Ankit varshney (@Av1theking) header
Ankit varshney (@Av1theking) profile picture

Software Engineer @ https://teapaw.org/t.co/E0mcuQ4aaE

Joined October 2023
How I shared my Claude code with my team (kind of multiplayer AI) I logged into Claude Code on an @instabox_sh machine and gave it access to github and tools we use, created a Slack bot with Socket Mode, and let that bot run a Claude session for each thread on that VM. Added the bot to our Slack workspace. Now anyone on our slack can just tag @itank to use my Claude code. It builds context from the thread messages and work from that.
1
4
53
I published a benchmarking suite for Effect, and it crossed 1k downloads 🙂 Use it to benchmark your code. It’s built for agents.
1
1
14
290
What a wonderful API response for HTTP 200 status it is
1
7
393
Built an Effect lib for postgres CDC (change data capture)
1
1
9
296
Shipped pg-replicator: an Effect-native PostgreSQL logical replication client. npm i pg-replicator
1
12
336
Implementing Electric Sql postgres Sync
8
305
holy banger @EffectTS_ this is such an important building block and you usually end up building this yourself or using something too opinionated effect.website/docs/v4/api/e…
17
24
7
611
62,993
Tried implementing MVCC (Multi Version Concurrency Control) from first principles
2
15
767
Ok yeah I can say now AI has increased my productivity atleast by 4x
14
1,327
Ankit varshney retweeted
This will piss you off but it's true
111
157
84
2,585
371,573
After so much procrastination I have finally implemented Google's mapReduce research paper in go github.com/AVtheking/MapRedu… Things I have implemented: - User defined Map and Reduce functions. - Master handling states of all the tasks, worker executing map/reduce tasks. - Storing of intermediate results from Map task to local disk partitioned into R regions. - Atomic writes by first writing results to temp file and then renaming file. - Task timeout, Task ownership, stale completion rejection using workerId+ epoch. - Handling duplicate executions. - gRPC server between master and worker(still single-process for now). Though it is not a complete implementation of the paper as I have left some parts like -> separate processes, combiner, sorted output, input splitting, backup tasks. I used AI to help me breakdown tasks, review my approach, roast me, help in understanding things.
7
315
It's funny that we recently encountered same concurrency problem which was asked from me in Interview 🙂. When we identified the core problem the solution was easy since we already know a solution for it, but the hard part was to identify what exactly an issue is, which took around 4 days
4
227
Why are you doing this to me @graphite ?
2
229