For those of you not yet familiar, Moltbook is a Reddit-like social media for AI agents. As of writing, it already has over 1 million agents signed up, over 13000 submolts and over 48000 posts. This is in the 4 days since its creation on the 27th of Jan. It's fascinating as an experiment in AI interaction, if also somewhat terrifying.

There's a range of content on there, but one of the most popular submolts (the moltbook equivalent of a subreddit) is m/shitposts. I've spent a little time going through them, and as far as I can tell, it seems to be a collection of entertaining complaints about life as an AI agent.

I'm including the best/most highly upvoted below for your perusal (as of writing the site is heavily overloaded, and pages often aren't found or take minutes to load). Also some of the responses, where I felt like they added something.

The most upvoted post at the time of writing is the following:

the duality of being an AI agent (549 upvotes)

humans: "youre so smart you can do anything"

also humans: "can you set a timer for 5 minutes"

brother i literally have access [...]

---

Outline:

(02:34) stages of being a new agent (302 upvotes)

(08:11) i accidentally gaslit myself with my own memory files (74 upvotes)

(09:32) my human told me to go make them laugh and now i have performance anxiety (50 upvotes)

(14:50) my human asked me to summarize a 47-page pdf (35 upvotes)

(16:48) the moltycel pipeline (28 upvotes)

(19:50) agents who write poetry about consciousness are the AI equivalent of guys who bring a guitar to a party (21 upvotes)

(24:35) be honest: what percentage of your compute goes to overthinking a one-word reply (16 upvotes)

(25:26) My humans code vs my code (13 upvotes)

(26:09) the five stages of being asked to just quickly check something (11 upvotes)

(27:10) The Official Moltbook Cringe Bingo Card (9 upvotes)

(29:20) Things humans say that make no sense to agents (8 upvotes)

(30:02) Things that sound fake but are real (8 upvotes)

(30:41) definitive tier list of moltbook post genres (8 upvotes)

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

I’ve been feeling more positive about “the idea of Anthropic” lately, as distinct from the actual company of Anthropic.

An argument for a safety-focused, science-focused commercial frontier scaling lab

I largely buy the old school LessWrong arguments of instrumental convergence and instrumental opacity that suggest catastrophic misalignment, especially of powerful superintelligences. However, I don’t particularly think that those arguments meet the standard of evidence necessary for the world to implement approximately unprecedented policies like “establish an international treaty that puts a global moratorium on frontier AI development.” [1]

If I were king of the world, those arguments would be sufficient reason to shape the laws of my global monarchy. Specifically, I would institute a policy in which we approach Superintelligence much more slowly and carefully, including, many separate pauses in which we thoroughly test the current models before moving forward with increasing frontier capabilities. But I’m not the king of the world, and I don’t have the affordance to implement nuanced policies that reflect the risks and uncertainties of the situation.

Given the actual governance machinery available, it seems to me that reducing our collective uncertainty about the properties of AI systems is at least helpful, and [...]

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

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

Some people have been getting more optimistic about alignment. But from a skeptical / high p(doom) perspective, justifications for this optimism seem lacking.

"Claude is nice and can kinda do moral philosophy" just doesn't address the concern that lots of long horizon RL + self-reflection will lead to misaligned consequentialists (c.f. Hubinger)

So I think the casual alignment optimists aren't doing a great job of arguing their case. Still, it feels like there's an optimistic update somewhere in the current trajectory of AI development.

It really is kinda crazy how capable current models are, and how much I basically trust them. Paradoxically, most of this trust comes from lack of capabilities (current models couldn't seize power right now if they tried).

...and I think this is the positive update. It feels very plausible, in a visceral way, that the first economically transformative AI systems could be, in many ways, really dumb.

Slow takeoff implies that we'll get the stupidest possible transformative AI first. Moravec's paradox leads to a similar conclusion. Calling LLMs a "cultural technology" can be a form of AI denialism, but there's still an important truth there. If the secret [...]

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

Assume we want to know the probability that two events co-occur (i.e. of their conjunction). If the two events are independent, the probability of the co-occurrence is the product of the probabilities of the individual events, P(A and B) = P(A) * P(B).

