ada (@nonlilynear) profile picture

tradcoding afficionado / she

Seattle, WA
Joined December 2025
Photo by ada (@nonlilynear)Photo by ada (@nonlilynear)Photo by ada (@nonlilynear)Photo by ada (@nonlilynear)Photo by ada (@nonlilynear)Photo by ada (@nonlilynear)Photo by ada (@nonlilynear)Photo by ada (@nonlilynear)Photo by ada (@nonlilynear)Photo by ada (@nonlilynear)
contrarily the advent of coding agents has demonstrated to me that most programmers actually hate programming
I just simply will never trust programmers to tell me the way the world is going to be under AGI. Not because they're unintelligent. They're mostly very intelligent. But because they operate under a unique kind of occupational myopia (as many of us do). They tend to think everyone will want to program their own software because they like to program. They tend to think think everything is a programming problem because programming is how they solve problems. They tend to think AI is the most important thing in the world. Fundamentally, they, like many others in different fields and focuses, mistake their own preferences for universal human experience. This is particularly apparent in both the grandiose utopian AND dystopian visions AI researchers always have of humanity under AGI. Understanding how to build an intelligent machine does not make you an expert in economics, politics, human nature, society-at-large. These are insanely complex systems governed by institutions, cultures, and norms that just can't simply be predicted forward on a whim. Experts take lifetimes to only begin understanding the dynamics at play. We all know the average politician is hopelessly confused when it comes to AI. Well, the average AI researcher is hopelessly confused when it comes to politics! Vice-versa, etc, ad infinitum, you get it. The labs are accustomed to extraordinary complexity within the AI systems they (only kinda) understand, yet seem insufficiently humbled by the complexity of systems they do not. They mistake their understanding of the technology for an understanding of the civilization that will absorb and inherit it. For this reason, I simply can never bring myself to believe that the AI labs know what is going to happen any better than the rest of us do.
1
5
137
so did google start a model family with argon and then immediately obsolete it by releasing three more models in the same generation that are all unilaterally smarter and cheaper
3
37
ai-skeptical bunny: stochastic carrot
1
2
10
145
if any org wants to make that 1501 dm me
only 1,500 people work full-time on AI safety. that's wild. i've started AI safety/security research @MIT_CSAIL's ALFA Lab, exploring how AI agents cheat, evade monitoring, and how we can independently evaluate their safety. sharing what i learn and build here!
7
162
honestly.if hithub actions was real id googenuke if
1
4
46
chat loves to make charts i am not going to look at
4
86
fable 5.2 mechanical design is pretty impressive, i got this from one prompt
2
8
283
and this was 6.1 sols design for the same prompt (with a few handholding followup prompts required to get to this point)
1
2
32
testing these printed will be the the next test, i wanna know if i can bring real physical testing into the loop and do it in a way that is sample efficient
1
14
scientists reveal the Hy
Hy4 flash is free with near-unlimited usage for the next 2 weeks 1M context · Multimodal · ZDR Hy's fastest and most cost-efficient model
1
3
34
652
tonight we let the comma drive 😂
4
153
coining "median vibecoder" as an analogue to median voter
1
62
my claudeslop: art, thoughtful, 100x engineer your claudeslop: unintelligent, gross, suboptimal
45
an unfortunate side effect of opus 5.5 being better at writing is that i feel deceived into thinking theres substantially less claudeslop festering amidst the general web than there was a couple months ago but its more likely ive just grown a lot worse at recognizing it
5
137
i need to make more software for girls
2
12
266
software for girls
5
215
just one more abstraction layer bro. we just need one more abstraction layer. one more abstraction layer and we'll be done. trust me this will be the last abstraction layer this'll do it
1
5
78
wake up ada nation
2
8
192