fritzo

8 ms·

91.
by fritzo 1y ago
I've used google's old diff-match-patch, a faster python binding of that C++ library fast-diff-match-patch, and biopython (which amazingly supports unicode!)
92.
by fritzo 1y ago
In my experience, the 2nd most costly function in agents (after LLM calls) is diffing/patching/merging asynchronous edits to resolve conflicts. Those conflict resolution operations can call out to low-level libraries, but they are
93.
by fritzo 1y ago
Honest question: are you a Malthusian? If so, what do you think is the ideal birth rate?
94.
by fritzo 1y ago
Is this just the more general problem of steadily decreasing cost to create & distribute content leading to a longer tail of low-quality content, requiring stronger filtering mechanisms? I don't see what's specifically AI-rela
95.
by fritzo 1y ago
Do you think katabasis' proposal is unhealthy for a society? It strikes me as healthy for parents to keep their kids at ideological home during an information pandemic, to provide access to educational material and shelter from social
96.
by fritzo 1y ago
It's exciting seeing grassroots censorship efforts like this.
97.
by fritzo 1y ago
A real stretcher, you might say
98.
by fritzo 1y ago
From scratch it's also easy dress up a quoted Y combinator as in Löb's theorem. Start with the usual Y combinator Y = \f. (\x.f(x x)) (\x.f(x x)) : (p -> p) -> p which can produce a thing of type p from a function
99.
by fritzo 2y ago
ARC 2 is brand new, but neurosymbolic approaches have performed well on the original ARC, e.g. https://arxiv.org/abs/2411.02272
100.
by fritzo 2y ago
Crawling is important, and PoW incentivizes the creation of a data broker as a cached middle layer to reduce negative effects on the website by amortizing PoW cost among crawlers.
101.
by fritzo 2y ago
This essay would be stronger if it backed up its anecdotes with data, e.g. trends of standardized test score over time.
102.
by fritzo 2y ago
Where are those ten houses going to get water?
103.
by fritzo 2y ago
Point taken, though I'd like to charitably think Americans are 'lazy' the way Haskell is lazy or Japanese Kanban or Jidoka is lazy. As MisterTea points out, the VCs invest in metrics, management pushes on metrics, and produc
104.
by fritzo 2y ago
Paddles for stand up paddle boards, kayaks
105.
by fritzo 2y ago
Specifically within a single file, I find LLMs can easily extend and maintain large sets of slightly varying concrete code, yet they struggle when that code is compressed into abstractions. The concrete examples form a dataset. Attention
106.
by fritzo 2y ago
Rule of three is obsolete in age of AI assist. New rule is rule of 10ish.
107.
by fritzo 2y ago
What a relief, what a sanctuary for sore eyes!
108.
by fritzo 2y ago
Right, like propane tanks for my bbq
109.
by fritzo 2y ago
i feel like you're overgeneralizing here
110.
by fritzo 2y ago
unpublished, i'm sure the idea has come up many times
111.
by fritzo 2y ago
I first saw the trick in 2009 when Linda Brown Westrick independently discovered it as a way to parallelize sequential sampling of MCMC chains.
112.
by fritzo 2y ago
A great source for stereo pairs is NOAA's aerial imagery data, consisting of various snapshots along an airplane's trajectory. For example here is a stereo pair of Desecheo Island: https://cdn.coastalscience.noaa.gov&#x
113.
by fritzo 2y ago
Math research is cheap, requiring only coffee + room + board + health insurance
114.
by fritzo 2y ago
Was expecting a map of bagel and donut shops
115.
by fritzo 2y ago
> Follow us on Instagram When Elon bought twitter, I didn't immediately see the conflict of interest. Now it's clear.
116.
by fritzo 2y ago
I like this writing style with unbound variables. Reminds me of Maya Binyam's novel "Hangman", or Kafka's novels.
117.
by fritzo 2y ago
Agreed. I learned much from Joe Studwell's book "How Asia Works" (2013), where he argues that among countries in the Asian region, those whose economies flourished were those with protectionist trade policies and export disci
118.
by fritzo 2y ago
It feels like mere language difference. Certainly every government official is advised by many Excel spreadsheets. Were those spreadsheets "appointed", no.
119.
by fritzo 2y ago
Sounds like a contender for the longest-ever lasting digital backup.
120.
by fritzo 2y ago
Labor has never been more expensive than now. Raw resources have never been under more competition than now. We as a species are more numerous and consume more scissors than ever before. Supply chains have never farther separated end user f

More ›