Rishav

Software Engineer @ Amazon

क्या तकल्लुफ़ करें ये कहने में
जो भी ख़ुश है हम उस से जलते हैं

[Rishav @aptrishu]

@KirtiAzaad @grok I always forget to look at such events from these “important” angles. What is wrong with me?

[Rishav @aptrishu]

@7signxx

[Rishav @aptrishu]

@ChShersh Jira is a must?

[Rishav @aptrishu]

@thechosenberg I am thinking of an intuitive way to explain that.

[Rishav @aptrishu]

@thechosenberg Probability depends on how likely each result is, not just how many possible outcomes exist.

[Rishav @aptrishu]

Is there any other LLM in market for that?

[Rishav @aptrishu]

Why are chatgpt(4o) and sonnet 3.5 so bad at interpreting sheet music?

[Rishav @aptrishu]

@ChrisMartzWX Just curious, as more ice melts, the rate of melting would also increase, no?

[Rishav @aptrishu]

@muratdemirbas @hillelogram Same here 🙋🏻‍♂️ learned TLA+ mostly from your sessions.
Hillel’s https://t.co/nduDcda7yh and emailing him the doubts were also a big help.

[Rishav @aptrishu]

@DominikTornow but it still gets everyone on the same page. Leadership can help, sure, but it doesn't seem like the core idea. It's all about that agreement in the end, however we get there.

[Rishav @aptrishu]

@DominikTornow The reason I had mentioned that 'leader election' isn't the essence of consensus is because it feels to me more about everyone agreeing on the same state. Classic Paxos, for example - it doesn't have a fixed leader, there could be multiple concurrent leaders running ....

[Rishav @aptrishu]

@DominikTornow something like agreement on a shared state.

"shared" is redundant in dist systems, maybe just "state agreement" then.
Not able to think of a better term.

[Rishav @aptrishu]

@DominikTornow And I don’t like the term “data replication “ here.

[Rishav @aptrishu]

@DominikTornow Data replication.

In many algorithms, leader election is just a means to an end.

[Rishav @aptrishu]

Thought I had consensus algorithms all figured out, then I tried writing formal specs for them... Turns out, it’s way easier to spot bugs when you’re *not* imagining them 😶

[Rishav @aptrishu]

@benjamenpyle @alexbdebrie DynamoDBMapper ruined it all, people started annotating their actual domain models with ddbMapper annotations and started using ddb like a relational db (this comparison to relational db may not be exactly correct)

[Rishav @aptrishu]

@benjamenpyle @alexbdebrie Yup, It makes so much sense. People get so pissed looking at my ddb tables because I have pk, sk instead of naming them better.
IMO, it keeps my table flexible, pk, sk can take any meaning, tomorrow I can put a new data collection there.

[Rishav @aptrishu]

@a_prout @MarcJBrooker Failure probabilities are multiplicative when they are independent?
Doesn't seem correct.
They should be multiplicative when they are interdependent.

[Rishav @aptrishu]

@scarletinked @threadreaderapp unroll

[Rishav @aptrishu]

@doer_uvc @AkasaAir Lol I agree ☝️
Had a different experience but equally shit.