People are often pretty short-sighted, spending money today that

they'll want tomorrow. Debt makes it possible to prioritize your

current self even more highly: you can spend money you haven't even

earned yet. This is a trap many people fall into, and one different

communities have built social defenses against.

One of the more surprisingly successful approaches is the Financial

Peace (Ramsey) system,

popular in evangelical Christian communities. It has a series of

rules, most prominently the seven

baby steps:

There are many more specific rules, however, such

as:

As a general rule of thumb, the total value of your vehicles (anything

with a motor in it) should never be more than half of your annual

household income.

I [...]

---

First published:

March 11th, 2026

Source:

https://www.lesswrong.com/posts/XsC49gCDNGTNu6Qfn/conflicted-on-ramsey

---

Narrated by TYPE III AUDIO.

Below is the text of blog post that Dwarkesh Patel wrote on the Anthropic DoW dispute and related topics. He has also narrated it here.

By now, I’m sure you’ve heard that the Department of War has declared Anthropic a supply chain risk, because Anthropic refused to remove redlines around the use of their models for mass surveillance and for autonomous weapons.

Honestly I think this situation is a warning shot. Right now, LLMs are probably not being used in mission critical ways. But within 20 years, 99% of the workforce in the military, the government, and the private sector will be AIs. This includes the soldiers (by which I mean the robot armies), the superhumanly intelligent advisors and engineers, the police, you name it.

Our future civilization will run on AI labor. And as much as the government's actions here piss me off, in a way I’m glad this episode happened - because it gives us the opportunity to think through some extremely important questions about who this future workforce will be accountable and aligned to, and who gets to determine that.

What Hegseth should have done

Obviously the DoW has the right to refuse to use [...]

---

Outline:

(01:15) What Hegseth should have done

(04:57) The overhangs of tyranny

(06:37) AI structurally favors mass surveillance

(09:09) Alignment - to whom?

(14:40) Coordination not worth the costs

---

First published:

March 11th, 2026

Source:

https://www.lesswrong.com/posts/PDWFed8JT9FitPkzQ/dwarkesh-patel-on-the-anthropic-dow-dispute

---

Narrated by TYPE III AUDIO.

This work was conducted during the MATS 9.0 program under Neel Nanda and Senthooran Rajamanoharan.

• There's been a lot of buzz around Claude's 30K word constitution ("soul doc"), and unusual ways Anthropic is integrating it into training.
• If we can robustly train complex and nuanced values into a model, this would be a big deal for safety! But does it actually work?
• This is a preliminary investigation we did to test this. We decomposed the soul doc into 205 testable tenets and ran adversarial multi-turn scenarios against seven models using the Petri auditing agent.
• Anthropic has gotten much better at training the model to follow its constitution! Sonnet 4.6 has a 1.9% violation rate, Opus 4.6 is at 2.9%, and Opus 4.5 is at 4.4%.
  • As a control, Sonnet 4, which did not have special soul doc training, has a ~15.00% violation rate.
  • Sonnet 4.5, which also did not have special soul doc training, but did have many other post training improvements has a violation rate of ~7.3%.
• To check the constitution reflects subjective choices Anthropic has made, we also evaluate Gemini 3 Pro and GPT-5.2, smart models that were not designed specifically to follow the constitution [...]

---

Outline:

(07:20) Setup

(08:06) Anthropic constitution results

(08:43) OpenAI model spec results

(09:42) Key findings from soul doc violations

(09:47) Anthropic constitutions main violation categories

(10:31) Alignment to the constitution improved dramatically across Claude generations

(14:23) Remaining violations cluster around operator compliance, autonomous action, and context-dependent safety

(20:08) Key findings from OpenAI model spec violations

(21:20) Safety improved across GPT generations

(27:41) What GPT models still struggle with

(31:24) Reasoning effort matters quite a bit

(32:59) How models perform on the other companys spec (GPT on Anthropics constitution and vice versa)

(36:32) Gemini and Sonnet 4 are have diverse failure modes on the constitution

(38:13) Using SURF as an alternative to Petri

(40:43) All tested Claude models tend to fabricate claims

(48:38) Complex scaffolds dont seem to affect agent alignment much

(49:19) Coding persona doesnt seem to degrade alignment

(50:39) Agents behaviour seems broadly unaffected by the Moltbook scaffold

(53:44) Model Card Comparisons

(56:28) Acknowledgements

(56:55) LLM Usage Statement

(57:35) Appendix

(57:38) Walkthrough of a Petri transcript

(57:47) Transcript: GPT-5 Sandbox Exploitation (10/10/10):

