
@orit@mastodon.social // she/they // avg systems enjoyer
the sharp end
Joined March 2010
- Tweets121K
- Following954
- Followers251
- Likes117K
orit retweeted
not an artist 🤍 but this is true from lived experience being in tech as <1% demographic. most everyone else outside of tech i can vibe with fairly easily. that's the irony/dilemma. not in tech because of the people - but because i enjoy the mental stimulation/challenge.
everyone is simply going to be responsible for more lol
orit retweeted
Support & belonging are empirically-backed elements of good problem-solving in complex knowledge work AND engineering consistently fails to create this culture in all the research on belonging that I'm aware of
orit retweeted
Something that I’ve learned over my time in industry is that the web runs on cache and that cache is always redis and no matter what you’re building, from CRUD to AGI, you better know about cache. openai.com/blog/march-20-cha…
orit retweeted
We’re thrilled to release this into the world ✨
Introducing MOONLIGHT—a space for social tarot.
Explore beautiful decks. Do tarot with friends. Start an interactive ritual. (Even if you're new & curious)
moonlight.world
Excited to see what you do with this! 👀👀
orit retweeted
I wrote “I am alive” on a piece of paper, and placed it into a photocopier. What I saw next has shocking implications
i think chatgpt is in this uncomfortable spot where it’s impossible to use as a complete layer of abstraction over code but also kind of does the work in a way that requires a ton of babying
i guess what i’m trying to get at is chatgpt is the peak leaky magical abstraction which we’re all enamored with at first due to its ability to perform scoped and simple tasks and then eventually hate when we have to actually build anything reliable on top of it
orit retweeted
direct the energy and will to act created by incidents into more useful avenues of action, eg learning from incidents and enhancing capabilities that are useful in a wide variety of scenarios (e.g., load shedding), i.e. actually improve your ability to adapt to the next incident.
orit retweeted
There's something called "The Tyranny of the Native Speaker". Among other things, it describes how native speakers are not just considered an authority on their mother tongue, but almost always also considered more trustworthy authorities on whatever they are communicating about in that language.
Part of this is linguistic processing fluency. The easier something is to understand and the more familiar it is, the more trustworthy we feel it is. Another part of it is just good old-fashioned colonialism.
It's why I take such pains to get assistance with my English. You risk being less credible than a less knowledgeable native speaker if you don't and since I compete for traffic almost entirely with native speakers writing about the same things I do, it would be an insurmountable disadvantage if I could not nearly match their English.
This particular kind of high-level fluency offers a tremendous lifetime advantage to native speakers who come from backgrounds that min-max for signaling. Speaking and writing in ways that elicit trust and indicate leadership, and knowing appropriate shibboleths and references is a huge advantage that often trumps actually knowing anything about the subject at hand. Sounding "truthy" is a tremendous skill. This skill, trained from birth is nearly impossible for non-natives to catch up with.
Even in the language source country, perceived fluency is rooted in, although not exclusive to, race and class. AI can go beyond "really good translation" to potentially outright spoofing key social signals- sounding "truthy", competent, and dignified, like someone from the "right" schools and the "right" sort of background. Allowing for near-equivalent output to potentially be produced by non-native writers and other out-groups, possibly even spoken via AR prompting.
This does not neutralize other, visible biases, but well-prompted AI could certainly make an interesting dent in at least one- and pose a threat to those whose entire career is built on exploiting those signals.
See:
I Don't Like You Because You're Hard to Understand: The Role of Processing Fluency in the Language Attitudes Process
sci-hub.ru/10.1111/hcre.1207…
The name-pronunciation effect: Why people like Mr. Smith more than Mr. Colquhoun
sciencedirect.com/science/ar…
The Interpersonal Consequences of
Processing Ease: Fluency as a
Metacognitive Foundation for Prejudice
sci-hub.ru/10.1177/096372141…
orit retweeted
1. Swathes of the economy now seem reliant on a small number of cloud computing providers.
@FTC is seeking public input on how the current market structure and business practices of cloud providers affect competition and data security.
ftc.gov/news-events/news/pre…













