161 Photos and videos
Pinned Tweet
12 Apr 2025
got accepted to work on project-loki on shipd!
shipd (by Datacurve YC W24) is an invite-only platform for devs to solve coding problems.
2
12
1,507
May 24
this one demo got me 1000 new users in under a week!
May 24
pushed an update after months and I gained around 1000 users in a week!
keep building!
ALT https://chromewebstore.google.com/detail/zelix-password-protected/eikknpnckalfaafkjkkjbikfbkbkdnfc
1
2
145
May 24
pushed an update after months and I gained around 1000 users in a week!
keep building!
ALT https://chromewebstore.google.com/detail/zelix-password-protected/eikknpnckalfaafkjkkjbikfbkbkdnfc
1
1
98
May 17
Building a clipboard history manager with semantic search capability.
Used NLEmbeddings from apple.
When the clipboard text lie distant from each other on the Vector space, the query seems to work well, not otherwise.
gotta try huggingface.co/MongoDB/mdbr-… from @MongoDB instead
1
60
Mano retweeted
May 4
Love seeing Naomi Osaka honor the CLRS Algorithms textbook at this year's Met Gala
109
1,735
18,112
614,345
May 5
Fixed a bug! I can sleep now!
Was stuck with my Solana NFT minted as a token failing to link with the metadata I already created,
This random video helped me out, gold
2
73
this is not a bounty this is a seed round
12
7
170
23,097
Mar 27
Talking to the end users is the best thing ever! You understand how you think your app is being perceived vs how it actually is being perceived!
3
7
69
Mar 13
finallyy ... solved the bug that was bugging me for a week!
I am building a dApp on Solana and was struggling to create a signer using in-browser wallets... after days of debugging its resolved.
I know its just so random, but this got me happy lol
1
3
81
Feb 28
over the couple of weeks I learnt about PDAs in solana and understood how proof of history helps consensus perform transactions in parallel.
- got started with anchor
- built a simple program to mint nft into a user PDA vault
- deployed Metaplex program on local test ledger for rapid testing development
- learnt about cross program invocation
building on @solana has been supa fun ... now im onto building the frontend for ux
6
127
Jan 26
spun up a local solana ledger and deployed the program!
now its time to build some games on solana
2
5
124
Jan 25
wrote my first program on solana!
used Rust and native solana-program crate
3
11
439
30 Dec 2025
Learn something new part #3
There are two models to replicate data in distributed systems:
Leader-Follower model
Leaderless model
In Leader-follower model, we have one DB that handles all the writes and its followers only handle read operations. When Leader gets a write request, it synchronously or asynchronously updates the follower nodes.
In Leaderless model, we have N number of nodes and each can be used to handle writes and reads. Here we use Read-Write quorums to make sure that data read at any point contains latest data. what is Read-Write Quorum now?
27 Dec 2025
Learnt something today new part #2;
What happens if two DBs from different zones receive record update on same data?
Lets say we have
Zone A -> berlin
Zone B -> tokyo
DB A has record with data 'cat'
So does DB B has record with data 'cat'
At some timestamp t0 someone edits 'cat' to 'cats' in DB A and someone else edits 'cat' to 'catz' in DB B
# When they communicate with each other what data will be written in record? cats or catz or what?
2
5
88
30 Dec 2025
Let us say we have N nodes. Read-Write Quorum says that if we write to atleast W nodes and read from atleast R nodes, such that W R>N, then it is guaranteed that whenever the user reads data from R nodes, atleast one node contains the updated data. In the diagram, N = 5, R = W = 3, W R > N
1
49
30 Dec 2025
For read heavy operation we can make W to be a very large value and R to be a small value maintaining R W>N rule. This reduces the number of reads across nodes to fetch data.
For example: Having R = 2 and W = 4, makes sure that data only needs to be read from atleast 2 nodes to make sure it is upto date, it is helpful in handling Read-Heavy operations.
Similarly, for write-heady operation we can make R to be very large value and W to be a small value. That way the number of writes across nodes becomes easy.
36
27 Dec 2025
Learnt something today new part #2;
What happens if two DBs from different zones receive record update on same data?
Lets say we have
Zone A -> berlin
Zone B -> tokyo
DB A has record with data 'cat'
So does DB B has record with data 'cat'
At some timestamp t0 someone edits 'cat' to 'cats' in DB A and someone else edits 'cat' to 'catz' in DB B
# When they communicate with each other what data will be written in record? cats or catz or what?
25 Dec 2025
learnt 2 new terms today;
forward and backward compatibility
forward compatibility: old code can read new data
backward compatibility: new code can read old data
2
3
211
27 Dec 2025
Why LWW fails? because timestamps across different time zones would be not consistent. Also this results in data loss
cats 10:00:39 DB A
catz 10:00:35 DB B
lets say catz was written after cats but becuase of difference in timestamps cats will be written across DBs. The other issue here is Data Loss.
1
24
27 Dec 2025
Manual Conflict Resolution:
Edits are persevered and a human comes into the picture to resolve the changes via GUI
21