g-leech

THE BEAUTIFUL AND THE SUBLIME

First published in Isaac Asimov's Science Fiction Magazine, June 1986.

May 30, 2070

My dear MacLuhan:

You, my friend, who know so well a lover's troubles, will understand my affair with Leona Hillis. Since my last letter to you, I have come to know... Leona's soul. Slowly, almost despite myself, I opened those reservoirs of sympathy and feeling that turn a simple liaison into something much deeper. Something that partakes of the sublime. It is love, my dear MacLuhan. Not the appetite of the body, easily counterfeited with pills. No, it is closer to agape, the soaring spiritual union of the Greeks. I know the Greeks are out of favor these days, especially Plato with his computerlike urge toward abstract intellect.

g-leech

MT: Hi, I'm Matt Turk. Welcome back to the Mad Podcast. For this first episode of 2026, my guest is Pavel Ismileov, a researcher at Anthropic and a professor at NYU. We kick off this episode by deconstructing a viral article about models evolving alien survival instincts. We also talk about the cultural differences between the major labs, the future of reasoning models in 2026 and the brand new paper he co-authored on a concept called epiplexity. Please enjoy this fascinating look at the frontier of AI safety and reasoning. Pavel, welcome.

PI: Thank you so much for having me.

MT: I wanted to start this conversation with an article that went viral during the holidays on X called Footprints in the Sand published by an anonymous account called I Rule the World Mo. The core thesis is that models across pretty much any lab are evolving unprogrammed what they call alien survival instincts. the ability for the model to realize that it's being evaluated and then react deceptively like faking alignment or engaging

g-leech

Matt: Hello and welcome to Elucidations, an unexpected philosophy podcast. I'm Matt Teichman, and with me today is Sam Enright, editor-in-chief of The Fitzwilliam, which is an online publication about Ireland policy, philosophy, literature. And he is also a non-resident fellow at the policy think tank Progress Ireland and an Emergent Ventures fellow. And he's here to discuss lifelong learning. Sam Enright, welcome. 


Sam: Thanks for having me. 


Matt: So, I would definitely put you at the top of my list of people who are r- really good at learning. Um, but before we sort of talk about some of the, uh, ways that you like to learn, maybe we could just address this question of, like, is it a good thing to learn? I mean, I think a lot of us assume the answer is yes. I certainly assumed that. I love learning. But, like, I don't know, is that wrong? Should we... Is it a waste of time to be learning? 


Sam: (laughs) I think the question of the returns to effort of learning about different fields is a really non-t

g-leech

ASAT

• "AGI Safety & Alignment"
  • amplified oversight
  • interpretability
  • ASAT eng (automated alignment research)
  • Causal Incentives Working Group,

Frontier Safety

• Risk Assessment (evals, threat models, the framework),
• Mitigations (e.g. banning accounts, refusal training, jailbreak robustness)

g-leech

ACTION BUTTON REVIEWS DOOM by Tim Rogers

introduction

Um.

Hello and welcome back to video games. I'm Tim Rogers. You are watching the Action Button review of DOOM — a video game developed by id Software and published for the Microsoft Disk Operating System on December 3rd, 1993; the Sega 32X and Atari Jaguar in 1994; the Super Nintendo Entertainment System and the Sony PlayStation in 1995; the Panasonic 3DO in 1996; the Sega Saturn in 1997; the Acorn Archimedes in 1998; the Nintendo Game Boy Advance in 1999; the Microsoft Xbox 360 in 2006; Apple iOS in 2009; the Microsoft Xbox 360 again in 2012; and on the Nintendo Switch, the Sony PlayStation 4, and the Microsoft Xbox One in 2019.

Updated with a fourth episode, "Thy Flesh Consumed," in The Ultimate DOOM — released for Microsoft personal computers on April 30th, 1995 — and by John Romero with a fifth episode, SIGIL, released for personal computers on May 22nd, 2019. Created by John Carmack, John Romero, Adrian Carmack, Kevin Cloud, and Tom Hall, with

g-leech

library(matlib)

gamma <- 2
rho11 <- 1.15
rho22 <- 0.05
mu1 <- 0.3
r <- 0.05
delta <- 0.05

x <- rnorm(1000000)

g-leech

https://twitter.com/g_leech_/status/1728731355029393447?t=VN50RyPqwWFfcC63tMP_ng&s=19
https://twitter.com/g_leech_/status/1723630099507871780?t=YFYUi0eNkkLXOzzmbMtoGA&s=19
https://twitter.com/g_leech_/status/1623044128190697483
https://twitter.com/g_leech_/status/1639725932448456705
https://twitter.com/g_leech_/status/1630276985963556865
https://twitter.com/g_leech_/status/1622321776595189765
https://twitter.com/g_leech_/status/1668960035068694533
https://twitter.com/g_leech_/status/1715654603973312949
https://twitter.com/g_leech_/status/1713930232632135986
https://twitter.com/g_leech_/status/1711441986878644597

g-leech

import numpy as np
import gjp
# https://github.com/niplav/iqisa/blob/master/iqisa.py
import iqisa as iqs
import matplotlib.pyplot as plt

df=gjp.load_markets()

df["isCorrect"] = (df.outcome == df.answer_option).astype(float)
fig = plt.figure(figsize=(9, 9))

g-leech

g-leech