Not everything needs to become a product.
Small tools, internal utilities, one-off experiments and prototypes that solved a real problem well enough to keep.
Personal short links
A tiny Cloudflare Worker-backed short-link system for clean links under my own domain.
Coding-agent benchmark fixtures
Intentionally broken repositories used to test whether agents can discover, explain and fix real issues without being handed the answer.
Live subtitle extension
A browser-extension prototype exploring live English-to-Persian subtitles across arbitrary video sites instead of only YouTube.
Token budget calculator
A small pricing utility for estimating how many input and output tokens a budget buys across different model rates.
Provider test harnesses
Small throwaway scripts and fixtures for checking endpoint compatibility, model behavior, errors and actual usage before a video or integration.
The next tiny thing
This shelf is intentionally allowed to grow without turning every idea into a roadmap commitment.
Coding-agent benchmark fixtures
Small repositories with known logical, security and completeness problems. The benchmark stays useful because the agent is not told exactly what is broken; discovery is part of the job.
Live subtitle extension
A browser-extension prototype for turning English video audio into Persian subtitles across arbitrary sites. The hard part is not the overlay; it is reliable browser audio capture, transcription compatibility and keeping latency reasonable.
Token budget calculator
A practical calculator for turning model pricing into something easier to reason about: given a budget and an input/output split, how many tokens can a workflow actually afford?
Provider test harnesses
Quick scripts and fixtures for checking whether an API really behaves like the interface it claims to support: endpoint shape, streaming, tool calls, model IDs, error behavior and usage reporting.