Pratik

@LadyAshBorg So much switch even with agents is very very costly. You cannot do it and expect the product to be of good quality.

Before Claude/Codex, programming was mostly single-threaded human execution. One task. Full focus. Context switching was expensive, and most importantly truly truly hated. I remember product & delivery teams endlessly complaining about how CS and sales teams were constantly asking them to context switch. I'd walk on eggshells before asking Anand a question while he was coding. Now, its like .. Gurllll who is this with all these tabs!?!? Vibecoding has turned developers into insane multi-taskers and schedulers. Kick off one job, supervise another, debug a third, and learn to manage the latency between agent runs.

@cristian_frunze @arankomatsuzaki same issues happening

@arankomatsuzaki I'm on Pro 20x and I still hit my weekly usage limits in less than 3 days. Is it just me who's bad at managing it or how are you getting so many tokens out of it?

I spent 1.7B tokens on Codex (Pro 5x) and 80M tokens on Claude Code (Max 20x) yesterday. Guess which one gave me a usage-limit warning.

@bindureddy Are you sure. I wasted time earlier on another Chinese model which had a huge hype and turned to be dud.

Finally figured out how to run DeepSeek V4 Pro correctly The evals are mind blowing... Yes, it is even better than Opus 4.7 Medium!! More coming...

@Physicsvers Models have become very expensive, is ai really replacing anyone in tech or is it just that the bigger companies are stopping new projects to fund datacentres. Also learning right usage of ai coding takes long time.

Sunday: Tech professionals are the first to feel the impact when AI hits hard, while the outside world keeps running as usual, most people don’t even notice or care about it.

@zuess05 Even 200$ is not sufficient and runs out, what will you do with 20$ subscription

Serious question. For decades, the standard advice was to "pick a niche" and become a highly paid specialist. Now Claude has the combined knowledge of every specialist on earth, instantly available for $20 a month. What exactly are we supposed to tell kids to major in when every technical skill is just a prompt away?

@Prathkum However the kind of resources required is changing. Probably juniors are no longer able to add much value unless they are really driven.

@Prathkum I don't see how AI is reducing work. Its only changing the kind of work. Plain vanilla surely is simpler and does not need a lot of engineers. Any place with domain knowledge, requirement for testing and product knowledge is not easier in terms of resources.

All the layoffs happening these days are not because of AI. They are because of a bloated workforce.

@IbinIssac600 @Kunal_K_Kumar What is this limit

@Kunal_K_Kumar Great real-world test. Static IP + AlgoID compliance changes the risk profile more than most expect. Strategies that relied on latency arbitrage took the biggest hit. At Firefly by Fintrens, we rebuilt around OTR limits from day 1 - cleaner executions, fewer disqualifications.

Is algo trading MORE profitable in 2026 under the new SEBI rules? I ran 5 real strategies (static IP, Algo-ID, new OTR limits). Here is what happened⬇

Day 62 — NiftyGram Never knew how days change from tea & biscuits to tea & laptop. Same cup. Different priorities. Strange thing is — it still feels good. Sitting down, working on this, figuring things out as we go. Still building NiftyGram.

Coming back to the desk today — laptop open, code in front — felt better than expected. Building hits different. Back to NiftyGram.

Day 61 — NiftyGram Funny how work follows you. Even when you step away for a bit, there’s always something pulling your attention elsewhere. Had to focus on a few other things around the business. Important, but different.

@NyshaDev Building quant finance

Quietly building something? 👀 Same here. If you're into Tech, dev, AI, data, design, or startups ... drop what you're working on below 👇 Let's connect...

Now back to doing what builders do best — finding it, fixing it, and learning from it. Still building NiftyGram.

Day 50 — NiftyGram Yesterday was one of those rare days when everything worked. So we did the responsible thing — took a small break and enjoyed it. Today… balance has been restored. A beautiful bug has appeared.

Because every builder knows when things run too smoothly, An interesting bug is probably loading in the background. Meanwhile, progress continues. Small improvements. Cleaner flows. Getting closer to launch. Still building NiftyGram. Let’s see what tomorrow breaks.

Day 49 — NiftyGram You know those magical days when your code runs perfectly and no errors show up? Today was that day. Which is great… but also slightly terrifying.

Turns out building products is mostly deciding what not to build. Still refining. Still debating tiny details. Building in public.

Day 48 — NiftyGram Product meetings be like: “Let’s keep it simple.” adds 4 more features So today we did the opposite. Removed a few things from NiftyGram. Because good platforms add value. Great ones remove noise.

@suresh_089 Thank you Suresh

@pratik1762006 Day 47 and still going strong! The stock market analogy is so accurate — every bug fixed reveals two more lurking underneath. Keep building, the insights feature for advisors sounds like a valuable addition!

Day 47 — NiftyGram Debugging a fintech product feels a lot like the stock market. You fix one issue. Two new ones appear. Today was about improving how advisors share insights on NiftyGram.

Because good research shouldn’t compete with “Buy this stock before 3:30 PM” WhatsApp tips. Still refining. Still questioning.