lukev

25 ms·

211.
by lukev 1y ago
You can absolutely extrapolate the results, because what this shows is that even when "reasoning" these models are still fundamentally repeating in-sample patterns, and that they collapse when faced with novel reasoning tasks abov
212.
by lukev 1y ago
I agree, obviously, but half the internet is still running around claiming we're on the verge of a singularity, so demonstrating the actual limitations of these systems concretely is important.
213.
by lukev 1y ago
Whenever I think about AI and labor, I can't help thinking about David Graeber's [Bullshit Jobs]( https://en.wikipedia.org/wiki/Bullshit_Jobs ). And there's multiple confounding factors at play. Yes, lots
214.
by lukev 1y ago
Capitalism isn't really just about free markets, though. Lots of economic systems have markets. Capitalism is more about ownership of the means of production and the relationship of labor to capital.
215.
by lukev 1y ago
To add, not just Jesus, but the entire Bible. The Year of Jubilee alone, from the Pentateuch, basically eliminates capitalism as we know it (if enacted at a full societal scale.)
216.
by lukev 1y ago
I mean, this is a little disingenuous. If you posit that there exists a supernatural mechanism for providing food and health, why would you admit that it exists for a small subset of the population, but not a larger (or whole) population?
217.
by lukev 1y ago
Interesting. I can make assumptions, but how would you label the novice and jedi roles in this particular instantiation of the meme?
218.
by lukev 1y ago
I saw a quote today that's been living rent-free in my head ever since: "AI utilitarians are in a suicide pact with the U.S. economy" (attribution: @bencollins.bsky.social‬)
219.
by lukev 1y ago
Can we get a disclosure on your experience and who you work for, if you’re going to weigh in so heavily on important public vs private sector issues like this?
220.
by lukev 1y ago
One of my favorite things about the JVM ecosystem is how stable it is. A 5-year-old library will almost certainly Just Work. And Clojure very much follows the same spirit. There's a lot of great, useful libraries that haven't been
221.
by lukev 1y ago
So, I know the battle has been lost for the definition of vibe coding ( https://simonwillison.net/2025/Mar/19/vibe-coding/ ). But we definitely need a word that makes a distinction between a LLM autonomous
222.
by lukev 1y ago
I think the concept that AI can be used to sell itself -- that a product is more valuable simply because it incorporates AI -- has to end, and soon. If you can actually use it to build a better product on it's own terms, great. But as
223.
by lukev 1y ago
You're right: it's probably fine then. My bad.
224.
by lukev 1y ago
Half the comments here are just pointing out that ad blockers exist, which is missing the point. The damage of an advertising-based internet economy is not limited to just "seeing ads." The entire content and structure of the inte
225.
by lukev 1y ago
I hate this app and I hate its code. I think the fact that this can happen at all is important. And whoever can contextualize this capability in the context of "real" software development will be the victor.
226.
by lukev 1y ago
That's a recommendation on how to conduct personal/unclassified communications. Approval for classified or military use is a completely different ballgame.
227.
by lukev 1y ago
There are whole agencies dedicated to this. In this case, DISA ( https://www.disa.mil/About/Our-Work )
228.
by lukev 1y ago
If this happened the way it's being reported, yeah, several people should lose their jobs. I suspect this is a case of being more afraid of saying "no" to the boss than of facing consequences for violating policy. Policies ar
229.
by lukev 1y ago
The failure mode is that the Secretary of Defense unilaterally bypassed security protocols to use technology that had not been evaluated for that use case in a national security context by the appropriate experts. It doesn't matter if
230.
by lukev 1y ago
So it's fair to say that effectiveness at monetization is an extremely strong evolutionary pressure on how technology evolves?
231.
by lukev 1y ago
So you think consumer engagement ultimately drives what types of tech that companies invest in building? I can see that argument. Why do companies want consumer engagement to start with?
232.
by lukev 1y ago
Let's follow this train of thought. What are the selective pressures on the "natural evolution of communications technology?"
233.
by lukev 1y ago
I don't follow. Are you saying the DOGE boys are trying to give Trump bad press?
234.
by lukev 1y ago
The Vision Pro could easily have been successful if they'd invested an additional 10% of it's R&D budget into software development, and released a suite of tools that actually leveraged what the platform was capable of. It&#x2
235.
by lukev 1y ago
Agreed. And it's bad enough within apps... it's even worse the way they use push notifications for this kind of thing. I installed your app so I'd know when my prescriptions are ready, or when my ride is here. Abusing that pr
236.
by lukev 1y ago
I mean there are definitely systems that let you hook up to a locally executed LLM, or an arbitrary cloud hosted one that you might either host yourself or pay for separately. So yeah I'd say those are "completely free."
237.
by lukev 1y ago
That's not what the article is talking about, at all. Try not to respond just to headlines.
238.
by lukev 1y ago
I'm not saying it'll cost the same, I'm saying it'll still be accessible. (Also, comfort level on a train is typically much better.) And it'll properly price in externalities, which is not currently the case. Also,
239.
by lukev 1y ago
I chose those examples specifically because they were completely successful.
240.
by lukev 1y ago
Ah yes. George Washington, Simon Bolivar, Vladmir Lenin, etc... famously incompetent.

More ›