yaj54

5 ms·

31.
by yaj54 2y ago
The web is distributed HyperCard. A website is basically a hypercard stack (albeit more text-centric than bitmap-centric) it's visual, card/page based, scriptable, and hyperlinked. Get him to put together a hello world page from
32.
by yaj54 3y ago
to riff on that: two selectors: (current mood, desired mood) -> recommendations. desired mood could be sorted by most commonly selected desired mood for the selected current mood. (sleepy, rested) -> nsdr playlist. (sleepy, focused)
33.
by yaj54 3y ago
I hate browser plugins - but this needs to be a browser plugin. Then I would use it while reading HN. ;-) I would suggest tackling dynamic difficulty and algorithmic selection of what words to learn, when, and how often, and then let impro
34.
by yaj54 3y ago
What resources do you suggest for learning how to do this kind of risk analysis well?
35.
by yaj54 3y ago
He's not trying to create 500 new Googles. He's trying to inspire the one kid that will actually create the next Google. Alternatively, if only one person in the upcoming generation tries to create the next Google it will probably
36.
by yaj54 3y ago
Based on your data your ideas have 1/10 odds of $1539 revenue and each idea takes on average 5.3 months to execute. You can: - Spend the next 5.3 months on a new idea. The expected value of this is $153.9 in 5.3 months. - Spend the ne
37.
by yaj54 3y ago
Are you using the SAT/SMT solvers to feed training data into a transformer or integrating the solver logic into the model code?
38.
by yaj54 3y ago
I like your UX a lot more. Modeling the llm calls as actual python functions allows them to mesh well with existing code organization dev tooling. And using a decorator to "implement" a function just feels like a special kind of
39.
by yaj54 3y ago
Is the theory tied to a specific llm? I'm interpreting it as, e.g., the llm writes the code, the solver verifies it, repeat until correct. In this situation the two are decoupled and the llm would be a drop in and thusly could be any
40.
by yaj54 3y ago
Pydantic can serialize both instances and classes/types to json and json schema. That seems quite helpful for this use case. How are you handling serialization to/from the llm? Role-your-own or are there additional libraries fo
41.
by yaj54 3y ago
gold
42.
by yaj54 3y ago
Weird. I tested with GPT4 Chat. I just tried again and got a differently worded incorrect answer. Interestingly my default responses are in the form "<answer><reasoning>." while it looks like your response was in the
43.
by yaj54 3y ago
GPT4 API and ChatGPT both get it wrong: "Sally has 2 sisters. Each of her brothers has 2 sisters because Sally and her 2 sisters make up the group of siblings each brother has." GPT4 w/ CoT system prompting gets it right: SYS
44.
by yaj54 3y ago
Interestingly the beginning of a lunar month is static (the new moon happens at the same moment everywhere on earth) while the beginning of a calendar month is sequential around the earth by timezone. The lunar cycle, earth orbital cycle (
45.
by yaj54 3y ago
Nice. Reminds me of a project I built a long time ago: https://yahnd.com/theater/?t=week . Makes me want to update it to have all the urls on the listing page so I can use your player technique on it. Or throw em in
46.
by yaj54 4y ago
What's the difference between me making food at home for me to eat and a restaurant? The restaurant is a business. What's a business? It's an economic & logistics model that produces value using known processes and wor
47.
by yaj54 6y ago
Fun and learning. And I do think my ranking algorithm generates different results.
48.
by yaj54 6y ago
Yes
49.
by yaj54 6y ago
Yep - parsing comments and submissions. I like the coreHN / non-coreHN idea. The search page [0] has links to categories that can allow you to dig into the different categories of book recs a bit but you still need to parse out which c
50.
by yaj54 6y ago
Sure - Heroku and Cloudflare for hosts. I have a lightweight JS server that handles ingestion of HN data and parses book references into a Postgres db. Site is built with Python using Starlette. There is quite a bit of caching with Redis
51.
by yaj54 6y ago
Yep - I'd for sure like to add other retailers - both for links out and for grepping for book references. Thanks for that resource - it looks helpful.
52.
Show HN: YaHNd – HN Books: The Best Books of Hacker News(yahnd.com)
55 points by yaj54 6y ago|17 comments
53.
by yaj54 6y ago
The thing that kills me is that soon enough, a "fake" paper written by GPT-3 will get published in an academic journal because it has actually contributed a new insight. It's easy to consider text generation models as "j
54.
by yaj54 6y ago
Very true. I try to mitigate this in some ways on my site, but it is certainly not perfect.
55.
by yaj54 6y ago
Nope - regular expressions all the way. :-) I keep thinking I'll move toward NER but keep finding other ways to find value first. Same answer regarding the blog...
56.
by yaj54 6y ago
As it happens I have catalogued and ranked all mentions of Coursera courses on Hacker News - you'll note that many of the mentions in this current discussion appear there as well. HN Academy: https://yahnd.com/academy&#
57.
by yaj54 6y ago
TIL about the Wakefield incident. TY.
58.
by yaj54 6y ago
@austincheney - Speaking as an American -- those in this country with the privilege of freedom have it because people have fought and died for it [0]. There actually aren't very many examples of people that have basic freedoms that ha
59.
by yaj54 6y ago
But when done systemically it likely creates the expectation that politics is done only or mostly by men, therefore discouraging women from entering politics now and in the future. Which, I would argue, is a substantial net negative.
60.
Twitter feed of the top videos posted to HN(twitter.com)
2 points by yaj54 6y ago|0 comments

More ›