Ezra

[Ezra @Revan]

I don't know who needs to hear it but a tech ceo winning a power struggle for moral authority over the State would not actually be good and it's much better for the long term stability of civilization that we not open that particular can of worms for as long as possible.

[NIK @ns123abc]

🚨BREAKING: Pentagon is now calling Claude a threat to national security

>pentagon embeds claude in military systems via palantir
>january: claude used in maduro extraction, people got smoked
>anthropic exec calls palantir like “hey did our AI help kill people”

Defense Secretary Hegseth reportedly “close” to classifying Anthropic as supply chain risk. All defense contractors must certify zero anthropic or lose contracts.

CEO Amodei wants guardrails on autonomous weapons and mass surveillance of Americans.

Pentagon says “all lawful purposes” or nothing:
>“we will make them pay”

ITS HAPPENING

[Ezra @Revan]

The transitive property of frame mogging

[Kylie Cheung @kylietcheung]

I don't think this is how frame mogging works you can't avenge someone.... if you frame mog the person who frame mogged your friend (brutally) then you frame mogged both of them

[Καλός @realKalos]

Got my answer. Full body chills.

[JT @jiratickets]

Knew a dude who emailed the IT guy directly instead of submitting a ticket and they blew his shit smoove off

[Ezra @Revan]

@MurrayHillGuy1 Until dating services have financial incentives that align with partnering people instead of maximizing screen time for lonely men, none of these or other well intentioned ideas will be implemented.

As it is, you're the guy politely requesting a thermostat adjustment in hell.

[Dave DeCamp @DecampDave]

Maybe sit this one out

[Ezra @Revan]

@fleshsimulator Not that two things can't be true at the same time but I'm not interested in hearing about federal jackboots from people who never gave a shit about Vicki Weaver

[Ezra @Revan]

@bencroyderived @Landeur All guns will D if you are N enough

[Ezra @Revan]

@9mmsmg If sig making a striker fire fulfills a prophecy just imagine the butterfly effect if keltec made something useful

[Ezra @Revan]

@KelTecOfficial I mean. Yeah I believe you when you said you didn't ask a focus group.

[Ink Blot @inkblotistan]

looksmatched couple

[Ezra @Revan]

@Bricktop_NAFO Did you think there's like an Indiana Jones warehouse full of every car somebody's died in, preserved for future generations? Or, what, are you not convinced she's dead?

[Ezra @Revan]

@utacult The other thing captured well here, surely just a coincidence, is the total indifference and narcissism of doctors in the face of people with serious medical problems more difficult to diagnose than a broken arm.

Hmm.

[Ezra @Revan]

This is largely my experience both with using AI for software development and for personal work. Getting good results usually happens within the first 3 prompts or not at all, and depends heavily on narrowly defining success criteria, and usually brevity.

[Robert Youssef @rryssf]

This paper quietly explains why so many people feel like LLMs are “almost smart, but somehow wrong.”

The core claim in this paper is very uncomfortable: most failures are not about missing information. They are about misreading intent even when all the relevant context is present.

The authors show that LLMs are very good at mapping text to plausible responses, but surprisingly weak at inferring what the user is trying to achieve. Two prompts can contain nearly identical information, yet imply very different goals. Humans pick this up instantly. Models often do not.

The paper separates “context understanding” from “intent understanding.” Context is the literal content: entities, constraints, instructions. Intent is latent: priorities, tradeoffs, what matters most if things conflict. Current models optimize for surface-level alignment, not goal inference.

One experiment makes this painfully clear.

Users asked questions that could reasonably be interpreted as either exploratory or decision-oriented. The models answered confidently but chose the wrong mode at high rates, giving verbose explanations when users wanted a recommendation, or giving a decisive answer when users were clearly still exploring. The information was correct. The response was wrong.

Another failure mode is over-literal instruction following. When users implicitly expect the model to fill gaps or challenge assumptions, the model instead treats the prompt as a closed specification. The result looks obedient but misses the point. This is not hallucination. It is misaligned helpfulness.

The authors also test paraphrasing. When the same intent is expressed with different phrasing, model behavior shifts significantly. That tells us the model is anchoring on linguistic form, not reconstructing an underlying goal.

"Humans normalize phrasing differences. Models react to them."

What’s striking is that longer context often worsens intent alignment. Adding more background increases the chance the model optimizes for local relevance instead of global purpose. More tokens give the illusion of understanding while diluting the signal of what the user actually wants.

The paper argues this is not solvable by bigger context windows or better prompting alone. Intent is not explicitly stated most of the time. It has to be inferred, tracked, and sometimes revised mid-conversation.

That requires models to reason about users, not just text.

The implication is brutal for agents and copilots. If a system cannot reliably infer intent, autonomy becomes dangerous. Tool use amplifies mistakes.

Confident execution based on a misunderstood goal is worse than asking a clarifying question.

The authors suggest future work should treat intent as a first-class object: something to model, update, and verify explicitly. Not just “what was said,” but “what outcome is being optimized.” Until then, many AI systems will continue to feel smart, fast, and subtly wrong.

This paper explains why that feeling keeps coming up.

Paper: Beyond Context: Large Language Models Failure to Grasp Users Intent

[Lomez @L0m3z]

@RichardHanania You should seriously reckon with the fact that your inability to understand fundamental moral intuitions felt by 99% of humabity disqualifies you from making recommendations for how those people organize their societies

[Ezra @Revan]

@DJSnM That would be meaningful and helpful though

[Ezra @Revan]

@BowTiedNiners @TheBigCigar1906 Attempting that turn on a B52 with ~800 ft agl...

[Ezra @Revan]

@WizardGoesBoom I don't know if I can really articulate this but there was something about the vibe of magic and spells in the AD&D books that just isn't there anymore. Maybe it was the inaccessiblity of the system itself enhancing the feel of "magic".

[Ezra @Revan]

@jxnlco Honestly don't try to help these animals.

[Ezra @Revan]

@RenoMayGuns <Joke about dropping them>