# Curiosity trails

The questions that connect Tanveer Riaz’s public projects, tools, and learning.

## Can better instructions become better visual stories?

Still unfolding

**Thread:** PromptZap → Chroa → ScriptToReel

A rough-prompt utility opened a deeper question about visual taste, agent-readable design systems, and eventually reproducible media pipelines.

**Topics:** tools, design, media

## Is the newest model the answer - or is the harness?

Deepening

**Thread:** multi-model experiments → coding agents → reusable skills

Comparing models led to a stronger conclusion: tools, permissions, context, feedback loops, and repeatable skills often matter more than a leaderboard position.

**Topics:** agents, models, skills

## How much useful AI can stay on my own machine?

Prototyping

**Thread:** OpenRouter → Ollama → MLX / oMLX

Cloud routing makes comparison easy; local inference tests privacy, control, latency, and resilience against real workflows instead of slogans.

**Topics:** models, local-ai, privacy

## What changes when agent ideas meet banking constraints?

Applying experience

**Thread:** global systems delivery → Nexus Bank demo → human review

Traceable data, bounded actions, explicit hand-offs, and clear stopping points matter when agent ideas meet banking constraints.

**Topics:** banking, agents, governance

## Can a very local problem become a useful product?

Live

**Thread:** Singapore context → verified venue data → HalalEats SG

A focused local need became a chat-first discovery product grounded in verified public data and the way people actually ask where to makan.

**Topics:** singapore, rag, product

## What can a game teach that a dashboard cannot?

Playground

**Thread:** voice game → pet leaderboard → provenance lesson

Play makes feedback immediate. It also exposes when delightful presentation outruns the evidence underneath it.

**Topics:** games, voice, data