DroneBetter

I guess the answer depends on whether you see NASA's goal as "to do cost-effective science" or "to do cost-ineffective rocketry" (since the express purpose of e.g. the SLS is to appeal to congress by providing local jobs as economy stimulus)

a few companies (mainly those in the open-source space) run their own Mastodon instances, though this is more common for FOSS organisations

well the Canadian and Japanese militaries are not engaged in imperialistic wars of terror currently

you should include a faster version of the fibonacci function with exponentiation by squaring

  def fibonacci(k):
    a,b=(0,1)
    for i in range(k.bit_length()-1,-1,-1):
      d=a**2
      c=2*a*b-d
      d+=b**2
      (a,b)=(d,c+d) if k>>i&1 else (c,d)
    return a

see https://oeis.org/wiki/User:Natalia_L._Skirrow/linear_recurre... (warning: old and bad and in need of revision), https://github.com/sympy/sympy/pull/30452 and https://github.com/sympy/sympy/pull/30541 for details of how to make similarly fast programs for arbitrary linear-recurrent sequences.

you can also encode polynomials into integers; see https://mathstodon.xyz/@peterluschny/116320199782572958 and the following prog from https://codegolf.stackexchange.com/a/279771

  lambda n:pow(p:=2<<n,n,p*p+~p)//p

both of these would be more intensive on the arithmetic side rather than control flow

also you could at least wrap the existing one in a `functools.cache`

if only there were a tag for llmery so that undisclosed use of it couud be a punishable offence

> There was also the Japanese JIS standard keyboard

is a particularly humorous instance of https://en.wikipedia.org/wiki/RAS_syndrome

I think orliesaurus meant the PR team is unable to rest because the rest of Deepmind is so prolific and keep coming out with new things for them to announce

for simulating gravity in particular you can do one better, since a sphere is in turn gravitationally indistinguishable from a point mass due to the https://en.wikipedia.org/wiki/Shell_theorem#Force_on_a_point...

it is difficult to motivate hyperconsumers with the narrow band of improvements that can be made with new LLM models and also are specific to smartphone OS integration and not superseded years in advance by datacentres

the CEO's pay in particular seems to have absorbed a large portion of the savings created by layoffs.

cf. Mozilla Firefox Is Dying (2025) https://news.ycombinator.com/item?id=44939271 and a list of posts from 2024 https://news.ycombinator.com/item?id=39303736

it would be useful to create an account and document your work in a post at https://conwaylife.com/forums, if you're not planning to make a blog post.

you mean... as opposed to removing the bricks with the art on them to auction off and replacing them with new bricks?

well, a bug in the Lean kernel was discovered last week by way of an LLM tricking itself and its handler into believing it had found a non-constructive proof of the existence of a nontrivial Collatz cycle, see https://infosec.exchange/@0xabad1dea/117002106099986943 and https://lipn.info/@mevenlennonbertrand/116997917683191056

the unit distance problem's paper was human-summarised and condensed significantly from the initial LLM output, so it seems the model did do some theory-building there (possibly providing motivation for the clever trick as a natural deduction, at least in its eyes) before humans cut off all the chaff; I imagine the development was similar here.

maybe by the platforms based on the non-publicly-disclosed behaviour (including usage patterns, like rates of opening specific menus), but not by other unprivileged users seeking to publicly out pseudomymous authors.

the bird's children may be crippled if it gets tangled in their feet while they're still developing, see https://en.wikipedia.org/wiki/Stringfoot (which is in general a common sight wherever pigeons can be found); when my flat's fibreoptic was being installed, the cable-layers made a big deal of collecting the cuttings where they spliced the cable and ensuring nothing would be in reach of any animals.

well it seems advantageous to be able to say "your honour, we're being unfairly singled out here; at the time of the incident this lawsuit concerns, none of our competitors had resolved this problem either" and have receipts.

(and also know whether the safeguards they're trying to implement/improve towards have been achieved or are theoretical)

also just as importantly (or moreso for cases of individual rooms used for social/hobbyist clubs like this) would be 'empty building taxes' to prevent the eventuality of landlords preferring to rent out fewer properties over reducing the cost of vacant ones because * housing is necessary for survival so people cannot afford to baulk at the prices of artificial scarcity, making the optimal price effectively arbitrarily high without regulation * landlords are generally not infinitesimal entities in a vacuum, and generally stand to lose existing customers who are still on the expensive contracts if they cheapen their unoccupied offerings although such a solution would just move the problem one layer further up (stop building properties to have them be vacant) and the solution is land tax

I think it becomes somewhat more defensible when considering that the alternative was operatiny Google's policy (before the advent of competition) of "these models would bring unknown dangers in the hands of the public, we shouldn't release them until we better understand the implications" (or perhaps more selfishly "these effectively nullify all our detectors of generated text, if released they would instantly lose us the war on SEO").

