Prior results have shown that LLMs released before 2024 can't leverage 'filler tokens'—unrelated tokens prior to the model's final answer—to perform additional computation and improve performance.

Repeating the problem statement generally works better and is more reliable than filler tokens, especially for relatively weaker models I test (e.g. Qwen3 235B A22B), though the performance boost is often very similar, especially for Anthropic models.

---

Outline:

(03:59) Datasets

(06:01) Prompt

(06:58) Results

(07:01) Performance vs number of repeats/filler

(08:29) Alternative types of filler tokens

(09:03) Minimal comparison on many models

(10:05) Relative performance improvement vs default absolute performance

(13:59) Comparing filler vs repeat

(14:44) Future work

(15:53) Appendix: How helpful were AIs for this project?

(16:44) Appendix: more information about Easy-Comp-Math

(25:05) Appendix: tables with exact values

(25:11) Gen-Arithmetic - Repetitions

(25:36) Gen-Arithmetic - Filler

(25:59) Easy-Comp-Math - Repetitions

(26:21) Easy-Comp-Math - Filler

(26:45) Partial Sweep - Gen-Arithmetic - Repetitions

(27:07) Partial Sweep - Gen-Arithmetic - Filler

(27:27) Partial Sweep - Easy-Comp-Math - Repetitions

(27:48) Partial Sweep - Easy-Comp-Math - Filler

(28:09) Appendix: more plots

(28:14) Relative accuracy improvement plots

(29:05) Absolute accuracy improvement plots

(29:57) Filler vs repeat comparison plots

(30:02) Absolute

(30:30) Relative

(30:58) Appendix: significance matrices

(31:03) Easy-Comp-Math - Repetitions

(32:27) Easy-Comp-Math - Filler

(33:50) Gen-Arithmetic - Repetitions

(35:12) Gen-Arithmetic - Filler

The original text contained 9 footnotes which were omitted from this narration.

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

Screwtape, as the global ACX meetups czar, has to be reasonable and responsible in his advice giving for running meetups.

And the advice is great! It is unobjectionably great.

It's one in the morning. I just ran the east coast rationalist megameetup. A late night spike of my least favourite thing to hear about a meetup I'm running means I'm not going to be able to sleep for a bit. One of my favourite organizers has recently published a list of opinionated meetup takes, saying I have to be reasonable and responsible.

I have to be reasonable and responsible in my advice giving, eh?

I'm the czar. Which one of you proposes to make me?

(Epistemic status: Written at one in the morning, after having slept about twelve hours in the last seventy-two, and a spike of cortisol. The odds I regret posting this are higher than pretty much anything I've put on the internet associated with my name before.)

Run meetups at a time convenient to you, a place convenient to you, and on a subject you find interesting

For a while now Boston has had a regular meetup Wednesday evening. Why Wednesday? Well, because I'm the [...]

---

Outline:

(01:13) Run meetups at a time convenient to you, a place convenient to you, and on a subject you find interesting

(02:46) Tell attendees to do the thing you want em to do

(04:20) Tell people who make organizing less fun for you to go away

(05:21) Get someone else to pay for it

(06:54) Optimize for what you want out of meetups, and what you want can be pretty weird

The original text contained 2 footnotes which were omitted from this narration.

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

Recent work by Anthropic showed that Claude models, primarily Opus 4 and Opus 4.1, are able to introspect--detecting when external concepts have been injected into their activations. But not all of us have Opus at home! By looking at the logits, we show that a 32B open-source model that at first appears unable to introspect actually is subtly introspecting. We then show that better prompting can significantly improve introspection performance, and throw the logit lens and emergent misalignment into the mix, showing that the model can introspect when temporarily swapped for a finetune and that the final layers of the model seem to suppress reports of introspection.

We do seven experiments using the open-source Qwen2.5-Coder-32B model. See the linked post for more information, but a summary of each:

Experiment 1: We inject the concepts "cat" and "bread" into the first user / assistant turn, and show that, while the model initially appears to not be introspecting, there's actually a very slight logit shift towards 'yes' and away from 'no' on injection when answering "Do you detect an injected thought..." with "The answer is...":

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

There's a stock market bubble currently in progress. The key question

There are a few ways that the current AI boom could turn out to be a

I'm betting that if any of those causes happen, they're more than two

Note that that I don't take seriously any claims that AI is approaching

The most likely alternative to AI being in a bubble is that some large

---

First published:

Source:

Linkpost URL:
https://bayesianinvestor.com/blog/index.php/2025/12/21/whats-the-current-stock-market-bubble/

---

Narrated by TYPE III AUDIO.

There is a standard pattern in philosophical conversations. People stumble upon their epistemological limitations and then assume that God's existence somehow would’ve solved the problem.

For instance, suppose we are talking about Münchhausen trilemma, demonstrating the inability to have completely certain foundation for knowledge. We inspect different horns of the trilemma and then conclude with a sigh:

“Oh well, it seems there is no absolute certainty in our world. Unless, of course, God Himself would tell us that something is true”.

This is deeply related to the idea of curiosity-stoppers. Standard non-answers that make people pretend that the problem is solved and, therefore, cease any further investigation. Naturally, saying “God!” is a very popular one.

But this time the situation is a bit different. The problem was already investigated and discussed. Dare I even say, understood. And yet we are still invoking “God!” as an afterthought. As if we are trying to appeal to His divine vanity and give Him some credit anyway, even if we’ve just explored all the reasons He wouldn’t help at all.

Imagine that it was extremely clear that God existed. Imagine everyone could close their eyes in prayer and in [...]

