Arjun
Shewalkar Computer Science · Software Engineer

I wanted to build something that feels like a deep breath.

Not another productivity app disguised as wellness, but an actual space where your mind can settle without being told what to do.

The interface strips away everything that doesn't serve stillness. No aggressive timers, no achievement badges, no guilt-inducing streaks. Just generous whitespace, gentle transitions, and typography that doesn't shout. Every interaction is designed to feel unhurried — like the app itself is breathing with you.

View Website →

Could I build a Web3 interface that doesn't feel like it's trying to sell you something?

A lot of crypto products lean heavily on flashy gradients and animated charts — always disconnected from the fact that these tools deal with real value and real decisions.

I thought about it like a driver's cockpit. Everything has a place you can rely on. You shouldn't have to hunt for information or guess what a control does. Color is used sparingly — only to show when something actually changes.

View Website →

Built in 8 hours. A study copilot that answers only from your own notes — never invents.

You upload notes across up to three subjects. Select one, ask a question, and the system answers strictly from that subject's material — citing the exact file, section, and confidence level. If the answer isn't in your notes, it refuses outright: "Not found in your notes for [Subject]." No hallucination, no guessing.

Phase 2 extended this into a voice-first teacher experience. Ask questions by speaking, hear answers read back, and maintain multi-turn context — follow up with "give an example" or "simplify it" and the system remembers the thread. All Phase 1 constraints remained fully intact: subject scoping, citations, confidence levels, and strict refusal behaviour.