kristjansson

4 ms·

61.
by kristjansson 4mo ago
It’s insane enough to consider undertaking the risk of a moon or mars colony. But doing so with him as the sole load bearing link in your supply chain? What if mars turns out to be woke and you’re fed into a wood chipper like USAID?
62.
by kristjansson 4mo ago
It's easy[1] to promise, it's hard to deliver. I hope the best for them. [1]: https://magic.dev/blog/ltm-1 (note the date)
63.
by kristjansson 4mo ago
>A full breakdown of the mechanism and how it compares to FlashAttention, DeepSeek sparse attention, and recurrent architectures is in the Technical Report. Oh they did publish details lets read the technical report! > The mechanism b
64.
by kristjansson 4mo ago
Seems like a lot of machinery over-and-above running the same review n times + aggregating the result. What led you to the design?
65.
by kristjansson 4mo ago
Learning that those shortcuts are arbitrary, application-defined conventions is the first step toward enlightenment.
66.
by kristjansson 4mo ago
Why does emacs have to be for everyone? Its attraction, to this user at least, is the lifelong process of adaptation and discovery that molds it to the user, and the user to it. It’s like a pet. I love my dog, I’m happy to tell you about my
67.
by kristjansson 4mo ago
C-u 0 M-x asi-optimize-everything (The prefix arg disables paperclips, for obvious reasons )
68.
by kristjansson 4mo ago
or that intelligence is merely the composition of many redundant, lossy, ~random components
69.
by kristjansson 4mo ago
https://thickets.mit.edu
70.
by kristjansson 4mo ago
It’s expensive, but not unfathomably, esp in an open source setting where capable people might contribute high quality data for post training (worked problems, code reviews, feedback, …) gratis instead of at immense cost.
71.
by kristjansson 4mo ago
It’s actually ok to be more critical of a government that’s capricious than one that merely advances polices one disagrees with
72.
by kristjansson 4mo ago
Dual use is definitely not a new thing; sacrificing the ostensible primary use for the secondary seems to raise the author's ire (and mine).
73.
by kristjansson 4mo ago
> the shelves were being cleared to make space—not for more books but for space itself. ... The new library has four floors. Two of them feature books Despite the frenzy of building at most American universities, the library is forced to
74.
by kristjansson 4mo ago
> 42 kW/kg Is that right? One could get ~200hp from something the size and mass of a small dog?
75.
by kristjansson 4mo ago
Update: not any more!
76.
by kristjansson 4mo ago
Obscuring core logic is the most egregious part of most agent frameworks. One needs a clear view of what, exactly, is being sent to the underlying language model, and what's coming back. Everything in an 'agentic' applicati
77.
by kristjansson 5mo ago
I think MTP Gemma4 support is still WIP https://github.com/ggml-org/llama.cpp/pull/23398 ?
78.
by kristjansson 5mo ago
> quantization 12b means 12G @ 8 bits/param (basically lossless) and 6G at 4 b/p (generally accepted 'pretty close' level). Not too bad? But TBD how well the base model performs before thinking too much about quantiza
79.
by kristjansson 5mo ago
yes yes the solution to some debatably valid social science research is to burn the entire epistemological method and practice to the ground. We'll surely find something better to replace it in its absence.
80.
by kristjansson 5mo ago
> 137B-A5B Yeah, not a 5B param model as the earlier title implied!
81.
by kristjansson 5mo ago
GOOG was a $20B company against a $10T S&P 500. SpaceX is going to be a $2T company against a $60T S&P 500.
82.
by kristjansson 5mo ago
Sure, it'll only be ~2% of the index if it opens where they want to. But in the downside case where it meanders long enough for significant amounts of its stock to make it in to public hands and then goes to 10x revenue (i.e. down 90%
83.
by kristjansson 5mo ago
Noting for reference that Gemma4 MTP work is in progress[0] on llama.cpp; similar work for Qwen3.6 landed recently and has been great thus far. [0]: https://github.com/ggml-org/llama.cpp/pull/23398
84.
by kristjansson 5mo ago
I mean catch and release is better than being dinner, and the prevalence of fly fishing as a hobby goes a long way toward garnering public and financial support for restoration activities.
85.
by kristjansson 5mo ago
The Bar is also a not-great signal of actual ability. It is a great signal for 1) base knowledge and reasoning capacity and 2) sticktoitiveness / willingness to meet ~arbitrary expectations to work in a given industry ... which is al
86.
by kristjansson 5mo ago
Yeah I wrote “by a third party” then thought of them and revised to “trusted neutral party”. Has to be someone with ~no financial interest in the candidates success
87.
by kristjansson 5mo ago
Internship etc. proposals miss that status quo job seeking is a parallel process for applicants. An internship model means each candidate can only “consider” one employer at a time. I think this is what Steve’s campfire proposal is trying t
88.
by kristjansson 5mo ago
Per TFA Steve got hired by Amazon and then kicked off the acquisition of his former company.
89.
by kristjansson 5mo ago
Whatever clause in whatever contract is just the starting point for negotiations if things go south.
90.
by kristjansson 5mo ago
Option C is, yknow, pass law (and ratify amendments) s.t. the interpretive power of the court is less of a factor

More ›