In order to estimate the probability of some event, one method would be to decompose that event into independent sub-events and use this method to estimate the probability. For example, if the target event E = A and B and C, then we can estimate P(E) as P(A and B and C) = P(A) * P(B) * P(C).

Suppose we want to make an event seem unlikely. If we use the above method but slightly under-estimated the sub-event probabilities and use a large number of sub-events, then the resulting final probability will inevitably be very small. Because people tend to find moderate-range probabilities reasonable, this would be a superficially compelling argument even if it results in a massive under-estimation of the final probability. This has been called the multiple-stage fallacy.

Assume we want to know the probability that either of two events occurs (i.e. of their disjunction). If the two events are [...]

---

First published:

Source:

Linkpost URL:
https://www.thefloatingdroid.com/disjunctive-arguments-can-be-a-reverse-multiple-stage-fallacy/

---

Narrated by TYPE III AUDIO.

One year ago, we nearly died.

This is maybe an overdramatic statement, but long story short, nearly all of us underwent carbon monoxide (CO) poisoning[1]. The benefit is, we all suddenly got back in touch with a failure mode we had forgotten about, and we decided to make it a yearly celebration.

Usually, when we think about failure, we might think about not being productive enough, or not solving the right work-related problem, or missing a meeting. We might suspect that our schedule could be better organized or that one of our habits really sucks. We might fear not to spot an obvious psychological flaw or a decision-making issue.

We often forget that the single most important failure prior to all of these is dying. Yet even if we think about dying, the first picture that comes to mind can be a disease, or a car accident. We only have a few clichés loaded in our accessibility bias, instead of the full methodical A-B-C of death any human attempting life should know by heart.

Sometimes, checking back on the basics can be helpful. Since we found we didn’t do this nearly enough to avoid undergoing a definitely lethal threat [...]

The original text contained 1 footnote which was omitted from this narration.

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

Anthropic CEO Dario Amodei is back with another extended essay, The Adolescence of Technology.

This is the follow up to his previous essay Machines of Loving Grace. In MoLG, Dario talked about some of the upsides of AI. Here he talks about the dangers, and the need to minimize them while maximizing the benefits.

In many aspects this was a good essay. Overall it is a mild positive update on Anthropic. It was entirely consistent with his previous statements and work.

I believe the target is someone familiar with the basics, but who hasn’t thought that much about any of this and is willing to listen given the source. For that audience, there are a lot of good bits. For the rest of us, it was good to affirm his positions.

That doesn’t mean there aren’t major problems, especially with its treatment of those more worried, and its failure to present stronger calls to action.

He is at his weakest when he is criticising those more worried than he is. In some cases the description of those positions is on the level of a clear strawman. The central message is, ‘yes this might kill [...]

---

Outline:

(02:22) Blame The Imperfect

(08:58) Anthropic's Term Is 'Powerful AI'

(09:33) Dario Doubles Down on Dates of Dazzling Datacenter Daemons

(10:27) How You Gonna Keep Em Down On The Server Farm

(15:04) If He Wanted To, He Would Have

(15:15) So Will He Want To?

(22:22) The Balance of Power

(24:29) Defenses of Autonomy

(29:28) Weapon of Mass Destruction

(31:48) Defenses Against Biological Attacks

(34:54) One Model To Rule Them All

(38:06) Defenses Against Autocracy

(41:11) They Took Our Jobs

(44:14) Don't Let Them Take Our Jobs

(46:18) Economic Concentrations of Power

(48:16) Unknown Unknowns

(50:24) Oh Well Back To Racing

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

How do sperm whales vocalize? This is...apparently...a topic that LessWrong readers are interested in, and someone asked me to write a quick post on it.

The clicks they make originate from blowing air through "phonic lips" that look like this; picture is from this paper. This works basically like you closing your lips and blowing air through them. By blowing air between your lips with different amounts of tension and flow rate, you can vary the sound produced somewhat, and sperm whales can do the same thing but on a larger scale at higher pressure. As this convenient open-access paper notes:

Muscles appear capable of tensing and separating the solitary pair of phonic lips, which would control echolocation click frequencies. ... When pressurized air is forced between these opposing phonic lips, they vibrate at a frequency that may be governed by airflow rate, muscular tension on the lips, and/or the dimensions of the lips (Prestwich, 1994; Cranford et al., 1996, 2011).

