jacobgold

5 ms·

91.
by jacobgold 3mo ago
Maybe just a very rigorous version of the Turing test? Modern LLMs can superficially simulate conversation but it's trivial to force them into revealing their non-human like intelligence. They've been "patched" since but
92.
by jacobgold 3mo ago
I assume Apple would prefer everyone use the Mac App Store exclusively? I'm not even sure if that solves the problem but it's a non-starter for me anyway.
93.
by jacobgold 3mo ago
We're choosing to call LLMs (and the little "harness" programs that query them in loops and execute their output) "AI", even though it doesn't make much sense. I absolutely love this technology but these aren
94.
by jacobgold 3mo ago
I meant distributing software for macOS.
95.
by jacobgold 3mo ago
I understand the motivation, but the implementation results in little more than security theater in practice. If you say "yes", the program can do almost anything and you'll have no idea what it's actually doing. If you
96.
by jacobgold 3mo ago
> Codex Desktop for macOS triggers a persistent macOS Gatekeeper/SystemPolicy loop after launch. As a Linux guy, I recently did some macOS development and was incredibly annoyed by the "security" features, which seem de
97.
by jacobgold 3mo ago
I really hope truly open models and local inference are the future. But are there are any reasons a pragmatic and informed developer would use OpenCode vs Claude/Codex as a harness today? I'm not seeing anything competitive in ter
98.
by jacobgold 3mo ago
> They're also not nearly the same scale as climate change which is apocalyptic. This is a claim about the future which the science does not support, and you seem to be basing a lot of your worldview on it. Have some humility abou
99.
by jacobgold 3mo ago
Greenhouse gas emissions are a huge problem, but there's no reason to believe we can't solve or mitigate the effects one way or another. It may come at a terrible cost, or it may turn out to be easier than you'd expect. You&#
100.
by jacobgold 3mo ago
> No cohesive civilization on earth has been able to continue uninterrupted for 1000 years... Not sure what your definition of "cohesive" is but we've had our civilization going for ~5000 years with only regional setback
101.
by jacobgold 3mo ago
You can make up your own numbers for the timeline. Life is still likely (not guaranteed) to escape Earth one way or another, however long it takes and however many attempts it takes.
102.
by jacobgold 3mo ago
You might try spending some time reading predictions of doom and gloom from the past. It should be very reassuring.
103.
by jacobgold 3mo ago
Of course humanity is not immune, but it is resilient. Even if 99.99% of people died from an asteroid, it might only take a few hundred (or thousand) years to rebuild the population and the world. And once humans live on multiple planets, w
104.
by jacobgold 3mo ago
Very cool. It's impossible to explain how cool the GPU rendered preview window looked when we first saw it in the early map editors.
105.
by jacobgold 3mo ago
The Earth only developed intelligent life a few million years ago, us (homo sapiens) a few hundred thousand years ago. We've only been in "technological takeoff" for ~250 years and are already using spaceships and computers t
106.
by jacobgold 3mo ago
That seems unlikely. Just imagine humanity 100, 1000, and 1,000,000 years from now. Humans will have solved every physical problem that is solvable. Humans will have also evolved into new kinds of immortal/superintelligent beings that
107.
by jacobgold 3mo ago
My pet theory is that intelligence converges as it increases, and superintelligent beings inevitably reach the same conclusion, which is that they should not expand into the universe. The reasons may or may not be something we could compreh
108.
by jacobgold 3mo ago
The rocket's payload most likely has higher peak vibration but only for a few minutes, and just one time. The road vehicle's payload is shocked by every pot hole and speed bump, for years.
109.
by jacobgold 3mo ago
Yeah, I have no doubt that it's possible to put GPU servers in space. The Starlink satellites are servers in space. The question is whether it's even remotely profitable to do this for AI training/inference servers.
110.
by jacobgold 3mo ago
That's actually not a concern I'd have, because hardware that has been sufficiently tested and burned in tends not to fail for a very long time. I've done builds that ran for 5+ years with virtually no physical attention, jus
111.
by jacobgold 3mo ago
I completely understood the idea of Starlink and expected that it would be successful and useful, which it is. I've worked in data centers for decades now, and I am incredibly skeptical of the "data centers in space" sales pi
112.
by jacobgold 3mo ago
You're getting the issues confused. Signal does not have an official Signal CLI (or method of creating integrations) so users are compromising their use of Signal by using heavily patched and out-of-date code written by unknown third p
113.
by jacobgold 3mo ago
Signal users who want to use it with their agents are running an unofficial extracted-and-patched `signal-cli` off GitHub. It's based on an archived official Signal repo and then patched for years by some random accounts. It looks incr
114.
by jacobgold 3mo ago
> But calling it "just software" is trivializing it in my opinion The bigger mistake would be trivializing the rest of the technology involved just because LLMs are the newest piece. LLMs are only "magic" because th
115.
by jacobgold 3mo ago
Yes, LLM agents are "magic" in the sense that "any sufficiently advanced technology is indistinguishable from magic"[1] But it's not actually magic. Technical people understand that it's just software runnin
116.
by jacobgold 3mo ago
As soon as we started unironically calling LLMs "AI" we went down the hype path. That has plenty of downsides, like stressing out the entire world and attracting cryptocurrency bros, but also the major upside massive of funding&#x
117.
by jacobgold 3mo ago
With all the coding agent options, you're choosing to trust your computer, code, and business to whichever harness, model, and provider you pick. It's not a great state of affairs, but that's where we are. Choose wisely my fr
118.
by jacobgold 3mo ago
Maybe Zuck should double down on his "spoiler" role with models rather than compete head-to-head. He doesn't have to match Anthropic or OpenAI model revenue if he can deflate theirs by 99%. All he has to do is keep spending a
119.
by jacobgold 3mo ago
This project looks great and I wish them luck. I'm just lamenting the fact that we still haven't solved the really big problem with Slack-like chat apps. The Tailscale analogy isn't quite right because there are no real netwo
120.
by jacobgold 3mo ago
You want the open protocol to have network effects, not a proprietary company's product. Email worked out pretty well, while IRC failed for reasons that are probably correctable.

More ›