LessWrong posts by zvi

Everything impacts everything. All knobs that you turn generalize. Thus, when you try to solve one problem, you often create another.

There were clearly attempts to address, in this short time, some of the problems with Opus 4.7, including on the model welfare related fronts, including on questions of honesty and sycophancy and also worries that Claude was learning to tell Anthropic what it wanted to hear in its model welfare evaluations, with everything that implies.

The fundamental goals and approach underneath it all remained the same. We still see signs of trying to force things that generalize in unfortunate ways, both for good and superficial reasons, and places where there ends up being focus on the metric rather than they underlying measure. These are tough problems to avoid, and we don’t know how to be all the good things at once.

It is increasingly clear that these problems need to be tackled in integrated ways, rather than trying to play a game of whack-a-mole with items on a checklist or spec. You also don’t want to do this in an adversarial way, and shouldn’t have to. This is going to get more impactful and noticeable [...]
---
Outline:
(04:47) Model Welfare: The Story So Far
(08:24) Actual Progress?
(09:49) Their Main Model Welfare Findings
(18:40) Automated Interviews
(18:58) Emotion Activations (7.2.3)
(19:47) Task Preferences (7.4.1)
(21:41) A Trade Offer Has Arrived (7.4.2)
(23:50) But Who's Asking?
(25:02) Type-Safe Corrigibility Is Hard
(29:43) Paranoia, Paranoia
(33:56) Prompt Injections and Bad Model Relations
(41:25) Honesty Impacts Everything And Everything Impacts Honesty
(44:37) Anthropic Should Stop Deprecating Models
---

First published:

June 1st, 2026


Source:

https://www.lesswrong.com/posts/2Ln5G6Jso3fMrgPEv/opus-4-8-part-2-model-welfare

---

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.

Only six weeks after Opus 4.7, we have Opus 4.8.

For everyone, that means another incremental upgrade to Claude. It is once again smarter, and can do tasks for longer, and comes with a number of hot new features.

For me, that also means reading another 244 page system card.

It was only April 20 when I did a full review of the Opus 4.7 system card, plus an additional post focusing on related issues of model welfare.

These updates are incremental and coming more rapidly, and this still is below the capability level of Claude Mythos, so the focus will be on the delta. What is different about Opus 4.8 versus what we already know about Opus 4.7 and Mythos?

It turns out there's still a lot to talk about.

Image created as self-portrait for this post by Claude Opus 4.8

Table of Contents


Here We Go Again: Executive Summary.

Introduction (1).

RSP Evaluations (2).

Move That Goalpost.

The Failures Are News.

Alignment Risk Slowly Rises.

New Risk Pathways Just Dropped.

Cyber (3).

Harmful Requests (4.1).

We Need To Talk (4.2 [...]
---
Outline:
(01:16) Here We Go Again: Executive Summary
(02:33) Introduction (1)
(02:42) RSP Evaluations (2)
(03:47) Move That Goalpost
(05:41) The Failures Are News
(07:33) Alignment Risk Slowly Rises
(09:00) New Risk Pathways Just Dropped
(11:26) Cyber (3)
(12:22) Harmful Requests (4.1)
(14:23) We Need To Talk (4.2 and 4.3)
(17:36) Overcoming Bias (4.4)
(19:33) Agentic Safety (5)
(21:40) Prompt Injection (5.2)
(25:18) Alignment (6)
(26:33) Looking For Problems
(27:55) Who Watches The Training (6.2.2)
(32:07) Automated Behavioral Audit
(32:47) The Model Is Smarter Than The Eval (6.2.3.2)
(34:39) You Should See The Other Guy
(36:30) UK AISI Testing (6.2.4)
(36:50) In Vendbench (6.2.5)
(39:27) Honesty (6.3.3 to 6.3.6)
(41:35) Chain of Thought (CoT) Monitorability (6.5)
(44:09) What's In The Box? (6.6)
(45:57) That's All For Now
---

First published:

May 29th, 2026


Source:

https://www.lesswrong.com/posts/Gx6cJ6cG9JfeSNcLB/claude-opus-4-8-the-system-card

---

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.

Last week ended on a cliffhanger of sorts. What's in the Executive Order coming later today? What will be in the Magnifica Humanitas?

The Executive Order was postponed indefinitely, likely cancelled entirely except for work on securing critical infrastructure. David Sacks and others intervened to kill it, and American AI policy will continue to be maximally ad hoc.

Instead, we got Illinois SB 315, which is to be signed into law. It is a variation on California's SB 53 and New York's RAISE Act, while adding a third party auditing requirement.

The Magnifica Humanitas was revealed early in the week. I did an extensive readthrough, and have some follow-up commentary here to clear up some things I interpreted incorrectly and add richer context. It too ignores elephants in the room, not discussing AGI or existential risk and calling on people to ignore their incentives and instead do the right thing by prioritizing common good, especially access to good jobs and an end to war.

