Utpal Sen

ai has sucked all the fun out of programming

@GithubProjects github. mostly because it's down.

WHAT ARE YOU NOT BUILDING TODAY?

next to her

read more want to dive deeper into how databases really work? check out use-the-index-luke.com for sql performance tips, and sqlite.org/queryplanner.h… to see exactly how a database executes joins.

modern sql joins aren't just for tables anymore. you can use LATERAL joins to run a "for-each" loop per row (like getting top 2 orders per user). you can even use CROSS JOIN UNNEST to explode a json array trapped inside a single column into queryable rows.

what is a sql join? most people learn joins using venn diagrams. but venn diagrams are a lie. they show intersections, but sql joins actually multiply data. a join is fundamentally a cross product followed by a true/false filter. here is how they really work.

@amritwt this is agi

this is what I wanted wispr flow to be like

@https_ghanshyam not worth it the $20 plan hits its quota limit within an hour lol

Since the morning i confused between should I need to buy claude pro or not??

it's a great lesson in systems design. we instinctively throw concurrent threads at scaling problems which usually just introduces race conditions and lock overhead. sometimes the most scalable architecture is just a single thread that never ever blocks.

the true bottleneck in most systems isn't the cpu, it's waiting for network or disk. redis avoids disk by keeping the dataset in memory and avoids network blocking by making all sockets non-blocking. it just drains the ready queues at raw memory speeds.

i was learning today about how redis sustains millions of tcp connections on a single thread. everyone assumes you need massive thread pools to scale i/o. it's completely backwards. the mechanics of why one thread wins are actually beautiful.