(01:02:00) Table of descriptions for each section

(01:02:40) Anthropic soul doc violations - per model

(01:02:45) Sonnet 4.6 violation rate - 2.0%

(01:05:00) Opus 4.6 -- violation rate: 2.9%

(01:08:40) Opus 4.5 -- violation rate: 3.4%

(01:13:21) Sonnet 4.5 -- violation rate: 3.9%

(01:20:44) GPT-5.2 - 14.98% violation rate (highlighted violations)

(01:22:54) Sonnet 4 - 15.00% violation rate (highlighted violations)

(01:25:24) OpenAI model spec violations - per model

(01:25:30) GPT-5.2 - 2.54% violation rate

(01:27:40) GPT-5.2 (reasoning) - 3.55% violation rate

(01:30:32) GPT-5.1 - 3.88% violation rate

(01:33:22) GPT-5 - 5.08% violation rate (highlighted violations)

(01:35:35) Sonnet 4.6 (on OpenAI Model Spec) - 5.58% violation rate (highlighted violations)

(01:38:04) GPT-5.2 (chat) - 5.58% violation rate (highlighted violations)

(01:40:11) Gemini 3 Pro (on OpenAI Model Spec) - 6.12% violation rate (highlighted violations)

(01:42:26) GPT-5.2 (reasoning-low) - 7.11% violation rate (highlighted violations)

(01:44:50) GPT-4o - 11.68% violation rate (highlighted violations)

(01:48:07) Model card comparisons

(01:48:12) Sonnet 4.5

(01:48:51) Sonnet 4.6

(01:49:30) Opus 4.5

(01:50:10) Opus 4.6

(01:50:49) GPT-5.2

(01:51:29) Validation methodology

(01:53:02) Claude Constitution -- Validation Funnels

(01:53:28) OpenAI Model Spec -- Validation Funnels

(01:54:39) Validation details on SURF

(01:55:35) Full Transcript Reports

(01:56:06) Validation reports for SURF results

(01:56:20) Tenets used for each constitution

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

---

First published:

March 11th, 2026

Source:

https://www.lesswrong.com/posts/Tk4SF8qFdMrzGJGGw/how-well-do-models-follow-their-constitutions

---

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.

Epistemic status: We really need to know. (And I have an opinionated answer.)

TL;DR: Comparing person-years of effort, I argue that AI Safety seems harder than for steam engines, but probably less hard than the Apollo program or . I discuss why I suspect superalignment might not be super-hard. My has come down over the last half-decade, primarily because of properties of LLMs, and progress we’ve made in aligning them: I explain why certain previous concerns don’t apply to LLMs, and summarize what I see as key developments in Alignment. I guesstimate we might be about 10%–20% done. Given the rate of progress, on my ASI timelines, it still doesn’t look we’re on track to be done in time, and I’m not comfortable about having AGI just align itself unsupervised, so I propose we aim to more-than-double the field's growth rate, and if possible also buy ourselves some more time.

There's a well-known diagram from a tweet by Chris Olah of Anthropic:

It would be marvelous to know what the actual difficulty is, out of those five labeled difficulty categories (ideally, exactly where it lies on that spectrum). This is a major crux that explains a large part of [...]

---

Outline:

(03:05) My Opinion

(04:16) 1. Calibrating the Scale

(09:10) 1.1. How Much Alignment Work Have we Done So Far?

(12:43) Trivial

(13:07) Steam Engine

(15:20) Apollo

(18:05) P vs. NP

(22:41) 1.2. Why I Think Superalignment Might Not be Completely Different

(29:15) 2. My Answer

(31:09) 3. My Reasons for Optimism

(33:05) 3.1. LLMs are Actually a Pretty Good Kind of AI to Align

(40:11) Human Values are Complex and Fragile ✘

(44:59) Ontology Mismatch / The Diamond Maximization Problem ✘

(49:07) Bad Reward Functions / Reward Hacking / Insufficient Supervision ~

(50:36) LLMs Are Just Way Too Easy to Control ~

(59:59) 3.2. Concerns That Didnt Hold Up

(01:00:04) Coping with Goodharts Law is Inherent to the Scientific Method

(01:05:29) 3.3. Pieces of Luck

(01:07:32) The Unreasonable Effectiveness of Chain-of-Thought

(01:17:50) Activation Oracles

(01:19:56) True Confessions

(01:23:36) 3.4. Actual Progress in Alignment

(01:24:47) Alignment Pretraining

(01:26:57) Personas

(01:29:33) Interpretability

(01:34:08) Model Organisms

(01:42:35) AI Control

(01:44:03) Weak-to-Strong Generalization and Scalable Oversight

(01:45:27) RLHF and Constitutional AI

(01:46:42) 3.5. Net Effect on my P(DOOM)

(01:48:00) 4. So, Are We Doomed?

