theyre not “predictors” theyre gamblers, hes trying to get 18-34 year olds addicted to gambling
more news: Mark Zuckerberg is telling lieutenants to pursue partnerships with Polymarket and Kalshi with Meta's new prediction market app called Arena, per sources
Meta is targeting 18-34 year old users with a goal of 100 million regular "predictors"
www.nytimes.com/2026/06/26/t...
Meta is targeting 18-34 year old users with a goal of 100 million regular "predictors"
www.nytimes.com/2026/06/26/t...
Zuckerberg Urges Meta to Explore Working With Polymarket and Kalshi
www.nytimes.com
June 27, 2026 at 11:56 AM
theyre not “predictors” theyre gamblers, hes trying to get 18-34 year olds addicted to gambling
wow, turns out super-predictors *are* real!*
*they're not
*they're not
This is a damning chart. Gonna have to dismantle these prediction markets posthaste. We need leaders who will do that
May 7, 2026 at 2:20 AM
wow, turns out super-predictors *are* real!*
*they're not
*they're not
Also worth noting that choking is one of the biggest predictors of future intimate partner homicide.
June 14, 2026 at 2:45 PM
Also worth noting that choking is one of the biggest predictors of future intimate partner homicide.
Whatever we do, we CANNOT let “predictors” gain traction as a term. He doesn’t get to sanitize this.
Meta wants gamblers.
100 million regular, 18-34 year old gamblers.
Meta wants gamblers.
100 million regular, 18-34 year old gamblers.
more news: Mark Zuckerberg is telling lieutenants to pursue partnerships with Polymarket and Kalshi with Meta's new prediction market app called Arena, per sources
Meta is targeting 18-34 year old users with a goal of 100 million regular "predictors"
www.nytimes.com/2026/06/26/t...
Meta is targeting 18-34 year old users with a goal of 100 million regular "predictors"
www.nytimes.com/2026/06/26/t...
Zuckerberg Urges Meta to Explore Working With Polymarket and Kalshi
www.nytimes.com
June 27, 2026 at 12:29 AM
Whatever we do, we CANNOT let “predictors” gain traction as a term. He doesn’t get to sanitize this.
Meta wants gamblers.
100 million regular, 18-34 year old gamblers.
Meta wants gamblers.
100 million regular, 18-34 year old gamblers.
You get a divergence between the star-gazers and the predictors over time though. The predictors are *generally* not watching the skies themselves, they’re relying on established charts and tables to determine horoscopes:
December 27, 2025 at 3:27 AM
You get a divergence between the star-gazers and the predictors over time though. The predictors are *generally* not watching the skies themselves, they’re relying on established charts and tables to determine horoscopes:
So the Epstein files dump
1. Protected a lot of predictors
2. Created another partisan food fight.
3. Failed to create any obvious pathway for these women to get accountability.
4. Included a number of clearly unverified claims that can be used to undermine the very legit emails
1. Protected a lot of predictors
2. Created another partisan food fight.
3. Failed to create any obvious pathway for these women to get accountability.
4. Included a number of clearly unverified claims that can be used to undermine the very legit emails
February 1, 2026 at 1:00 PM
So the Epstein files dump
1. Protected a lot of predictors
2. Created another partisan food fight.
3. Failed to create any obvious pathway for these women to get accountability.
4. Included a number of clearly unverified claims that can be used to undermine the very legit emails
1. Protected a lot of predictors
2. Created another partisan food fight.
3. Failed to create any obvious pathway for these women to get accountability.
4. Included a number of clearly unverified claims that can be used to undermine the very legit emails
betting markets are dogshit predictors until the votes start coming in and then they are pretty good
how I found out this happened
polymarket.com/event/wiscon...
polymarket.com/event/wiscon...
