Aha, it makes total sense. This number sounds much more reasonable than “669 GB”, since the actual total size of processed frames would be like 10-30 GB.

(Not downplaying anything. Doing-at-home always requires some math on practicality)

> Total compute time 67h 40m 42s

I’m just curious tho — is there any paying options that can accelerate this kind of process? Just spin up GPU instances?

In fact, Opus does the same. It finishes the job, and redo it from scratch before presenting the result to the user. This happens even for simpler writing tasks especially when I instruct it to create a text file.

The game that I suggested earlier assumes that (1) those in power seek to maximize their gains, (2) and such behavior is NOT aligned with social gains. So, basically, a never-ending arms race between the authority and the people. (check Leviathan by Thomas Hobbes if you’re interested)

However, China (and also South Korea) got a weird alignment in interests between the authoritarian government and the people — both of them somehow sought after national economic growth. Since their interests are already aligned, they didn’t play the game that I suggested above.

I think such alignment cannot be reproduced through games nor social interaction b/w powers — rather, it is more of a humane part of the history.

It’s a sort of game theory stuff here. A structured class hierarchy makes it inherently more difficult for individuals to challenge the authority in power, even under the democratic government. The system imposes an additional risk of social backlash/punishment/retaliation for anyone who attempts to disrupt the established order, thus people have more reasons to stay back. This kind of risk is largely absent in more egalitarian/classless societies.

I think seeing it as an “align”-ment problem puts too much blame on the local side. Also, autonomy has nothing to do with the problem of misalignment.

In authoritarian systems like China, mis-alignment with authority can carry serious political and social risks, so people are easily pushed toward dishonesty. What happened under the Mao’s rule is simply this; local officials were too afraid of criticizing the very father of the revolution, which could be interpreted as attacking the legitimacy of the revolution itself. It was a side effect of over-concentration, and gaining more control over local would have not made any differences.

Deng was successful only because he was exactly aware of this problem. In his speeches on the government reform (the Open-Door policy), he explicitly pointed out over-concentration as a major issue. He not only eased the concentration of power, but also redesigned the incentive structure, so that officials can adopt objective measures and even try their own experiments.

So my pipeline is mostly stuck with: (1) brainstorm with Gemini (2) plan with GPT (3) augment the plan with Gemini (4) execute the plan with Opus.

Perhaps, given all the materials filtered and prepared, Opus is good for verification.

I think this shows how Noctua value their customers, including myself. I really love how they are nice to their customers — both their products and services — especially because experience like this is getting more and more scarce. I really appreciate their work.

The “AI” “technology” is an easy excuse to create artificial information gap in the era of the interconnected.

My view is actually the opposite. Software now belongs to cattle, not pet. We should use one-offs. We should use micro-scale snippets. Speaking language should be equivalent to programming. (I know, it's a bit of pipe dream)

In that sense, exe.dev (and tailscale) is a bit like pet-driven projects.

Personally, I think a possible angle of attack for a new practical social network protocol is data management, as the amount of data people generate, consume, store, and share is enormous these days. More like, manage data conveniently, and share them easily as a side-effect.