quavious le fou

[quavious le fou @realeigenvalues]

@doodlestein @mattshumer_ does 150k LoC to check if a bash/git command should be blocked seem 'solved'?

[quavious le fou @realeigenvalues]

classic sniff walk experience

[quavious le fou @realeigenvalues]

its interesting (but i guess somewhat expected) how much the stylistic choices seem to change for an LLM based on the project domain. specifically anything financial, the var names or code quality in general read like 2015-esque python

[quavious le fou @realeigenvalues]

@willccbb @dylan522p he also 'leaks' MNPI as part of his newsletter and would be sanctioned if there was any sort of enforcement

[quavious le fou @realeigenvalues]

@xmjEE hope she doesn’t become a barker

[quavious le fou @realeigenvalues]

adopted a beautiful midwit, just like me, a week ago

[quavious le fou @realeigenvalues]

@0xfdf @post555s they tried this already 1-2 years ago and seemed like it didnt help much

[quavious le fou @realeigenvalues]

@max_spero_ problem is he doesn’t seem well suited to build the next thing. he’s much better at acquiring something which at the time seems like whatever but pans out quite tremendously in the long run. he’s probably aware of that, but does he not see that as feasible for this next stage?

[quavious le fou @realeigenvalues]

dont think ive ever run into a "talent sourcer" who uses the correct acronym for daylight/standard time

[quavious le fou @realeigenvalues]

is fast mode on the mac gui app just completely fake? seems like it has basically no effect

[quavious le fou @realeigenvalues]

@jacobmparis @jxnlco has never worked for me. computer use too, think it worked at one point but now completely broken with warnings everytime i update mac codex app trying to have me trash 'computer use' (despite approving it in Privacy & Security system settings)

[quavious le fou @realeigenvalues]

@jxnlco @jacobmparis is this expected? if so, it’s been working great!

[quavious le fou @realeigenvalues]

gotta love aligning about job postings if you want to make it in modern society

[quavious le fou @realeigenvalues]

if you ever presented a classifier this bad as a Data Scientist/ML engineer, you would be fired from most jobs

[quavious le fou @realeigenvalues]

@krishnanrohit that’s the “obvious” one but analogy falls short if you think about it for all of 10 seconds

[quavious le fou @realeigenvalues]

@VDN_00001 yeah, think I largely agree w/ what you are saying although you seem more skeptical on it being a 'real' pathology than me. its such a relatively novel phenomenon that there is no 'real' research on it yet

app just seemed like 𝘴𝘰𝘮𝘦𝘵𝘩𝘪𝘯𝘨 to start with (and flag planting)

[quavious le fou @realeigenvalues]

Reposting that I made an anti-AI-psychosis iOS app as the first post did not get much traction.

The issue of AI over-reliance isnt going to fade away; the number of viral posts by others related to AI skill deterioration or AI psychosis seems to be increasing.

[quavious le fou @realeigenvalues]

@ryxcommar @buccocapital @MikeBurgersburg do you see any way to incentivize yourself/others to not fall into this local optimum and become AI-overreliant?

from what ive seen (and saw in myself), the average knowledge worker is totally fine with and largely unaware of possible downsides of just being a conduit to an LLM

[quavious le fou @realeigenvalues]

@ryxcommar @buccocapital @MikeBurgersburg totally. had a somewhat similar realization awhile back although more framed around how junior engineers not getting to do that rote 80% means the intuition for the harder stuff isnt there. but then what happens?

https://t.co/BODB9WHd8B

[quavious le fou @realeigenvalues]

feel like we are going to have some fucked up feedback loop that plays out in 5-15 years from automating so many jobs that maybe aren’t the most vital or creative but allowed people to gain experience/exposure and go on to start interesting co’s and have actual novel ideas

[quavious le fou @realeigenvalues]

@VDN_00001 its a good question. in general, i dont think people want to feel as if they are cognitively declining (specifically because they want to take the 'easy way' to do things), but is an app the best way to go about mitigating that?