English | 日本語

I follow problems, not layers. If an idea could move technology forward and I can test it in code, I want to build it.

AI-assisted development lets me explore a wider range of systems. The goal is not just to generate code, but to understand what runs beneath it. I publish what I learn as open-source tools, and I want to carry the same ideas into web services and apps so more people can access emerging technology.

"Imagination is more important than knowledge."

— Albert Einstein

albert-einshutoin is a small homage to Albert Einstein, the person I respect most. For me, imagination is the starting point; engineering is how I test it in the real world.

1. Start with a concrete problem from production or development.
2. Question whether the problem is being solved at the right layer and whether the current abstraction still makes sense.
3. Form a technical hypothesis and test it in working software.
4. Publish the result as OSS so others can inspect, challenge, and extend it.

I am currently exploring programmable edge computing, native performance, safe local and CI infrastructure, and AI memory systems. These are areas I am exploring now, not boundaries around what I build.

• cdn-security-framework — Released JavaScript

  Born from production work with WAFs, application-level rate limits, and Nginx. It explores how lightweight filtering in Edge Functions and protection in a WAF can work as separate layers to reduce obvious anomalies before traffic reaches the backend.

• quant-cache — Released Rust

  Built to explore whether offline QUBO and simulated-annealing search can improve CDN cache-policy selection under economic and capacity constraints.

• TenantScript — In progress TypeScript

  Builds a control plane around Cloudflare Workers primitives for tenant-specific plugins, permissions, approvals, versioning, and audit logs. Live validation with Dynamic Workers is still in progress.

• lazy-image — Released Rust Node.js

  Started after I chose sharp for image processing in a production Node.js system. Maturity and performance explain why an engine becomes the default—but could a new engine earn its place through different strengths? Image processing was a new domain for me, so I used AI-assisted development to explore bounded memory, secure defaults, and smaller JPEG outputs in the current reference benchmarks with a Rust-powered Node.js engine.

• i18next-turbo — Released Rust N-API

  Started from wanting the fastest possible i18n checks in CI. Rust and SWC accelerate key extraction, while N-API keeps the native implementation accessible from the existing Node.js workflow.

• mockport — MVP Go Docker

  Started with a practical question in AI-assisted development: how can agents run integration tests without reading .env or receiving production-like secrets? Instead of only changing how credentials are passed around, mockport emulates dependencies locally so tests can switch endpoints and environment variables.

• roomci — Experimental Rust MQTT

  Grew from contributing to floci and building mockport. It explores whether local emulation and contract testing can reproduce IoT and smart-home failures in CI instead of waiting for them to happen in the field.

• qzt — In progress Rust zstd

  qzt is a queryable zstd text-evidence container with partial retrieval and verified search. I am exploring it as a storage foundation for MemoryPager, a higher-level system for preserving and recovering long-running human–AI context.

• floci — Local AWS RDS persistence and runtime restoration

  Merged: #945, #1014, #1071

• X / @forte_grapher
• Note / albert_forte