As we roll into the end of 2025 I think it'd be good to have a thread in the rationalist tradition of putting down predictions with confidence ratings for 2026. Feel free to take the my list and make your own predictions off them and/or add new things to predict and I'll add them to the op with my own prediction. I'll put my predictions in spoilers at the end. Feel free to critique other people's predictions but for fairness if you're going to say someone's prediction is wrong you should supply your own odds as well. Of course let me know if any of these categories are under defined.
By December 31st 2026
AI:
- Yudkowski declares AGI/ASI Achieved
- We have federal regulation passed through congress to regulate AI
- Anthropic announces IPO
- A Chinese model is released that is widely considered to be the best coding model(not just on price per token)
- AI System wins an award for a significant contribution in mathematics
- A lab releases a fully autonomous drop-in worker agent that at least five fortune 200 companies implement
- OpenAI revenue exceeds $30 billion
- A frontier lab experiences a security incident that requires public disclosure
US Domestic politics:
- Democrats have a majority in the House
- Democrats have a majority in the Senate
- Trump Approval rating exceeds 50% at any point
- A government shutdown exceeds 14 days
- total deportation in 2026 exceed 500k
- Mamadani implements fare free busses city wide
- Mamdani implements at least three state run grocery stores
- Trump is impeached
- A Major political figure is assassinated(Congress/SCOTUS Judge/Executive cabinet member)
Wars:
- Israel-palestine conflict reignites
- Ukraine war ceasefire lasts greater than 30 days at any point
- China invades Taiwan
- US officially at war with Venezuela at any point
China:
- Official GDP growth below 4%
- Youth unemployment (16-24) reported above 15% at any point in 2026
- Major Chinese property developer (top 20) enters formal bankruptcy
- SMIC achieves volume production at 5nm or below
Economics:
- US enters recession (2 consecutive quarters negative GDP growth)
- S&P 500 higher on 12/31/26 than 12/31/25
- US unemployment rate exceeds 5% at any point
- Bitcoin above $150k at any point
- YoY inflation exceeds 4%
Technology:
- Starship upper stage (Ship) successfully lands (caught or propulsive landing)
- Blue Origin New Glenn completes 5+ successful launches
- Tesla releases vehicle with SAE Level 4 autonomy to consumers
- US approves new nuclear reactor construction (not SMR)
Jump in the discussion.
No email address required.
Notes -
Very doubtful, as that would imply (to a lot of people at least) that Bad Things should happen in the very short-term future, and if those Bad Things don't materialize then he'll look foolish. Yudkowsky has steadfastly refused to make any falsifiable predictions so far, so I don't see why he'd get close to one now.
Any regulation? You should probably specify that you think it will be big regulation, like how the DMCA was for Internet copyright. Obviously there will probably be something that at least affects AI in some small way.
Is this a "China's model robustly surpasses the West" take, or is it a more modest take that you think something like Deepseek will join the rotating list of 4 frontier companies? I could see the latter perhaps with like a 30% chance this year, while the former is highly unlikely.
Hell no, not gonna happen. As I've said, there's too many missing capabilities in terms of general computer use and long context management, and those are not easy problems to solve. Well, maybe a lab will release a broken version and a few companies will try it before dropping it, but if you consider that to satisfy your prediction then it's probably already been satisfied in 2025 or even 2024.
Yeah, my odds aren't very high for it, It was hard to come up with something to capture if this is the year we get recursive self improvement and things go hyperbolic.
Congress doesn't pass a lot of bills. I guess I should say that a minor rider on another bill doesn't count, stand alone legislation that's main purpose is the regulation of AI.
It passes if for any time China has the top coding model so it wouldn't need to be particularly robust, just be state of the art and hit in between the top US lab cycles. In practice they would probably just need to join the list of the top frontier companies as all of them but grok have traded off that position occasionally in my opinion. I should probably pick a benchmark but there are a lot and it's hard to say which will even be relevant in a year. I'm trusting Dec 31st 2026 aqouta's taste with evaluation.
Only time will tell, things are moving fast. I don't need the company to be run entirely autonomously, just some slots in some roles being done entirely by an ai drop in worker. I'd accept promising trial announcements that aren't rolled back.
More options
Context Copy link
More options
Context Copy link