chr15m

5 ms·

61.
by chr15m 2mo ago
Because you can build on Nostr without anybody's permission or API keys for any server or anything like that.
62.
by chr15m 2mo ago
> network effects favor the largest No. You can write a client that does not use popular relays and it will work fine.
63.
by chr15m 2mo ago
You can use a technology without agreeing with the views of other users of the technology. You don't have to follow those people.
64.
by chr15m 2mo ago
I think the opposite can be true. Billions of humans could be happy on the Earth we have now for a billion years.
65.
by chr15m 2mo ago
> micro-managed Perfect, this is exactly how I refer to my LLM workflow too!
66.
by chr15m 2mo ago
40 trillion coconuts per year.
67.
by chr15m 2mo ago
No. Around 3.5 million tonnes of coconut oil per year. Around 4.5 billion tonnes of fossil fuel oil per year.
68.
by chr15m 2mo ago
Around one million coconuts required to fly from Tokyo to Los Angeles. Around 40 million flights per year. 40 trillion coconuts required for all flights, or 1000x current production. To scale up coconut production we need 12 billion hectare
69.
by chr15m 2mo ago
Yep local models will be good enough for most things you need to do, in the same way as most people need a laptop not a supercomputer.
70.
by chr15m 2mo ago
> Explicitly ask the harness to reload agent.md. "Reload agent.md" is enough when I see code quality dropping. Having the LLM re-read the file is really silly and a common bug in harnesses. Even sillier is when the harness allo
71.
by chr15m 2mo ago
Brevity tho.
72.
by chr15m 2mo ago
Ethics should be part of the RL loop.
73.
by chr15m 2mo ago
This was useful to read, thanks.
74.
by chr15m 2mo ago
We are the chauffeurs. They're going to drive their own computers now.
75.
by chr15m 2mo ago
True, but the richest government and military in the world depends on them. Too load-bearing to fail.
76.
by chr15m 2mo ago
> enable cheats in a game Great analogy. Life does not have cheats though. Things are changing fast but the limits of physics still apply and reality has a surprising amount of detail[1]. I find myself feeling the way you are feeling whe
77.
by chr15m 2mo ago
The leveraged US labs are cooked. Debt's coming home. There will be bailouts.
78.
by chr15m 2mo ago
Meta's Glimmer is cracked for local inference!
79.
by chr15m 2mo ago
I have a good feeling about this. Unbelievably cheap and remarkable technology is coming. History repeats.
80.
by chr15m 2mo ago
> worked on it for an estimated 300 hours If humans aren't needed why have you had to spend 300 hours? I upvoted your comment BTW because I think you might be right, but I'm not sure. I still see people doing a lot of work, des
81.
by chr15m 2mo ago
Interesting, thanks! > habitually drinking... sugar-sweetened-beverages Yes, terrible idea, avoid!
82.
by chr15m 2mo ago
Interesting, thanks!
83.
by chr15m 2mo ago
You can control your CVD risk. Cutting saturated fat to under 15g per day and increasing intake of viscose fibre (e.g. beans) will reduce your LDL particle count in a few weeks, which reduces your CVD hazard ratio. You can measure your LDL
84.
by chr15m 2mo ago
When prices spike people generally try to make more of the thing so that they can profit from the high prices. It's very likely this is what will happen. The hardware for running inference will get very cheap and abundant, and the soft
85.
by chr15m 2mo ago
Yes, this is literally what they are doing in countries where these measures have already been implemented. First hand experience.
86.
by chr15m 2mo ago
> use a regex engine with runtime guarantees That's a good idea. If you have to use one that supports NP-hard features I think a configurable CPU time timeout is also a reasonable backstop, just as configurable timeouts are reasonab
87.
by chr15m 2mo ago
Why would you use wall clock time for this?
88.
by chr15m 2mo ago
I'm just using the defaults, through Claude.ai on the web.
89.
by chr15m 2mo ago
I'm just using the defaults.
90.
by chr15m 2mo ago
Yes, in the same way constant software malfunctions and errors are normal and tolerated.

More ›