beacon294

16 ms·

241.
by beacon294 3y ago
In product sale terms, $15 once is also very close to $80 once. They are differentiating on "the main segment of toaster buyers". It would be nice if someone had an incentive to differentiate for people willing to pay more.
242.
by beacon294 3y ago
You can also consider a virus a loosely coupled form, which really increases the size of the virus, since it requires its platform to propagate.
243.
by beacon294 3y ago
This kind of an old axe to grind at this point, many engineers involved in these kinds of choices would rather be coding but see the need for good navigation. It is hard to decide when to differentiate roles as a company grows.
244.
by beacon294 3y ago
Yes, i do this with 3x Archer C7 V2.
245.
by beacon294 3y ago
If you enjoyed Echopraxia, you may want to read it twice. I thought I got it the first time, then I read a Peter Watts commentary and realized how much I missed. Planning to reread both soon.
246.
by beacon294 3y ago
They have lower saturated fat than lean beef, good protein content and blend, and overall less strain on the organs.
247.
by beacon294 3y ago
Things don't have to be momentous to have a measurable impact in the long term. Unwinding intangible property rights has tremendous value for the public, and we should get it sooner.
248.
by beacon294 3y ago
Two words: "rounded corners"
249.
by beacon294 3y ago
Which books did you find worth it lately?
250.
by beacon294 3y ago
Curation and curation tooling both add value, similar to how purifying chemicals or making food adds value. And that isn't the only dimension either
251.
by beacon294 3y ago
Imagine how much more complex things are now :)
252.
by beacon294 3y ago
That can be an organizational problem, reliability should be impact.
253.
Ask HN: How do you plan your personal work?
13 points by beacon294 3y ago|9 comments
254.
by beacon294 3y ago
> In other words no one should be at liberty to seek the ceiling until EVERYONE is out of the basement. Can you explain your reasoning?
255.
by beacon294 3y ago
I don't know, hopefully it will help set general expectations.
256.
by beacon294 3y ago
It is Llama-2-70b-chat. I quantized it to 2q_k using `quantize` with llama.cpp.
257.
by beacon294 3y ago
The answer seems kind of low quality, but it got further than others. It could probably self correct.
258.
by beacon294 3y ago
I have quantized Llama-2-chat-70b to 2q_k with llama.cpp then run on a 4090 for 50/83 layers. I was able to get 1.3 tokens per second. It answers this question, which smaller models have trouble with: <s>[INST]Hello. The functio
259.
by beacon294 3y ago
What are your top 5 (and even 10)?
260.
by beacon294 3y ago
It's quite interesting in its own context, otherwise, you may not be interested.
261.
by beacon294 3y ago
draft data exists, might be a fast alternative