experimentwith.abhijeetkakade.in

Experimenting
with ideas.
Shipping what sticks.

Small tools, late-night builds, random ideas.
Some worked. Most did not. All are here.

Explore ExperimentsView Latest Build

// what this is

Not a portfolio.
A playground.

01

No perfection

Shipped imperfect beats never shipped perfect. Every time.

02

No overthinking

Idea at 11pm, prototype by 3am. That is the whole plan.

03

Just building

If it solves a problem I had, it belongs here. Period.

04

Failure counts

Failed experiments get their own section. No hiding.

// all experiments

Everything
I shipped.

Crop

Sprite sheet cropper built because I was tired of doing it manually at 3am.

Success#001

RoboForge

GitHub plus HuggingFace for simulation environments. On Avalanche. India-native.

In Progress#002

AVAXVERSE

Web3 freelance marketplace on Avalanche. DID plus soulbound reputation plus DAO governance.

Success#003

NextFlow

Krea.ai workflow builder clone. Got too complex. Scope crept. Abandoned.

Failed#004

OnlyPngs

Autonomous NFT trading agent on Avalanche using Gemini 1.5 Flash. IEEE paper published.

Success#005

ShadowRealm

Isometric canvas game with AI pathfinding. Built in one sitting. Combat system works. Game doesn't.

Failed#006

This Site

Building a living lab of experiments. Meta-experiment: can a portfolio be anti-portfolio?

In Progress#007

Blog v3

Next.js blog from scratch with Markdown parsing. Notion-dark aesthetic. Actually ships content.

Success#008

Portfolio v1

Traditional portfolio site with a skills list, project cards, and a contact form.

Failed#009

// things that didn't work

Failed experiments.
Learned lessons.

Failure isn't the opposite of success here. It's part of the process. These are the experiments that didn't work out, and what I took from them.

NextFlow

What was attempted

Full Krea.ai-style AI workflow builder with React Flow, Trigger.dev, and multi-model support.

Why it failed

Scope expanded every day. No clear MVP. Got lost in building the builder, never the product.

What was learned

Define the smallest possible useful version first. Ship that. Expand later if it earns it.

ShadowRealm

What was attempted

Isometric canvas game with AI villain pathfinding, traps, and combat in a single HTML file.

Why it failed

Game loop worked. Combat worked. But gameplay wasn't fun. A working game is not a good game.

What was learned

Technical completion is the easy part. Design and feel are what make things worth playing.

Portfolio v1

What was attempted

Traditional portfolio site with a skills list, project cards, and a contact form.

Why it failed

It looked like every other developer portfolio. Nothing stood out. No one remembered it.

What was learned

The format is the message. This site exists because that version failed.

// my building philosophy

Why I build
like this.

Ideas > Perfection

A rough idea that ships teaches you more than a perfect idea that does not.

Speed > Planning

The plan changes the moment you touch the keyboard. Start faster.

Shipping > Thinking

Thinking is free. Shipping creates feedback. Feedback is what you actually need.

Failure = Data

Every failed experiment tells you something true. That is worth more than never trying.

Got an idea?
Build it tonight.

Let's Build Together