Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
snowstormsun
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
8 ms
·
61.
▲
by
snowstormsun
3y ago
It's not uncommon to have docker containers without a shell for security reasons. For example distroless.
62.
▲
by
snowstormsun
3y ago
https://sso.tax/
63.
▲
by
snowstormsun
3y ago
While you're at it, skip the internet and burn the database on DVDs, then distribute them!
64.
▲
by
snowstormsun
3y ago
Thanks, that's very interesting!
65.
▲
by
snowstormsun
3y ago
Thanks, interesting take!
66.
▲
by
snowstormsun
3y ago
Isn't one big problem of vector search that it is bounded by the complexity of the knn problem in multiple dimensions, whereas keyword search like bm25 is very fast and scalable because of inverted indices? Annoy and others need to sac
67.
▲
by
snowstormsun
3y ago
Isn't webrtc leaking the IPs of call participants anyways, regardless of which service?
68.
▲
by
snowstormsun
3y ago
Don't write your own CMS!
69.
▲
by
snowstormsun
3y ago
A fine tool would have a functional full text search...
70.
▲
by
snowstormsun
3y ago
How did you recover?
71.
▲
by
snowstormsun
3y ago
Why store raw events for 21 days in an S3 buckets when you could hash all IDs cryptographically in-memory before they're persisted to any disk?
72.
▲
by
snowstormsun
3y ago
From the explainer: > Detect non-human traffic in advertising to improve user experience and access to web content As if the goal to do that would be improving UX...
73.
▲
by
snowstormsun
3y ago
New Directions in Cryptography (WHITFIELD DIFFIE AND MARTIN E. HELLMAN, MEMBER, IEEE), https://www-ee.stanford.edu/~hellman/publications/24.pdf
74.
▲
by
snowstormsun
3y ago
I didn't mean to insult anyone. The idea of not knowing the actual performance of the model just intuitively seems to me like it's a bit of a gamble. I have only trained models in a scientific context before, where this was never
75.
▲
by
snowstormsun
3y ago
Yeah but in competitions there's a secret test set used to evaluate the model.
76.
▲
by
snowstormsun
3y ago
Why would you want to ship an untested model? That's insane.
77.
▲
by
snowstormsun
3y ago
Oh yeah, that says it all. See also: https://news.ycombinator.com/item?id=36444839
78.
▲
by
snowstormsun
3y ago
Thanks, but this reference doesn't exactly answer the original question. > Will my private code be shared with other users? > No. We follow responsible practices in accordance with our Privacy Statement to ensure that your code s
79.
▲
by
snowstormsun
3y ago
Does GitHub use code from private repos for AI training?
80.
▲
by
snowstormsun
3y ago
I see
81.
▲
by
snowstormsun
3y ago
So, couldn't it be that the authors of the paper ran the model with a random tie break and got lucky? This blog post seems to assume they had the "rand" flag deactivated. Please correct me if I am wrong.
82.
▲
by
snowstormsun
3y ago
Yeah I think it was similar for me.
83.
▲
by
snowstormsun
3y ago
The cookie banner on this website is not gdpr conform.
84.
▲
by
snowstormsun
3y ago
Wow, that's so cool!
85.
▲
by
snowstormsun
3y ago
Beware that often you need to append -n to echo to not include a newline when you base64 encode/decode something.
86.
▲
by
snowstormsun
3y ago
Password Phishing is a problem. WebAuthn solves that problem.
87.
▲
by
snowstormsun
3y ago
Does someone know for reference how big of a number that computer can factor? Or some other complex problem that is not highly theoretical...
88.
▲
by
snowstormsun
3y ago
Certificates also need to be rotated at some time.
89.
▲
by
snowstormsun
3y ago
Also, the lawn guy now has a pricing tier for different grass heights that all in the end take the same amount of work.
90.
▲
by
snowstormsun
3y ago
3. Nice idea! But how will that prevent any pinned script from loading new code by making some request to the server of Eve (simple get request, for example) and running it using something like `eval`. The problem is that the pinned scripts
More ›