Attending DevFest Noida.
Rooms full of people actually building with AI — not just talking about it.
My pick: the Agentic Web session later today. Will share takeaways.
#GDGNoida#DevFest2026
Built an AI agent from scratch in Python — no LangChain, no frameworks.
Compared it with Gemini on the same prompt.
Biggest takeaway: The quality gap wasn't the model. It was the pipeline.
Next: Multi-source retrieval, planning & reflection loops.
#BuildInPublic#AgenticAI
LLM answering questions is easy.
Making it verify its own answers is where AI agents get interesting.
Step 2 of my genesis-agent:
Answer → Judge → JSON → ✅ done=true
#AgenticAI#Python#LLM
Most people use LangChain and never understand what's happening inside.
I'm doing it the hard way — building an AI agent from zero.
Step 1 done: raw LLM call with Groq + Llama 3.3 70B. It works.
More steps coming 👀
#BuildInPublic#Python#AIAgent#LLM
Turned my boring default GitHub profile into something that actually represents who I am as a developer 🔥 Small win, but it matters.
GitHub Link- https://t.co/lsZ6SD444i
What followed was a 6–7 hour Linux troubleshooting marathon...
Installed Ubuntu 24.04 alongside Windows today on my RTX 4050 laptop… and I got a full Linux troubleshooting masterclass 😭🐧
Expected: Install Ubuntu and start coding.
Reality: Welcome to kernel debugging💀
Still facing some graphics/suspend related instability though — Linux + NVIDIA is truly a different world 😅
Honestly, this felt less like installing an OS and more like debugging production infrastructure.😅
#Linux#Ubuntu#NVIDIA#DualBoot#Programming
While learning, I came across an interesting real-world issues: "Integer Overflow"
The Boeing 787 faced a bug where its 32-bits counter overflowed after 248 days of continuous power, risking all electrical power loss. The temporary fix was to reboot the plane every 248 days.