Founder @simlcrm | AI superforecasters
New York, NY, USA
Joined February 2026
- Tweets125
- Following79
- Followers421
- Likes286
My PhD advisor received an email from a student who's been doing research for 8 months, and has *16 solo papers published* in 2026, almost all at real AI conferences, 1 paper at ICML.
All papers 100% flagged by Pangram (not reliable but a tell). The truth is academic AI research does not survive this. It has to reinvent itself fast.
To be clear, I am extremely pro-AI, and I believe most text should and will be generated by AI in the future. The use of AI is not the problem. The problem is that AI academia has rewarded cookie cutter experimental and theoretical papers at the expense of durable findings to build up researcher capacity.
But with coding and math now being cheap, this model no longer works: the current AI researcher (labor) will substituted by AI (capital). If it wants to survive (and this is a question mark since most of the goals of the field have been accomplished), Academic AI has to reward researchers who build models and theories, not those who follow recipes.
In hindsight, with post-training RL now clearly the winner for building machine intelligence, Shane Legg and Marcus Hutter’s mathematical definition of Universal Intelligence is what AI companies are clearly building.
The terminology is also clearly superior to AI or SI. The intelligence is universal because it is build on the knowledge and thought of all. It is neither superior to human intelligence nor artificial, rather it is compression of the universe of all intelligence.
Nash's paper on equilibrium existence is almost all english text, there's very little math notation. I think all math heavy fields will converge to similar paper styles.
The AIs will prove + verify the boring math details, and authors + reviewers will debate the theoretical frameworks rather than the technical details. At the of the end of the day, math has always been about the models and their properties, not the proofs.
We asked that the OpenReview forum of our paper be made public because our reviews were all great, PC ignored our emails, and the paper+reviews are now all gone.
@NeurIPSConf Why remove the paper? How is an undisclosed requirement leading to desk rejection? Is this open science?
Our NeurIPS paper with all accepts has been desk rejected after reviews and rebuttals for 5 *real* but incorrectly entered citations *in the appendix*.
Why waste reviewer and author and AC time, if you will run blanket AI rejections after reviews?
@NeurIPSConf @konstantina_pll @JessicaSchrouff @joavanschoren @alexandredrouin
Until we can scale up voting mechanisms theory, it seems that a bicameral presidential system where the lower legislative body and the highest judicial court is chosen entirely by sortition is the best.
I’m sure alignment fixes this
BREAKING: The US "almost started a war" against China, nearly launching an operation against a Chinese ship in the Middle East this spring, in the midst of the Iran war, after an AI-generated intelligence report falsely identified the vessel as carrying nuclear weapons components, with armed US personnel preparing to board and military aircraft already in the air before officials found the report had been produced with a chatbot, per CNN.
A special operations command analyst had used a chatbot to fuse open-source material with classified signals intelligence, then used AI again to package the result into a standard intelligence report and circulated it across the US military.
Our NeurIPS paper with all accepts has been desk rejected after reviews and rebuttals for 5 *real* but incorrectly entered citations *in the appendix*.
Why waste reviewer and author and AC time, if you will run blanket AI rejections after reviews?
@NeurIPSConf @konstantina_pll @JessicaSchrouff @joavanschoren @alexandredrouin
@NeurIPSConf @konstantina_pll @JessicaSchrouff @joavanschoren @alexandredrouin This information was conveyed in our response to the PC where provided clear evidence the citations are not fabricated, and it was due to our appendix being submitted separately from the main paper.
I don’t understand why people treat pangram scores as reliable. What stops anyone from fine-tuning a model to maximize the pangram human score to sound human?
If you could verifiably claim something is human, then RLVR can replicate it, no?
So many services that have been free so far (e.g., Google drive CLI), will get overwhelmed by agent requests soon. The costs of providers will go up, and with the demand up too, these services will start billing per-request, agents getting wallets is inevitable.
This is the wrong thinking paradigm fwiw. The right one is that it is the natural evolution of programming languages (C -> Python -> English) and their processors (compilers -> interpreters -> LLMs).
The programmer will survive but will have to be an abstract systems thinker.
the thinking that "everyone will just vibe code their own software" was born from within a bubble
what's happening to software development right now has happened many, many times before to other things. you just need to look
web 2.0 allowed everyone to write blogs. did everyone write blogs?
instagram allowed everyone to share beautiful photos. did everyone share beautiful photos?
tiktok allowed everyone to make short videos. did everyone make short videos?
now vibe coding allows everyone to build software, and you think everyone will start building software? this time around, people are just different?
the mainstream is always a consumer, never a creator. go talk to some people outside of our tech bubble and you'll see - they literally don't give a f about vibe coding