August 12, 2026 at 2:07 AM
betting markets are dogshit predictors until the votes start coming in and then they are pretty good
deterministic algorithms outperform text predictors when the task is not text prediction
June 15, 2025 at 2:24 AM
deterministic algorithms outperform text predictors when the task is not text prediction
Note that because gradients flow back to previous positions, it's somewhat misleading to say LLMs are purely next-token predictors (even though they output one token at a time). Rather, they are sequence predictors.
Ah, now I see the issue!
When processing token N, attention doesn’t just receive information from tokens (1…N), it also receives information from intermediate representations of (1…N) at prior layers. With autoregression, this allows a model to “set up” useful representations for future generations
When processing token N, attention doesn’t just receive information from tokens (1…N), it also receives information from intermediate representations of (1…N) at prior layers. With autoregression, this allows a model to “set up” useful representations for future generations
October 16, 2025 at 2:56 AM
Note that because gradients flow back to previous positions, it's somewhat misleading to say LLMs are purely next-token predictors (even though they output one token at a time). Rather, they are sequence predictors.
It is. One of the predictors of conservative behavior is low inter-group contact.
April 18, 2024 at 5:30 PM
It is. One of the predictors of conservative behavior is low inter-group contact.
This is our tragic reality. 👇
"...in both the United States and Canada, political affiliation is now one of the single strongest — if not the strongest — predictors of vaccination status."
My take: www.thestar.com/opinion/cont...
"...in both the United States and Canada, political affiliation is now one of the single strongest — if not the strongest — predictors of vaccination status."
My take: www.thestar.com/opinion/cont...
On measles and vaccines, Canada can’t keep ignoring the elephant in the room
Study after study has shown that political affiliation is one of the single strongest predictors of vaccination status.
www.thestar.com
January 13, 2026 at 6:55 PM
This is our tragic reality. 👇
"...in both the United States and Canada, political affiliation is now one of the single strongest — if not the strongest — predictors of vaccination status."
My take: www.thestar.com/opinion/cont...
"...in both the United States and Canada, political affiliation is now one of the single strongest — if not the strongest — predictors of vaccination status."
My take: www.thestar.com/opinion/cont...
My dating advice for straight men is dump your misogynistic friends.
Really, though. If you dig into the research, being friends with men who blame all their problems on women is one of the biggest predictors of a lot of bad things for men. Especially in dating.
Really, though. If you dig into the research, being friends with men who blame all their problems on women is one of the biggest predictors of a lot of bad things for men. Especially in dating.
March 18, 2025 at 3:24 PM
My dating advice for straight men is dump your misogynistic friends.
Really, though. If you dig into the research, being friends with men who blame all their problems on women is one of the biggest predictors of a lot of bad things for men. Especially in dating.
Really, though. If you dig into the research, being friends with men who blame all their problems on women is one of the biggest predictors of a lot of bad things for men. Especially in dating.
it's crazy that people can't recognize THIS is the position that requires a lot of unfounded confidence about how brains work. "they're just next-word predictors" okay and how do you know that we aren't?
or 3) there isn’t any evidence that the human brain works like a large language model and since you‘ve provided none, we can comfortably say that “machine that predicts the next word based off probability“ is not conscious.
You have to *prove* it and not just go “well it looks like it thinks”.
You have to *prove* it and not just go “well it looks like it thinks”.
Ultimately you get backed into one of two positions:
1) the AIs are, by any reasonable standard, thinking and understanding, albeit in a way very different from us
2) “thinking” and “understanding” require a soul, which we have and they do not
The territory between these has basically collapsed
1) the AIs are, by any reasonable standard, thinking and understanding, albeit in a way very different from us
2) “thinking” and “understanding” require a soul, which we have and they do not
The territory between these has basically collapsed
September 20, 2026 at 10:44 PM
it's crazy that people can't recognize THIS is the position that requires a lot of unfounded confidence about how brains work. "they're just next-word predictors" okay and how do you know that we aren't?
