I do; it looks like the issue was that claude does not read AGENTS.md by default

T-Mobile is ramping up starlink coverage

I switched this recently so I could dual-wield, it seems plausible claude has been missing it. this case is certainly evidence

o rly.

taps sign
bsky.app/profile/tbab...

and it stacks them too, so you have to answer _all_ the questions _in the order asked_. you can't redirect or address the entire plan in prose. it's gotta be split up the way the agent diced it

do you know what the option is called?

4.7 seems much more prone to using them. they are an extremely unwelcome disruption to dialogue

like, imagine sending another person a multiple choice poll in a 1:1 chat. unhinged

how do I stop claude code from using those infuriating multiple choice options. prompting in AGENTS.md didn't work

where's that tweet that goes "my T-shirt saying that I always defect in prisoner's dilemma is gathering fewer cooperators than I'd hoped"

IMV choosing to shore up your position rather than offer an olive branch is an entirely justified strategy, given both the content of the video and how he ended up responding

in that sense I think it _can_ inform what groups of people think. e.g., there are serious arguments against the proliferation of AI, but the anti-AI egregore does not use these and instead picks weak, easily-refuted positions. IMV this truthfully shows it isn't aware of/concerned about those issues

but also the absence of coherent argument does reveal 'state of mind' information, i.e. that the speaker _believes_ they have no better option, regardless of what arguments might be out there; this may be true at their own level of awareness

of course, otherwise nutpicking wouldn't be a thing. see also dynamics where the loudest, most insane people go at each other, while reasonable people go about their business, distorting the representation of viewpoints

to the extent that I follow your points, I think this would be my way of phrasing what you're saying?

The Brain’s Learning Algorithm Isn’t Backpropagation YouTube video by Artem Kirsanov

(endorsement from neurology on the idea of 'brains are prediction machines with error correction: www.youtube.com/watch?v=l-OL...; sidenote I am bullish on the technique in this video— predictive coding— eventually supplanting backprop + gradient descent)

so it is with phenomenological experiences, we need to include enough of the extent of the brain in space and in time to see them play out (which is really including the underlying causal structure, at the deeper level of ontology)

cf. 'wetness' as a causal pattern with structure (adhesion, cohesion, viscosity, thermal transport) that does not appear until you include a large number of molecules in your model. a water molecule is not 'wet'; you have to include many of them in your view before the above properties appear

and as for the 'extension' of awareness over time, like many phenomena in causality, the pattern does not appear until you widen your window enough to enclose all its parts. and predictions/anticipations of the world play out over finite time

and so 'what is in the spotlight' could probably be thought of as 'what the mind is currently occupied in predicting'

which puts the activity of 'computing stuff' at the center of awareness. feels right and satisfying to me

and second, on the nature of the 'spotlight of attention', I think a good chunk of our experience of 'awareness' can be explained by a mind which is silently predicting things and continuously sensing the differences between its anticipations and what occurs

and the above is why I am what you might call a "computational realist"

because really, 'substrates' like neurons and atoms are themselves patterns in the influence of information on other information

we must first have a notion of 'states transforming by rules' before we talk about which states and which rules (i.e., the Standard Model, neurology, whatever)

two thoughts that come to mind:

endorse that the most elementary ontology on which to build others is "causal relationships exist"; that is, there exist patterns in how information affects other information, it's these patterns/structure which are real; more than the substrates that implement them

heuristic: people will try to defend their positions with the best argument available to them. if they put a sh*t argument at the forefront, it implies they have no better arguments available

heuristic: people will try to defend their positions with the best argument available to them. if they put a sh*t argument at the forefront, it implies they have no better arguments available

a useful thing to notice is when someone's argument doesn't depend at all on the substance of the objections to it; something you could say to literally any criticism at all. in other words, a complete absence of response to the original objection

same shape as "nuh uh!"

(I dunno if this is true for others, but my mental image of Mythos was a 'professorly lion' evolution of thebesian claudesona basically right away. banana has not done a particularly good job here, but you get the idea)

masley might have taken a look at the disingenuousness in the original video and (perhaps rightly) calculated that his opponent would not respond with intellectual good faith

we'll never know the counterfactual for sure, but benn's behavior doesn't speak well for him

The left ceded talk radio, then ceded social media strategy, then ceded algorithmic amplification, each time on the theory that the medium was beneath them or inherently corrupting. We cannot afford to keep doing this with AI because its bigger than all of them.