(recall that OpenAI thought GPT-2 was too powerful to release for approximately tantamount reasons)

i hate every time i hear about spaghettisort.

it is still O(n) weight to transport, so O(n^2) amortised; if you liken having a stronger hand that can carry more spaghetti to parallelisation, it's beaten by O(log^2 n) sorting algorithms on parallelised classical computers.

this is explained by the symbolic method (https://en.wikipedia.org/wiki/Symbolic_method); in the language of exppnential generating functions, we say that bell(n) = A000110(n) = [x^n/n!] exp(exp(x)-1), since exp() means "sets of ()," exp()-1 means "nonempty sets of ()" and x means "singletons" (it can be read as "bell(n) is the # of partitions of [n] into sets of nonempty sets").

see the asymptotic formulae by Vaclav Kotesovec and Natalia L. Skirrow (me!) on https://oeis.org/A000110; the former (together with Stirling's approximation) says bell(n)/n! ~ exp(n/W(n) - nlog(W(n)) - 1) / √(τ(1+W(n))), and the latter says that the average number of subsets in a partition on [n] counted by the Bell number is ~ n/log(n), while for a permutation it's only = H_n ~ log(n).

(using W(n) ~ log(n), although it's more precisely log(n) - log(log(n)) + o(1); its appearance here comes from solving for the radius from the origin at which |exp(exp(z)-1) / z^(n+1)| has a saddle point in the complex plane, which lets you pretend your contour integral from Cauchy's formula is a Gaussian and approximate it in terms of the generating function's logarithmic second derivative)

another cool adjacent result that comes out of saddlepointery on e.g.f.s is that (# of sets of lists on [n], A000262(n))/(# of sets of cycles on [n], n!) ~ exp(2√n - 1/2)/(2√π * n^(3/4)); this is also the expected number of preimages under the surjection A000262(n) -> n! from closing lists into cycles, or equivalently the expected product of cycle lengths in a random permutation ∈ Sₙ, and is my favourite example of a function whose growth is between polynomial and exponential

this is not necessarily the case; the coursework could have been produced by a different person from the teacher (although generally at my alma mater the 'module organiser' fulfils both roles).

it would be nice to show an amplified pixelwise difference between the before and after images

I don't think that's true, at least in the 13" model.

and the F keys come back on the touch bar if you hold the fn button

this is basically the plot of the end of Dark Star (1975)

> done already in the 1990s by human-written programs that iterated through the finite casework that human thought had reduced the theorems to (four-colour theorem, FLT, etc.), which recent developments (eg. LLMs autonomously resolving Erdős problems) seem meaningfully distinct from. > human effort to make the results cleanly understandable well, perhaps loops of "derive proof through reasoning in English, formalise in Lean, use AST size of formal proof as a metric to optimise (via an LLM-guided search), translate back into English" could improve this? a lot of resources are being spent to make frontier LLMs more resistant to hallucinations via Lean, perhaps cogency will increase as a byproduct.

from the "DeepSeek is a ploy to undermine usamerican models' duopoly" theory's perspective, "now everyone has this" helps them achieve this goal more efficiently.

especially if it's something that the major companies had already stumbled upon (something equivalent to) and regarded as a trade secret.

yes, that is pretty much what he disclosed in the article

> He turned to Google’s Gemini AI for advice and decided to create a “hot girl” crafted specifically for the “MAGA/conservative niche,” after the software told him that “the conservative audience (especially older men in the US) often has higher disposable income and is more loyal,” according to Wired.

age doesn't inherently make math less useful, and the parts it does affect it does non-uniformly.

i have undergone an undergrad differential equations module that taught exclusively ad-hoc methods for certain families of equation that no working mathematician needs to know since they were all subsumed into and superseded by computer algebra systems, but the subject i would enjoy replacing it with (generating functions) is similarly old in origin (perhaps even earlier, since Euler used most of the techniques that an undergrad class would cover before diffeqs were considered an object of study) but has happened to become more useful with the advent of CASes instead of less.