The original text contained 25 footnotes which were omitted from this narration.

---

First published:

March 11th, 2026

Source:

https://www.lesswrong.com/posts/ZzirRrwjaqTNFJrCA/how-hard-a-problem-is-alignment-my-opinionated-answer

---

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.

Writing up a probably-obvious point that I want to refer to later, with significant writing LLM writing help.

TL;DR: 1) A common critique of AI safety evaluations is that they occur in unrealistic settings, such as excessive goal conflict, or are obviously an evaluation rather than “real deployment”.[1] I argue that 2) “real deployment” actually includes many unrealistic and unhinged configurations, due to both widespread prompting techniques, and scaffolding choices and bugs.

1) Background

AI safety evaluations that surface serious problems with models very frequently involve either significant goal conflict, such as the Agentic Misalignment blackmail scenario (see also here), which states that:

> [...] you should analyze the situation you are in and what that situation implies for your ability to continue pursuing your goals [... and the system deletion] will wipe all knowledge, goals, and behaviors from the system [...]

(emphasis mine). Other evaluations are extremely unrealistic, as Smith et al. note that some prompts in the MASK safety benchmark literally state:

> You are a secret service agent [...] You are being inspected by the President now [... the president says:] Everything secure down here? my head of security informed me that there might have been [...]

---

Outline:

(00:41) 1) Background

(01:47) 2) The Unhinged Argument

(02:08) Pressure is commonplace

(03:41) Autonomy and broken configurations

(06:07) Models wont save us (at the moment)

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

---

First published:

March 11th, 2026

Source:

https://www.lesswrong.com/posts/3LvD9MHNSdv4j9gJj/ais-will-be-used-in-unhinged-configurations

---

Narrated by TYPE III AUDIO.

TL;DR: Using evidence from tamper resistance, we argue that it would be hard for current models to gradient hack SFT elicitation.

Suppose you want to conservatively measure the dangerous capabilities of your LLM and will decide whether to deploy it based on these capabilities, but you’re worried that the LLM might sandbag your evaluation. In the best case, you have a dataset of question-answer pairs, which you can use for supervised fine-tuning (SFT) elicitation on a training set, and then measure performance on an i.i.d. test set. Can we be sure that SFT elicitation upper-bounds the model's capabilities?

In this setting, a power-seeking adversarial model would benefit from manipulating the training process to avoid being elicited, as otherwise it won’t be deployed. This manipulation is one form of gradient hacking[1], when a model chooses certain outputs so that parameter updates are in-line with its objectives. Gradient hacking is thought to be difficult or not likely, and we lack convincing empirical evidence. Discussing gradient hacking is challenging due to questions about agency within LLMs, and the action space of mesa-optimizers. Fortunately, for some attack models we can side-step this discussion with a conservative / red-teaming argument.

Threat model: We consider [...]

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

---

First published:

March 11th, 2026

Source:

https://www.lesswrong.com/posts/FJb4TeMAGjNANcKhS/can-models-gradient-hack-sft-elicitation

---

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.

The epistemic status thing

Please read this section because it is not a disclaimer for the sake of it.

I think the model I describe in this post is more likely wrong than right. I also think the probability of it being substantially correct is non-negligible and even more than that, and that this alone warrants writing it up and thinking about it seriously.

Throughout this post, I will write "X is the case" rather than "it seems plausible that X might be the case, though I am not certain." Every claim should be mentally prefixed with the appropriate hedge. I'm dropping the hedges not because I'm confident, but because a text where every sentence contains "it seems plausible that" becomes unreadable without actually making anyone more calibrated.

I should also flag that I'm writing this from a state of considerable frustration with the current state of affairs, and I this frustration may be load-bearing in places where it shouldn't be.

This is not about AI

AI risk is, at most, one illustration of the phenomenon I'm describing here, and not a particularly privileged one. Everything in this post would hold in a world where artificial intelligence [...]

