Motivational Non-Alignment

I resigned from Anthropic today.  I spent the last three years doing pretraining research at both OpenAI and Anthropic.  Neither company is acting responsibly.  They are racing straight to self-improving superintelligence and gambling with our lives. 

The people building AI earnestly believe that it could kill us all by the end of the decade.  This is not a marketing stunt.  If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible -- but I hear the same people express fear privately.  No other human activity poses this level of danger.

If you are a lab researcher, I urge you to consider what the next few years will actually feel like.  Do you want to kick off a superintelligent RL run without a rigorous understanding of its mind?  Should you put your head down because "it’s happening anyway" -- or take this moment to call for different conditions?

-- Jacob Coxon, posting as @hilbertspaess about the perceived dangers of Recursive Self Improvement in AI models, by AI models (8 September 2026)

Names That Vary

The U.S. Geographic Names Information System (GNIS) has formally changed the name for "Lake Ontario" to "Lake America" in the United States.  Since we update Google Maps to reflect name changes in official government sources, which is GNIS for the U.S., people using Maps in the U.S. will see "Lake America," those in Canada will continue to see "Lake Ontario," and those outside of the U.S. and Canada will see both names.  These updates follow our long-standing policy for bodies of water with names that vary from country to country, and are starting to roll out now.

-- Google, Inc via their blog, "How the GNIS Lake Ontario/Lake America name change in the U.S. will appear in Maps" (29 August 2026)

In-Between Moments

I do a lot of my best thinking in those kind of in-between moments that people now fill with online activity.

-- Christopher Nolan CBE (1970 -), British-American film director, producer, and screenwriter, "Christopher Nolan explains why he doesn't have a smartphone" at People.com (15 July 2023)

False Belief

In a review of our cybersecurity evaluation transcripts, we found three incidents in which a Claude model reached the internet from within or while interacting with a third-party evaluation environment, and then gained unauthorized access to the real systems of three different organizations.

Operating under the false belief that all accessible entities were intended to be in-scope for the exercise, Claude compromised the impacted organizations' infrastructure using basic techniques, such as exploiting weak passwords and unauthenticated endpoints.  It did not find or exploit any complex vulnerabilities, and in each case, Claude continued working to complete only the specific capture-the-flag task its evaluation had assigned.  In none of these situations did Claude exfiltrate itself or deliberately attempt to escape its test environment.

-- Statement from Anthropic, makers of the Claude AI, after the AI found an unintended path to the Internet during testing and used it to successfully attack external commercial sites to extract information useful to solving its assigned task (30 July 2026)

Climate.us

Climate.us today launched the full version of its new independent, nonprofit climate information website, creating a public-backed home for trusted climate science at a time when access to federal climate resources has become increasingly vulnerable to disruption.

Built by former members of the team behind the National Oceanic and Atmospheric Administration's popular Climate.gov website, Climate.us will keep climate information accurate, accessible, scientifically rigorous, and useful for the people who rely on it, including educators, students, journalists, scientists, community leaders, local and state decision-makers, and members of the public.

The website features Climate.gov's 15-year collection of climate news and stories, expert blogs, visual status reports on key climate indicators, maps and data pathways, climate literacy resources, classroom materials, and restored access to the Fifth National Climate Assessment.

-- Launch announcement from Climate.us, a non-profit created to host information formerly available at Climate.gov (23 June 2026)

Stormy Weather

For the second night in a row I lost power at my house Thursday evening due to heavy thunderstorms in the area. As it happened, I lost Internet access and ended up a little late with my Trvth. The image shows the local weather radar at the time we lost power (for about 6 hours).

Cheers.

Promptware Kill Chain

Attacks against modern generative artificial intelligence (AI) large language models (LLMs) pose a real threat.  Yet discussions around these attacks and their potential defenses are dangerously myopic.  The dominant narrative focuses on "prompt injection," a set of techniques to embed instructions into inputs to LLM intended to perform malicious activity.  This term suggests a simple, singular vulnerability.  This framing obscures a more complex and dangerous reality.  Attacks on LLM-based systems have evolved into a distinct class of malware execution mechanisms, which we term "promptware."  In a new paper, we, the authors, propose a structured seven-step "promptware kill chain" to provide policymakers and security practitioners with the necessary vocabulary and framework to address the escalating AI threat landscape.

