🗓️ Timeline Devlog: From Legacy to Launch (Gemini CLI → Meta-Search Engine)
Kshanik Search is a tribute to my grandfather, Kshanik Kumar Biswas (May 9, 2021).
He served in the police force, and his life was a constant pursuit of knowledge and duty.
1) Late 2025 — The Brink of Shutdown
In my final year of college, I built Kshanik Search as more than just a project—it was a tribute to a life guided by service and learning.
But by late 2025, I hit a hard reality:
- I relied on third-party search engine APIs like SerpApi and Google Custom Search
- the low free-tier quotas and rising costs made the project unsustainable
At that point, I was stuck—because “keeping it alive” started to look impossible.
2) Turning Point — Gemini CLI changed how I built
Instead of giving up, I leaned into Gemini CLI (and its upcoming successor, Antigravity CLI).
This wasn’t just “help me write code.”
Gemini helped me understand agentic AI as a practical engineering approach:
- break complex work into orchestrated steps
- iterate with feedback loops
- architect the solution instead of patching problems
Gemini became my collaborator—not only generating outputs, but shaping how I think through systems.
3) Build Phase — A Meta-Search Engine from Scratch
I decided to stop relying on expensive aggregators and build a self-sustained meta-search backend in Go, guided by Gemini’s research + execution support.
What I implemented:
-
⚡ Concurrent Fan-out
- query multiple providers in parallel:
- DuckDuckGo, Wikipedia, GitHub, Reddit, Stack Overflow
-
🧩 Hybrid Ranking
- merge results, deduplicate, and rank them using relevance + source reliability signals
-
🧷 Memory Efficiency
- designed to run efficiently even on minimal hardware
This phase was about resilience: building a system that could survive beyond quotas.
4) May 9, 2026 — Old Laptop + ngrok = Launch Day
The deadline was emotional and real: the 5th anniversary of my grandfather’s passing — May 9, 2026.
Frontend âś…: polished and hosted on Netlify
Backend ❌: no production server ready
So I ran the Go backend on my old laptop, and used ngrok to expose it publicly.
To my surprise, it worked flawlessly.
That moment completed the transition from:
API-dependent “maybe it lasts” → self-sustained meta-search engine that actually can last.
Website link
You can view the website here: - https://kshaniksearch.netlify.app
5) After the Build — Career Growth Unlocked
This journey didn’t only rebuild the product—it reshaped my career.
By applying what I learned through system design and AI orchestration, I transitioned from a traditional software role into:
Gen AI Engineer at EY.
The key lessons that carried over were practical and industry-relevant:
- how to guide an AI agent effectively
- how to manage state in complex LLM interactions
- how to build resilient, real-world systems
6) Now — Antigravity and the Next Chapter
As Gemini CLI transitions to Antigravity CLI, I’m not sad—I’m excited.
Kshanik Search proved that AI is leverage when you orchestrate it properly.
Right now, I’m:
- moving the backend to a dedicated Oracle Cloud ARM instance
- planning a “GPT-like Knowledge Search” tool for human genealogy
Coming later: Before → After website transition
Later on, I’ll add a before and after transition phase of the website in the devlog—showing how it evolved from the earlier API-dependent setup into the self-sustained meta-search engine.
Closing: Kshanik Search Lives On ❤️
Kshanik Search lives on not just as software, but as a tribute—an ongoing way to honor my grandfather’s pursuit of knowledge and service, while building forward with the next wave of AI.
Top comments (0)