---

First published:

Source:

Linkpost URL:
https://apeinthecoat102771.substack.com/p/no-god-can-help-you

---

Narrated by TYPE III AUDIO.

Someone on reddit said, "Remember robots still can't go into a new house and make a coffee." And I thought

I actually wonder whether, if I provided the physical actuation, current LLMs would be capable of doing this? Like, through a conversation like:

Me: I'm in a house. Your job is to instruct me to make a coffee. I can take photos of my surroundings, I can follow basic directions, and if you ask me to do something too complicated I'll ask for clarification. Here is my current surroundings: (photo)

LLM: Okay, we need to find the kitchen. There's a door on the right of the photo, go through that.

Me: Here's where I am now: (photo)

LLM: That looks like the kitchen on the left, go there.

Me: It looks like this: (photo)

LLM: Now we need to find either a coffee maker or a kettle. Look through the cupboards.

Me: I don't know what those things look like.

LLM: Then open the cupboard on the left and show me a photo.

...and so on.

It wouldn't shock me either way if they can or can't do it. I think I weakly predict that the models have the [...]

The original text contained 3 footnotes which were omitted from this narration.

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

Master version of this on https://parvmahajan.com/2025/12/21/turning-20.html

I turn 20 in January, and the world looks very strange. Probably, things will change very quickly. Maybe, one of those things is whether or not we’re still here.

This moment seems very fragile, and perhaps more than most moments will never happen again. I want to capture a little bit of what it feels like to be alive right now.

1. 

Everywhere around me there is this incredible sense of freefall and of grasping. I realize with excitement and horror that over a semester Claude went from not understanding my homework to easily solving it, and I recognize this is the most normal things will ever be. Suddenly, the ceiling for what is possible seems so high - my classmates join startups, accelerate their degrees; I find myself building bespoke bioinformatics tools in minutes, running month-long projects in days. I write dozens of emails and thousands of lines of code a week, and for the first time I no longer feel limited by my ability but by my willpower. I spread the gospel to my friends - “there has never been a better time to have a problem” - even as [...]

---

Outline:

(00:34) 1.

(03:12) 2.

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

Imagine you were born in 2007. The Internet has always existed. Facebook has always existed. Google has always existed. Smartphones have always existed.

By the time you became old enough to social media, social media had already evolved from "truly social" to "superintelligent hostile optimization system working tirelessly to grab your attention and mind control you via brainrot, ragebait, and parasocial relationships". By the time you were old enough to drive, all new cars had annoying touchscreen interfaces designed as if their purpose was to distract you and crash the car. Screens don't even advance. There's no point in building a display with higher resolution than a retina display. Instead, the Corporate Overlords managing the software arbitrarily change up the user interfaces of your software just as soon as you start to like the old interface.

The blogosphere is unheard of, by which I mean you have literally never heard of it. You don't know what a blog is or how they work. "Media" means YouTube, Instagram and Tiktok. Hosting your own website is unheard of too. You don't know that the first webcomic artists managed their own physical servers—or even what a real server is—everything is in the [...]

---

Outline:

(03:03) Retro Tech

(07:25) Technological Responsibility

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

Going into November, I wanted to write a sequence of blog posts that would convince people to practice digital intentionality, and show them how to do it.

The theory of change was something like: A lot of people have written books on this topic, and they sometimes work to change people's behavior. But the people who need the message the most aren’t reading books, because they can’t. The only things they read are online, so I should write to them online, on a platform that they’re already on, like Substack.

This was not crazy. But I now feel that it was confused.

Where was I coming from?

Reading Digital Minimalism led me to completely change my life. I was catastrophically miserable, and digital minimalism allowed me to fix that.

This experience filled me with a proselytizing urge. I looked around at my friends and family, and saw them all enslaved to their screens, and I imagined how much happier and more fulfilled they could be if they only got free.

Despite my zealousness (and/or perhaps a little bit because of it), I’ve only had one friend ask me to help them do a digital declutter in the two years [...]

---

Outline:

(00:45) Where was I coming from?

(02:20) What do I want?

(02:58) Regulation

(04:02) Digital intentionality at the community level

(04:43) So what now?

(04:58) Community assurance contract

(05:38) More posts!

(05:56) Coaching?

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

Audio note: this article contains 162 uses of latex notation, so the narration may be difficult to follow. There's a link to the original text in the episode description.

One of the weirdest things in mathematics is how completely unrelated fields tend to connect with one another.
A particularly interesting case is the play between number theory (the study of natural numbers _mathbb{N}={0,1,2dots}_) and complex analysis (the study of functions on _mathbb{C}={a+bi:a,bin mathbb{R}}_). One is discrete and uses modular arithmetic and combinatorics; one is continuous and uses integrals and epsilon-delta proofs. And yet, many papers in modern number theory use complex analysis. Even if not directly, basically all of them rely on some result with a complex analytic flavor.

What's going on?

This post gives an example of a fundamental result in number theory that uses complex analysis, and tries to explain where that usage is coming from. My writing is semi-technical, I'm not trying to formalize everything, but mostly doing this as an exercise to clear out my own confusion (and hopefully to convince others of why this mathematical connection is valuable).

Technical details in the proof follow Ang Li's notes; exposition, intuition and possible mistakes [...]

---

Outline:

(01:31) I Dirichlets Theorem

(03:58) II Eulers proof for infinitude of primes

(08:12) III Trying the same for Dirichlet

(16:25) IV Wrapping up (or: getting started)

(19:20) V Looking ahead

The original text contained 4 footnotes which were omitted from this narration.

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.