Almost like casual transphobia and ableism are often predictors of other bad behavior
Shocking that the politician who made a “butch body” joke and a “governor hot wheels” joke is a Chris Brown fan
September 14, 2026 at 12:36 PM
Almost like casual transphobia and ableism are often predictors of other bad behavior
The main predictors of adult (financial) success are nepotism, wealth, and sheer dumb luck. Thank you for coming to my TED Talk.
December 29, 2025 at 1:13 PM
The main predictors of adult (financial) success are nepotism, wealth, and sheer dumb luck. Thank you for coming to my TED Talk.
in my experience studying extremists of all stripes, the best predictors for understanding them are
-what they say they believe, particularly in intracommunity texts
-what they say they want to do
-what they say they believe, particularly in intracommunity texts
-what they say they want to do
November 16, 2024 at 7:35 PM
in my experience studying extremists of all stripes, the best predictors for understanding them are
-what they say they believe, particularly in intracommunity texts
-what they say they want to do
-what they say they believe, particularly in intracommunity texts
-what they say they want to do
Study: Predictors of online exposure to disinformation www.sciencedirect.com/science/arti...
Predictors of exposure to disinformation:
- right-wing populism attitudes
- trust in non-traditional media
- social media reliance
- lower levels of education
Predictors of exposure to disinformation:
- right-wing populism attitudes
- trust in non-traditional media
- social media reliance
- lower levels of education
August 7, 2025 at 7:49 PM
Study: Predictors of online exposure to disinformation www.sciencedirect.com/science/arti...
Predictors of exposure to disinformation:
- right-wing populism attitudes
- trust in non-traditional media
- social media reliance
- lower levels of education
Predictors of exposure to disinformation:
- right-wing populism attitudes
- trust in non-traditional media
- social media reliance
- lower levels of education
Statements on values and stances as predictors of future action, and whatnot, too
February 28, 2026 at 4:19 PM
Statements on values and stances as predictors of future action, and whatnot, too
The now viral, incorrect meme that LLMs are just next token predictors is causing so much confusion
May 21, 2025 at 12:12 AM
The now viral, incorrect meme that LLMs are just next token predictors is causing so much confusion
but it is only anecdote. There’s no evidence that family and friends checking in has generalisable protective effects. But there is evidence that suicide rates in close family members go up sharply after a loved one takes their own life. Shame and guilt are strong predictors of suicidal behaviour.
September 24, 2026 at 7:13 PM
but it is only anecdote. There’s no evidence that family and friends checking in has generalisable protective effects. But there is evidence that suicide rates in close family members go up sharply after a loved one takes their own life. Shame and guilt are strong predictors of suicidal behaviour.
So I wrote a bunch of words about what election predictors you should (and shouldn’t) trust. www.washingtonpost.com/politics/202...
Column | Which election predictors deserve your trust — and which probably don’t
A meta-guide to assessing how the election will unfold.
www.washingtonpost.com
October 10, 2024 at 7:59 PM
So I wrote a bunch of words about what election predictors you should (and shouldn’t) trust. www.washingtonpost.com/politics/202...
problem is they are just kneejerk next token predictors and can't engage in actual discourse
January 9, 2026 at 2:28 AM
problem is they are just kneejerk next token predictors and can't engage in actual discourse
maybe the most enduring lesson i've been forced to learn by coming out is that status, class, and power are much better predictors of behavior than gender or sexuality
April 5, 2025 at 1:20 AM
maybe the most enduring lesson i've been forced to learn by coming out is that status, class, and power are much better predictors of behavior than gender or sexuality
This is especially concerning because persistence is foundational to skill acquisition and one of the *strongest* predictors of long-term learning.
AI conditions you to expect immediate answers, removing the productive struggle that builds real competence.
AI conditions you to expect immediate answers, removing the productive struggle that builds real competence.
April 7, 2026 at 5:36 PM
This is especially concerning because persistence is foundational to skill acquisition and one of the *strongest* predictors of long-term learning.
AI conditions you to expect immediate answers, removing the productive struggle that builds real competence.
AI conditions you to expect immediate answers, removing the productive struggle that builds real competence.