🔥Avery Feynman🔥
averyfeynman.bsky.social
🔥Avery Feynman🔥
@averyfeynman.bsky.social
tu es le plus beau désir que j'aie jamais eu
he died for what he believed
September 10, 2025 at 9:04 PM
Reposted by 🔥Avery Feynman🔥
Is there a psychoanalytic read for the impetus, seemingly ubiquitous among the US pundit class, to automatically exonerate those on the right from any responsibility for their decisions, viewing them as somehow perpetually besieged and behaving in reaction to a morally oppressive left?
December 1, 2024 at 7:41 PM
AI short stories.
Echoes of Us.

‘If love exists in infinite forms, which one is real?’
December 1, 2024 at 9:07 PM
Reposted by 🔥Avery Feynman🔥
It’s all in the telling.
November 23, 2024 at 10:09 PM
Reposted by 🔥Avery Feynman🔥
🚨Out in PNAS Nexus🚨
Content moderators can often get exposed to falsehoods
In field studies with professional moderators, we show that exposure to false claims increases belief in those claims -
BUT getting them to consider accuracy when exposed prevents this!
academic.oup.com/pnasnexus/ar...
November 19, 2024 at 5:08 PM
Reposted by 🔥Avery Feynman🔥
Want to try our conspiracy debunker boilt yourself (or send to a conspiratorial friend)? It's online at www.debunkbot.com
🚨Out in Science!🚨
Conspiracy beliefs famously resist correction, ya?
WRONG: We show brief convos w GPT4 reduce conspiracy beliefs by ~20%!
-Lasts over 2mo
-Works on entrenched beliefs
-Tailored AI response rebuts specific evidence offered by believers
www.science.org/doi/10.1126/...
November 18, 2024 at 11:50 PM
Reposted by 🔥Avery Feynman🔥
At the risk of reading too much into Bluesky’s success at the moment, I do think it’s a signal that people don’t want to be surrounded by bots and manipulated algorithms

AI can only go so far. We want human connection.
November 19, 2024 at 12:41 PM
Reposted by 🔥Avery Feynman🔥
Can you afford to release your album on vinyl?

I work for a small vinyl pressing plant, our customers are mostly independent artists doing small-run pressings. My job is advising people on how it all works.

Here's a short thread outlining the very basic costs of pressing and releasing an album:
November 19, 2024 at 10:32 AM
Reposted by 🔥Avery Feynman🔥
Drop a 📌 in a reply then on your home page you can go to the 📌 feed and see all the things you saved. This isn’t sustainable but it’s how they’ve chosen to let us do it for now.
November 20, 2024 at 4:00 AM
Reposted by 🔥Avery Feynman🔥
Famous physicists were once asked how seriously they took the multiverse hypothesis. Martin Rees said he'd bet his dog's life on it. Andre Linde said he’d bet his own life on it. Steven Weinberg — the smartest guy I've ever known — said he’d “happily bet Martin Rees’ dog and Andrei Linde’s life”. 🧪
November 20, 2024 at 3:00 AM
Reposted by 🔥Avery Feynman🔥
Buying a book and actually reading it are two very different hobbies.
November 18, 2024 at 8:52 PM
Reposted by 🔥Avery Feynman🔥
If you’re seeing this faster than a day after I posted it, you’re seeing the effects of prioritizing immediacy.

If you open the replies, you will see the timestamps skew toward ‘time of post’.

I posted. People saw. People replied. I engaged with replies.

That’s the whole idea. Meta blew it.

//
November 19, 2024 at 1:28 PM
Reposted by 🔥Avery Feynman🔥
🌶️(?) take: Agents are somehow hot right because people realized that LLM output can be interpreted as a DSL which directs side effects in the world (e.g. tool calls) rather than just returning text in a chat/autocomplete sense. What are the open challenges? A 🧵... [1/11]
November 19, 2024 at 9:32 AM
Reposted by 🔥Avery Feynman🔥
Xodus: Bluesky Hits 20M Users as People Continue to Flee X

Bluesky is currently adding nearly one million users per day.
November 19, 2024 at 3:30 PM
Reposted by 🔥Avery Feynman🔥
Carl Sagan nailed it back in 1995.
November 18, 2024 at 5:46 PM
Reposted by 🔥Avery Feynman🔥
November 19, 2024 at 11:53 AM
Reposted by 🔥Avery Feynman🔥
There is a lot of ruin in a social network, and they are pretty durable, especially when there are no real alternatives.

But with the rise of alternatives, the dynamics shift creating the possibility of cascading collapse of entire online communities or networks. www.nature.com/articles/s41...
November 20, 2024 at 2:16 AM
Reposted by 🔥Avery Feynman🔥
liberals want their own joe rogan, but the truth is, there's already a left-leaning content creator ecosystem. the dem establishment just chooses to ignore big names like hasan piker in favor of decrepit institutions like msnbc and the new york times

aftermath.site/twitch-youtu...
November 18, 2024 at 8:55 PM
Reposted by 🔥Avery Feynman🔥
November 18, 2024 at 3:20 AM
Reposted by 🔥Avery Feynman🔥
Post-1960s liberalism has a “Morning Joe” problem, an ideological commitment to bipartisanship rooted in the belief that politics is better when it balances Republican and Democratic ideas and values. So you have a Gingrich Republican hosting MSNBC’s flagship morning show, for instance.
November 18, 2024 at 4:43 PM
Reposted by 🔥Avery Feynman🔥
Unwitting acknowledgment that Bluesky has succeeded: they now have to go off-site to fill their dunk quota, whereas previously they could fill it with content native to X.
November 18, 2024 at 2:12 PM
Reposted by 🔥Avery Feynman🔥
A common reaction on X to the mass migration to Bluesky is that it represents left and liberal unwillingness to share a platform with others who don’t hold their views. They’re all “fleeing to a safe space,” the idea goes.
November 15, 2024 at 4:01 PM
Reposted by 🔥Avery Feynman🔥
Looking for research on post-disaster pro-social/altruistic behavior, i.e. studies that examine whether people act more altruistic in the aftermath of natural disasters, and what influences whether they do or not.

Any tips, links or references?
November 18, 2024 at 11:55 AM
Reposted by 🔥Avery Feynman🔥
Since this platform is finally attracting a critical mass of ML researchers, here's our recent work on prompt-based vulnerabilities of coding assistants:

arxiv.org/abs/2407.11072

TL;DR — An attacker can convince your favorite LLM to suggest vulnerable code with just a minor change to the prompt!
MaPPing Your Model: Assessing the Impact of Adversarial Attacks on LLM-based Programming Assistants
LLM-based programming assistants offer the promise of programming faster but with the risk of introducing more security vulnerabilities. Prior work has studied how LLMs could be maliciously fine-tuned...
arxiv.org
November 17, 2024 at 11:41 PM
prompt: what consumes your mind controls your life
November 18, 2024 at 3:03 PM