Surya Moorthy

HTTP/1.1 vs HTTP/2 vs HTTP/3 Each uses TCP Protocol, but the evolution of each one has significant advantage over Networking. Follow the Thread 🧵to learn more 👇👇:

@YearsProgress Almost half way and I am in progress , this reminder gives me more attention on my time, thanks bro!

2026 is 43% complete.

@btw_iamShreyaa Totally worth it!! , small steps are the stepping stone of success, that's what this books clearly and how to approach it too.

Started reading Atomic Habits today ✨✨!!!!! Everyone seems to recommend this book 😭 Is it really that good?

@avrldotdev @ConsciousRide It doesn't, how?

@ConsciousRide Who would've thought the traditional concepts would've stayed the same for ai backend as well

90% of Agentic AI is just these 12 concepts - repeated, combined, and improved every single day. 1) Goal Decomposition & Planning: 
Turn a vague high-level goal into a clear sequence of steps. The agent doesn’t just “do stuff” - it thinks ahead, creates a plan (or multiple plans), and decides the best order. This is the brain that turns “book my trip” into actual subtasks. 2) Tool Use & Function Calling 
Give the agent real tools (web search, code interpreter, APIs, email, calendar, databases, etc.) and teach it when + how to use them. The magic isn’t the LLM chatting - it’s reliably calling the right tool at the right time. 3) ReAct / Reasoning + Acting Loops 
The classic loop: Think → Act → Observe → Think again. The agent reasons about what to do next, takes an action, sees the result, updates its understanding, and continues until the goal is complete. 4) Memory Systems (Short + Long-term) 
Short-term memory for the current task, vector memory for relevant past experiences, and structured memory for user preferences/facts. Without good memory, the agent keeps forgetting what it already did or learned. 5) Self-Critique & Reflection 
After every few steps (or at the end), the agent looks back and asks: “Did this actually move me toward the goal? What went wrong? How can I do better?” This turns dumb trial-and-error into actual learning. 6) Multi-Agent Collaboration 
Instead of one giant agent, you have specialized agents (researcher, writer, critic, executor, verifier) that talk to each other. They divide work, debate, and combine strengths - just like a real team. 7) Dynamic Workflow / Graph Execution 
Move beyond fixed linear chains. Use graphs or state machines where the next step is decided at runtime based on results. This handles complex, branching, real-world tasks. 8) Error Handling & Retry Strategies 
Agents fail a lot. Smart systems automatically retry with different tools, rephrase the request, break the step smaller, or switch strategies - all without asking the user every time. 9) Human-in-the-Loop & Escalation 
Know when to ask the human for clarification, approval, or input. Good agentic systems are confident when they should be and humble when they shouldn’t. 10) Evaluation, Guardrails & Safety Layers 
Automatic checks for correctness, safety, cost, and alignment with user values. You don’t want your agent booking expensive flights or sending wrong emails just because it “felt” right. 11) Cost & Latency Orchestration 
Use cheap/fast models for simple steps, expensive/smart models only when needed, cache results, parallelize where possible, and stop early if something fails. Otherwise your agent burns money. 12) Observability, Logging & Continuous Improvement 
Log every thought, tool call, and decision. Monitor in production, analyze failure patterns, and improve the system over time. Agentic AI in production is 10% building + 90% iteration. These 12 ideas power almost every serious agentic system in 2026. Master them, ship small agents that actually finish tasks, and you’ll be way ahead. Save this post. 
Which concept do you want me to explain deeper? Drop a number below.

I think based on @elonmusk speaking, soon might gonna have space distributed systems like how we have distributed data centers Power is the main hidden mine here which cost more than anything , if it is done by all the companies I think electricity production cost can make way less in future than today. Space engineering, communication engineering by gpus are gonna be new subjects..

@X is something else and 10x better than any social media Without doing anything and almost no quality posts today. But today with investment on previous posts I reached 1k+ impressions today, @X is something else.

@rpraggnachess @annamalai_k Congrats Pragg , you have been a role model to play chess 2 years ago during my college days..

Thank you very much @annamalai_k sir. Your support and encouragement means a lot.

@avrldotdev Nope, we have alternatives now, since ideation and problem solvers are more , we can see many solutions now..

@suryabuilds Not very performant apps now are they

Remember the times when devs utilised each byte of the RAM? Pfft, today Slack alone eats 2GB. VSCode takes 4GB & don't get me started on Claude Code & it's hunger for memory. We replaced optimization with "just add more RAM & CPU." The irony is that 2026 is shipping the most resource-hungry software while hardware prices keep climbing. Hardware got better but software got lazy faster.

Looking for an Full stack or backend Related Internship/job Have shipped around 20+ projected, contributed to multiple open source repo, have won some hackathon , have shipped freelance project too . if you hiring then do consider me wont disappoint you, or any of your network hiring do lmk Repost appreciated

name an app that's basically impossible to replace. I'll go first:

@elonmusk X is Stronger in the world for developers!!

𝕏 is strong in the east too. #1 social network of any kind in Japan.

@suryabuilds yeah pretty much feels like we're just waiting for the next disaster

> Study time : Covid > Adult time : Wars > Job time : AI > Investment time : Stock 📈 Gold📉 > Marriage time : Peak divorce rate

@GergelyOrosz if that kinda job can be by cheap people , why not meta outsource it to other company and do others things with the their engineers?

Just learned: Software engineers used to do manual data labeling at Scale AI while Alex Wang was CEO. After he left, new leadership joined, and were HORRIFIED to learn this. Stopped it ASAP Now at Meta, software engineers are assigned manual data labeling... see the pattern?

@BuildWithOm Linux still dominate the future.... can't predict which gonna die

You can know the exact answer to one question: • What technology will dominate in 2035? or • Which technology will die by 2035? Which do you choose?

if i have 100,000 users to watch a live score or a notification system, which way i can handle the communication? and why it is used? A) WebSockets B) Long Polling C) Server Sent Events(SSE)

@NG_Nikhil_Gupta I have posted what's mostly used right now bro, HTTP 1.0 not something mostly now HTTP/2 has captured over 28-30% system now...

@suryabuilds Bro , Http 0.9 or Http 1.0

HTTP/1.1 vs HTTP/2 vs HTTP/3 Each uses TCP Protocol, but the evolution of each one has significant advantage over Networking. Follow the Thread 🧵to learn more 👇👇:

@VazeKshitij Need this kinda spike for me too!!

I don't get it man.....

@tiwarisuhani_11 == , do type conversion do match them. === is a strict equality , so you can't match with different types. ex : 5 == "5" -> True , 5 === "5" -> False.

Why is == dangerous compared to === in JavaScript?

@Taniyatweets_ Should have install 26.04, it is better than the current you are using..

Finally installed Linux. What’s the first thing you’d install?