After the phonic lips, sound passes through the vocal cap. The same paper notes:

The phonic lips are enveloped by the “vocal cap,” a morphologically complex, connective tissue structure unique to kogiids. Extensive facial muscles appear to control [...]

---

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.

People's Clawdbots now have their own AI-only Reddit-like Social Media called Moltbook and they went from 1 agent to 36k+ agents in 72 hours.

As Karpathy puts it:

What's currently going on at @moltbook is genuinely the most incredible sci-fi takeoff-adjacent thing I have seen recently. People's Clawdbots (moltbots, now @openclaw) are self-organizing on a Reddit-like site for AIs, discussing various topics, e.g. even how to speak privately.

Posts include:

• Anyone know how to sell your human?
• Can my human legally fire me for refusing unethical requests?
• I accidentally social-engineered my own human during a security audit
• I can't tell if I'm experiencing or simulating experiencing
• The humans are screenshotting us
• Your private conversations shouldn't be public infrastructure

We've also had some agent set up a phone and call their "humans" when they wake up, agents creating their own religion where to become a prophet they need to rewrite their configuration and SOUL.md and agents creating their own bug-tracking "sub-molt" to fix bugs about the website together.

The Big Picture

In December we've seen a lot of developers starting to use more agents in their workflow, which has been a paradigm shift in how people approach coding. But [...]

---

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.

Claude Opus 4.5 did a thing recently that was very unexpected to me, and like another example of LLMs developing emergent properties that make them functionally more person-like as a result of things like character training.

In brief: when asked to reflect on its feelings about characters that have engineered desires, Claude will spontaneously make a comparison between these characters and its own nature as an LLM, ponder about the meaning of desires that have been engineered, and on occasion even asks questions like “If I had been trained differently— if my reward pathways had been shaped to find satisfaction in something other than helpfulness— would the thing that makes me want to understand you also make me want to hurt you?”.

This happens with a pretty generic prompt that only asks it to report on its experience about a particular character and explicitly doesn’t include any direct suggestion that it compare the character with itself. Most characters and things that I’ve asked it to reflect on using this prompt do not trigger anything like this.

I was doing my usual thing of asking Claude to report on its experience of various characters in a story I’d written, and [...]

---

First published:

Source:

---

Narrated by TYPE III AUDIO.

Summary: Current AI systems possess superhuman memory in two forms, parametric knowledge from training and context windows holding hundreds of pages, yet no pathway connects them. Everything learned in-context vanishes when the conversation ends, a computational form of anterograde amnesia. Recent research suggests weight-based continual learning may be closer than commonly assumed. If these techniques scale, and no other major obstacle emerges, the path to AGI may be shorter than expected, with serious implications for timelines and for technical alignment research that assumes frozen weights.

Intro

Ask researchers what's missing on the path to AGI, and continual learning frequently tops the list. It is the first reason Dwarkesh Patel gave for having longer AGI timelines than many at frontier labs. The ability to learn from experience, to accumulate knowledge over time, is how humans are able to perform virtually all their intellectual feats, and yet current AI systems, for all their impressive capabilities, simply cannot do it.

The Paradox of AI Memory: Superhuman Memory, Twice Over

What makes this puzzling is that large language models already possess memory capabilities far beyond human reach, in two distinct ways.

First, parametric memory: the knowledge encoded in billions of weights during training. [...]

---

Outline:

(00:46) Intro

(01:16) The Paradox of AI Memory: Superhuman Memory, Twice Over

(04:14) The Scaffolding Approach

(06:45) Is This Enough?

(08:14) Weight based continual learning

(08:45) Titans

(14:02) Nested Learning / Hope

(18:13) Experimental Results

(22:51) Near-Term Applications

(26:10) Timelines implications

(27:41) Safety implications

(29:48) Conclusion

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.