests that are 'broadcasted' to others everywhere you go."

Venues personalize based on customer preferences

Status: In Progress πŸ”„
Confidence: Almost Certain
Date Made: March 2014
Estimated Date: Open
Reference: Blog Post
Where Said: Blog
Quote: "You could walk into a sports bar and have all the TVs turn to MMA because the restaurant's daemon knows you like it."

People will lack meaning and look for it in games

Status: In Progress πŸ”„
Confidence: Almost Certain
Date Made: July 2006
Estimated Date: Open
Reference: Blog Post
Where Said: Blog
Quote: "So, the whole idea of the world is to substitute for the failing one that exists in reality. The one where most people are weak and powerless."

Masive jumps in AI will be made through slack-in-the-rope advancements, i.e., "tricks"

Status: Confirmed βœ…
Confidence: Almost Certain
Date Made: August 2024
Estimated Date: 2024-2030
Reference: Essay
Where Said: Blog post
Quote: "Major AI breakthroughs will come from clever techniques rather than pure scale"

Extreme liberals will get Trump re-elected in 2024

Status: Confirmed βœ…
Confidence: Almost Certain
Date Made: Mid 2020
Estimated Date: November 2024
Reference: Blog Post
Where Said: Blog
Quote: "Extreme progressive positions will alienate moderate voters and hand Trump victory"

Trump will officially end the war in Ukraine by April 1, 2025

Status: Miss ❌
Confidence: Likely
Date Made: March 2023
Estimated Date: April 1, 2025
Reference: Post
Where Said: Blog post
Quote: "Trump will quickly negotiate an end to the Ukraine conflict"

People will return to Twitter within 6 months

Status: Miss ❌
Confidence: Almost Certain
Date Made: December 2022
Estimated Date: Mid 2023
Reference: Multiple
Where Said: Various conversations
Quote: "Elon will fix Twitter and people will come back quickly"

Notes

  1. Once I make a prediction here I will not materially change it. The whole point of this is to lock it in so I can see my mistakes and tally my record over time.
  2. AI is great for projects like this because you can feed your whole list to a model and have it tell you the biases that are causing your errors. I do this frequently within my TELOS file and it's quite powerful.
  3. I have omitted a number of predictions I've made that came true just because they seem so obvious at this point, e.g., how deepfakes would make it so people can't tell reality from fiction. They just don't seem worth mentioning at this point.