Seems like a good time to look back at all the talk about preference falsification and cascades from 2024 and see how it might apply to the midterms
Preference falsification has been central to the trajectory of DEI. People who abhor DEI principles and methods came to favor these publicly through a preference cascade. Every instance of preference falsification induced others to pretend they consider DEI just, efficient, beneficial to marginalized groups, etc. In time, a false consensus effectively displaced the search for truth as the university’s core mission, replacing it with DEI. Most professors watched in concealed horror the transfer of enormous powers from themselves to rapidly growing DEI bureaucracies. In countless contexts, they endorsed policies they considered harmful, participated in the defamation of scholars they admired, and sheepishly submitted to DEI training—all to be left alone, to avoid being called racist, to advance their careers. But the resulting equilibrium was self-undermining. In emboldening DEI officials, it increased privately felt anger and resentment. The stage was set for a preference cascade in reverse. The shock that unleashed the ongoing cascade in reverse was the Hamas massacres of October 7. The chain of events that they triggered in the U.S.—anti-Jewish demonstrations, the Congressional hearings, the plagiarism revelations—brought to the surface outrage that had been building up quietly for years. As public criticism of DEI grew, and as it became clear that broad segments of the left share the outrage, the DEI-favoring false consensus disintegrated. The process is by no means over. As preference falsification on DEI continues to diminish, more scholars and administrators will find in themselves the courage to stand up for values that they truly believe in. Though the DEI complex will fight back, it won’t be easy to quash the honest discourse finally under way. Public critics of DEI know they aren’t alone.
2
35
David Moon retweeted
Preference falsification has been central to the trajectory of DEI. People who abhor DEI principles and methods came to favor these publicly through a preference cascade. Every instance of preference falsification induced others to pretend they consider DEI just, efficient, beneficial to marginalized groups, etc. In time, a false consensus effectively displaced the search for truth as the university’s core mission, replacing it with DEI. Most professors watched in concealed horror the transfer of enormous powers from themselves to rapidly growing DEI bureaucracies. In countless contexts, they endorsed policies they considered harmful, participated in the defamation of scholars they admired, and sheepishly submitted to DEI training—all to be left alone, to avoid being called racist, to advance their careers. But the resulting equilibrium was self-undermining. In emboldening DEI officials, it increased privately felt anger and resentment. The stage was set for a preference cascade in reverse. The shock that unleashed the ongoing cascade in reverse was the Hamas massacres of October 7. The chain of events that they triggered in the U.S.—anti-Jewish demonstrations, the Congressional hearings, the plagiarism revelations—brought to the surface outrage that had been building up quietly for years. As public criticism of DEI grew, and as it became clear that broad segments of the left share the outrage, the DEI-favoring false consensus disintegrated. The process is by no means over. As preference falsification on DEI continues to diminish, more scholars and administrators will find in themselves the courage to stand up for values that they truly believe in. Though the DEI complex will fight back, it won’t be easy to quash the honest discourse finally under way. Public critics of DEI know they aren’t alone.
Really, @columbiasport - you don’t even have the @columbia Twitter handle. Did you actually think anyone would mistake shoddy outdoor wear for a fine university?
46
What a great time to suggest that people start raising chickens in their backyards desertsun.com/story/news/nat…
36
Welcome to the new war on Christmas nytimes.com/2025/04/29/busin… #tariffs
39
I guess @cnn will also be forced to report this, @bennyjohnson
New consumer sentiment data just out for April:
Americans’ expectations for how the economy will fare over the next year are now at their worst in 45 years.
Consumer confidence is down 40% since Trump’s inauguration, including a 20% drop among Republicans.
72
I don’t get it. Are you not a capitalist? Isn’t the market just a big information machine that drives the price mechanism and sorts winners and losers?
Are you saying we would be better off if the administration made those decisions and planned the economy?
68
David Moon retweeted
Time to re-up Milton Friedman and the making of a pencil…
xitter.gq/KingLibertarian/status…
Poor @levie — he was booked in cnbc to talk about his developments with ai agents, but the hosts zeroed in on his talks with enterprise ceos about tariffs. Proud of him for not shrinking from the discussion and calling out how damaging trumps tariffs are, unlike @BizRoundtable
81
We are living in a timeline where a non-US company is keeping the original mission of OpenAI alive - truly open, frontier research that empowers all. It makes no sense. The most entertaining outcome is the most likely.
DeepSeek-R1 not only open-sources a barrage of models but also spills all the training secrets. They are perhaps the first OSS project that shows major, sustained growth of an RL flywheel.
Impact can be done by "ASI achieved internally" or mythical names like "Project Strawberry".
Impact can also be done by simply dumping the raw algorithms and matplotlib learning curves.
I'm reading the paper:
> Purely driven by RL, no SFT at all ("cold start"). Reminiscent of AlphaZero - master Go, Shogi, and Chess from scratch, without imitating human grandmaster moves first. This is the most significant takeaway from the paper.
> Use groundtruth rewards computed by hardcoded rules. Avoid any learned reward models that RL can easily hack against.
> Thinking time of the model steadily increases as training proceeds - this is not pre-programmed, but an emergent property!
> Emergence of self-reflection and exploration behaviors.
> GRPO instead of PPO: it removes the critic net from PPO and uses the average reward of multiple samples instead. Simple method to reduce memory use. Note that GRPO was also invented by DeepSeek in Feb 2024 ... what a cracked team.
60
Democracy is the theory that the common people know what they want, and deserve to get it good and hard.
41
The aggressor is always peace-loving (as Bonaparte always claimed to be); he would prefer to take over our country unopposed.
—Carl von Clausewitz, On War
74