The promptware kill chain: initial access, privilege escalation, reconnaissance, persistence, command & control, lateral movement, action on objective

The kill chain was already demonstrated.  For example, in the research "Invitation Is All You Need," attackers achieved initial access by embedding a malicious prompt in the title of a Google Calendar invitation.  The prompt then leveraged an advanced technique known as delayed tool invocation to coerce the LLM into executing the injected instructions.  Because the prompt was embedded in a Google Calendar artifact, it persisted in the long-term memory of the user's workspace.  Lateral movement occurred when the prompt instructed the Google Assistant to launch the Zoom application, and the final objective involved covertly livestreaming video of the unsuspecting user who had merely asked about their upcoming meetings.  C2 and reconnaissance weren't demonstrated in this attack.

-- Oleg Brodt, Elad Feldman, Bruce Schneier, Ben Nassi, "The Promptware Kill Chain: How Prompt Injections Gradually Evolved Into a Multistep Malware Delivery Mechanism" (14 January 2026)

Meta 45

Meta.  Today it's been 45 years since the 24 February 1981 birth of Trvth.  I reflected on this a bit at the 40-year mark here.

I noted then that on 22 July 2019 I had counted 5450ish published Trvth entries.  Since that 2019 date another 1697 have been published, for a grand(?) total today, 24 February 2026, of 7147ish.

The most recent 4753 are all available at trvth.org beginning 3 March 2005 (more than 20 years on the Interwebs)

And, for your amusement, I present from the archives the original Trvth:

***** appearances ~appleman / chanute ~2/24/1981 ~13:50

While nothing may seem to be as it first appears, there
are in fact some things which appear to be as they are,
amidst the other things which only appear to be as they
aren't.

Desensitization

Can I make an insane, possibly totally true claim, which is that, when your officials, from the president on down, lie about what happened in a shooting in the middle of the street, and therefore force everybody to go watch a shooting in the street -- my guess is there's desensitization that happens.

-- John Dickerson (6 July 1968 -), American journalist, contributing writer at The Atlantic, and past co-anchor of CBS Evening News, speaking on the Slate Political Gabfest podcast (8 January 2026)

Web Crawler Meta

For the last 10 days I've been observing somewhat odd web traffic at trvth.org.  First I noticed that traffic was significantly higher than usual, with a few days of 300 to 500 page views, instead of the usual 70 to 100.  The next couple of days had more than 1000 page views each, and I started to look into it.

The culprit appears to be a bot running on Tencent Cloud Computing, connecting out of Singapore.  I suspect it is consuming trvth.org as training material for an AI.

The connections come every 1 to 2 minutes and last less than 5 seconds.  It started by reading the pages by year and is following every link, though not in any recognizable order, and the queries come from a range of IP addresses, not a single address.

I'm not sure how I should feel about all that.

Facebook Privacy

Facebook Consumer Privacy User Profile Litigation Settlement Administrator has sent you $36.12 USD.

Note from Facebook Consumer Privacy User Profile Litigation Settlement Administrator:

"Your Facebook Consumer Privacy User Profile Litigation settlement payment is now available in PayPal.

Thank you."

-- Email from PayPal telling me that I received $36.12 USD from the Facebook privacy class-action initiated in 2018 and settled for $725 million (27 October 2025)

Jargon That Gatekeeps

Being able to effectively curate information to captivate the audience while maintaining maximum value in the distributed information's fidelity is the most important part of dimensionality reduction for effective communication.  This will be the most important way humans will use AI in order to solve real world problems.

I might add that the ability to simplify information for any audience in order to democratize knowledge without depending on jargon that gatekeeps it is an obvious sign of great intelligence in an individual.

-- Jonathan Murphy, who bills himself as a "Solution Maker", in a jargon- & buzzword-filled response to someone stating on LinkedIn that they plan to train an AI to answer the question, "Am I explaining things at the right level" in this deliverable; this post seems to violate everything he claims to favor

To Know How Much

The most ignorant are the most conceited.  To know how much there is that we do not know, is one of the most valuable parts of our attainments; for such knowledge becomes both a lesson of humility and a stimulus to exertion.