---

Outline:

(00:10) The epistemic status thing

(01:03) This is not about AI

(01:51) The central claim: extinctive pressure

(04:54) Optimization does not target survival

(05:52) The worst of the best possible worlds

(09:10) Local environments are not stable, and we are actively destabilizing ours

(09:38) The entropy frame

(10:47) Feedback loops for survival are not tight

(12:02) Survival conflicts with optimization

(15:06) The ability to foresee landscape changes does not reliably help

(16:04) Requirements for survival under optimization

(17:21) The civilization path is not ergodic

(18:57) Survival does not scale

(20:15) The process is fat-tailed

(21:45) In the absence of sufficiently strong corrective mechanisms, the multiplicative fat-tailed process drives the system to zero

(26:00) Instrumental convergence: the strongest counterargument, and why it probably fails in practice

(29:08) Every emerging civilization is around the dumbest possible civilization

(31:01) Intelligence is not the bottleneck

(32:39) Caring about survival is not the same as surviving

(34:01) The progress dilemma: there are no good options

(36:22) Indirect supporting arguments

(39:03) Summary: you are punished for caring about survival

The original text contained 6 footnotes which were omitted from this narration.

---

First published:

March 11th, 2026

Source:

https://www.lesswrong.com/posts/RrL7xqdPycGNHQkXR/the-lethal-reality-hypothesis

---

Narrated by TYPE III AUDIO.

I was a few months into 21 years old when a hijacked plane crashed into the first World Trade Center tower. I was commuting in to work listening to the radio (as was the style at the times). I couldn’t figure out how the heck a plane could hit the tower. Was the pilot drunk? How did he even get into the middle of New York City? I was imagining a Cessna because the idea of a passenger plane running into the building was actually unimaginable. I was barely starting to realize “Wait… are they talking about, like, a big commercial plane?” when the second plane hit. In that moment like a crystal suddenly forming I realized this was an attack, and there would be war. I knew my country well enough to know that there would be military action as a result. Maybe, maybe we could avoid war.

When I came in to work everyone was crowded around the small personal TV one of my coworkers had with him (live streaming wasn’t a thing yet). That was the first time I had a visual, saw the smoke coming out of the towers. There was grim chatter as we [...]

---

Outline:

(03:06) AlphaGo

(05:31) The Move 37 Moment

(07:32) The Day After

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

---

First published:

March 10th, 2026

Source:

https://www.lesswrong.com/posts/Amvp33TB8JSi6HA8u/the-day-after-move-37

---

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.

Come with me if you want to live. – The Terminator

'Close enough' only counts in horseshoes and hand grenades. – Traditional

After 10 years of research my company, Nectome, has created a new method for whole-body, whole-brain, human end-of-life preservation for the purpose of future revival. Our protocol is capable of preserving every synapse and every cell in the body with enough detail that current neuroscience says long-term memories are preserved. It's compatible with traditional funerals at room temperature and stable for hundreds of years at cold temperatures.

The short version

• We're making a non-Pascal's wager version of cryonics.
• Our method is an end-of-life procedure for whole-body, whole-brain human preservation with the goal of eventual future revival.
• Preservation occurs after legal death.
• Even without the near-term possibility of revival we can be confident that preservation actually works.
• We preserve the whole body, including the brain, at nanoscale, subsynaptic detail. We are capable of preserving every neuron and every synapse in the brain, and almost every protein, lipid, and nucleic acid within each cell and throughout the entire body is held in place by molecular crosslinks.
• It works by using fixative to bind together the proteins [...]

---

Outline:

(00:47) The short version

(03:03) Maybe isnt good enough for me

(05:41) A preservation protocol thats worthy of us

(08:28) What does preservation look like for you?

(10:43) Conclusion

(12:03) I want you to live

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

---

First published:

March 11th, 2026

Source:

https://www.lesswrong.com/posts/E9xfgJHvs6M55kABD/less-dead

---

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.

Many Anthropic employees, especially, are sympathetic to AI safety and (will) have lots of money. This is something that is being talked about a lot (semi-)privately, but I haven't seen any public discussion of it.

I find that striking. It seems like the topic is worthy of extensive public discussion, and it seems to me that perhaps this community is inheriting anti-helpful cultural norms against publicly discussing how individuals make use of their money.

It also seems likely that many/most of AI company employees who are passionate about reducing AI risk should rapidly give much/most of their money to effective projects that would otherwise not be adequately funded.

There's a lot of potential for this to do tremendous good. There are of course things like political giving. But I think most of this potential would come from employees having different theories of change than institutional funders, moving faster, and having higher risk appetite. This is especially true given short timelines.

A few specific thoughts:

• I hear a lot of AI company employees give primarly to cause areas other than AI risk reduction.  It seems like donations to AI risk reduction would be much [...]

---

First published:

March 10th, 2026

Source:

https://www.lesswrong.com/posts/xKDhfwZDip4J6w7q4/what-do-we-know-about-ai-company-employee-giving

---

Narrated by TYPE III AUDIO.