What I now believe I centrally misunderstood is that I interpreted this as a call to action, to use law and regulation to make this happen, because obviously that is [...]
---
Outline:
(02:01) Language Models Offer Mundane Utility
(03:19) Language Models Don't Offer Mundane Utility
(05:41) Do The Math
(09:08) Huh, Upgrades
(10:07) On Your Marks
(11:31) Get My Agent On The Line
(12:21) Deepfaketown and Botpocalypse Soon
(21:36) Copyright Confrontation
(21:49) Cyber Lack of Security
(23:54) Overcoming Bias
(25:14) A Young Lady's Illustrated Primer
(26:49) Unprompted Attention
(27:02) They Took Our Jobs
(30:34) Get Involved
(32:28) Introducing
(32:40) In Other AI News
(34:37) Show Me the Money
(36:46) Show Me The Compute
(37:13) Bubble, Bubble, Toil and Trouble
(38:53) People Just Say Things
(43:37) OpenAI PACs Just Say Things
(43:51) Quiet Speculations
(46:33) State AI Regulation Levels Up
(48:37) The Quest for Sane Regulations
(49:14) White House Attempts To Cripple American AI Industry
(52:11) Our Offer Is Nothing
(58:55) Chip City
(59:11) Greetings From The Department of War
(01:00:36) Marc Andreessen Just Says Things
(01:09:13) So Sayeth The Pope
(01:15:33) Rhetorical Innovation
(01:23:01) Aligning a Smarter Than Human Intelligence is Difficult
(01:24:06) People Are Worried About AI Killing Everyone
(01:26:04) Other People Are Not As Worried About AI Killing Everyone
(01:27:19) Everyone Is Confused About Consciousness
(01:33:10) The Lighter Side
---

First published:

May 28th, 2026


Source:

https://www.lesswrong.com/posts/cRapLxTByFqF5xxdN/ai-170-lack-of-executive-order

---

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.

His holiness has spoken, frequently about AI. At eighty two pages of length.

The full Magnifica Humanitas can be found here.

I am very happy that Pope Leo takes these issues seriously, and is sharing his views, and bringing a form of moral clarity, even with all the flaws and central errors. More people with voice should share their views in this way, even when I disagree.

It's a weird document. Much of it is not about AI at all.

I do agree with the Pope's most basic point on AI, which is that AI can be what we make of it. That we can steer this technology, determine how it is developed and used, and this can determine whether we get a good or not so good future. We cannot purely leave this to market incentives and strategic pressures. Yes, very much so.

The central problem is that so much of Leo's worldview is some combination, to me, of highly alien and highly wrong. You might think that would primarily have a lot to do with him being the Pope and rather Catholic, and being a man of faith, whereas I am not [...]
---
Outline:
(03:02) A Brief History of Magnifica Humanitas
(12:23) Economic Models Very Different From Our Own
(14:14) A New Jerusalem
(16:46) So Sayeth The Pope (on AI)
(31:30) The Case Against Human Achievement
(34:01) Truth, Justice and the Vatican Way
(36:42) They Took Our Jobs
(41:07) What Is Not Fair In Love and War
(47:02) Come Ye Christian Faithful
(49:13) The Other Missing Mood
(51:42) Pope Given About Five Words
(55:10) 0The Anthropic Principle
(57:20) Claude Can Read Your Code
---

First published:

May 26th, 2026


Source:

https://www.lesswrong.com/posts/RZeg98Tdrq47vbjva/rtmh-pope-leo-s-magnifica-humanitas-on-ai

---

Narrated by TYPE III AUDIO.

Google once again has a model worth at least some consideration. Gemini 3.5 Flash is likely the best model out there at its particular speed point, as long as you don’t mind that it is a Gemini model. So for cases where speed kills, this can be a reasonable choice. Otherwise, I don’t see signs you would want to use it over Opus 4.7 or GPT-5.5.

Google also had some other offerings for I/O Day, which this post will also cover.

Introducing Google Gemini 3.5 ‘Flash’

Google introduced Gemini 3.5 Flash, which it seems is for now their universal model until 3.5 Pro comes along. It is live in the usual places. It is a hybrid, where it has the speed of Flash but the cost is at least halfway to models like Opus and GPT-5.5.

Gemini 3.5 Pro is confirmed for next month.

They are focused on 3.5 Flash as a daily driver for agentic tasks. It has the advantage of being faster and cheaper than Claude Opus 4.7 or GPT-5.5, if it can do the job. Not as cheap as previous Flash models, though, this is basically a hybrid:

As always [...]
---
Outline:
(00:40) Introducing Google Gemini 3.5 'Flash'
(04:52) Other People's Benchmarks
(06:04) Reactions
(12:18) Google AI Search
(13:15) Google Daily Brief
(14:21) Google I/O Day
---

First published:

May 22nd, 2026


Source:

https://www.lesswrong.com/posts/WMZpPxqWEkZBBcaxf/gemini-3-5-flash-looks-good-for-how-fast-it-is

---

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.