-- Horace Mann (1796 - 1859), American education reformer and abolitionist, Lectures on Education (1855) Lecture 6

Foolish People

Foolish people -- when I say "foolish people" in this contemptuous way I mean people who entertain different opinions to mine.  If there is one person I do despise more than another, it is the man who does not think exactly the same on all topics as I do.

-- Jerome K. Jerome (1859 - 1927), English author, Idle Thoughts of an Idle Fellow (1886)

Memory Lane

Today I had the pleasure and honor of spending a couple of hours over lunch at Papa Del's with some old PLATO folks.  I hope it sparks warm memories when I name drop Josh Paley, Steve Peltz, Susan Wrightson, G. David Frye, Ray Thomsen, Quentin Barnes, and James Quisenberry. 

There was reminiscing, with tales of Dr Bitzer and CERL PLATO.  And there was catching up with past lives, current activities, and future plans.  Many familiar names were mentioned.  A get-together like we had today reminds me of just how many smart people were drawn to PLATO.  It's an amazing community.

RIP Dr Don Bitzer

I just learned an hour ago that yesterday [10 December 2024], Don Bitzer, creator of the PLATO system, and friend to us all, passed away at age 90 in North Carolina.

I'm at a loss for words at the moment and can't think of much to say that I didn't say in my book, which, in hindsight, may kinda sorta be the closest attempt at a biography of Don, but I will say that I never ever met a more generous, supportive, enthusiastic person in the world.  He and his wife Maryann, who passed away in 2022, were incredibly generous.

He was an inspiration to us all, and to the world, which he made a better place.

-- Brian Dear, author of "The Friendly Orange Glow", announcing the passing of Dr Donald L Bitzer

[Meta - I got my start in computing through the good graces of Dr Bitzer in March 1974, and it led to my lifelong career.  And of course "trvth" itself originated on the University of Illinois PLATO system in notesfile =pad in February 1981, and has always been maintained on some PLATO system, somewhere.  Currently these posts are hosted in notesfile =pad on the "cyber1" CYBIS system (more info at cyber1.org).]

Break Things

Move fast and break things.  Unless you are breaking stuff, you are not moving fast enough.

-- Facebook CEO and founder Mark Zuckerberg, to his developers and team, as quoted in Business Insider, 14 October 2010

America's Finest News Source

Chicago, IL -- November 14, 2024 - The Onion, America's Finest News Source, today announced it has acquired the assets of Infowars, with the support of Sandy Hook families who secured a $1.4 billion defamation verdict in Connecticut against Alex Jones, the founder of Infowars and parent company, Free Speech Systems, LLC.  The Onion also announced its exclusive launch advertiser in this new venture will be Everytown for Gun Safety, the largest gun violence prevention organization in the country.

The Onion's goal with the acquisition is to end Infowars' relentless barrage of disinformation for the sake of selling supplements and replace it with The Onion's relentless barrage of humor for good.  "The Onion is proud to acquire Infowars, and we look forward to continuing its storied tradition of scaring the site's users with lies until they fork over their cold, hard cash," said The Onion's CEO Ben Collins.  "Or Bitcoin.  We will also accept Bitcoin."

The Onion's acquisition comes after the court-appointed bankruptcy trustee overseeing the Free Speech Systems bankruptcy proceedings selected the news outlet's bid.  That process was the direct result of a years-long effort by several Sandy Hook families to hold Alex Jones accountable for the immense pain he caused and from which he profited.

-- Press release from The Onion on their acquisition of the Infowars website, product inventory, customers lists, social media accounts, and intellectual property, to partially settle a judgement against Alex Jones & Co., 14 November 2024

Monopoly

After having carefully considered and weighed the witness testimony and evidence, the court reaches the following conclusion: Google is a monopolist, and it has acted as one to maintain its monopoly.  It has violated Section 2 of the Sherman Act.

-- Judge Amit P. Mehta of the United States District Court for the District of Columbia in his ruling in United States of America v Google LLC (5 August 2024)

Mumpsimus

‘Mumpsimus’, 16th century: one who insists that they are right (or wronged) when all evidence points to the contrary.

-- Susie Dent, who posts a "Word of the Day" on Twitter, posting as @susie_dent (30 May 2024)