I am moving deliberately into a Rust-first stack: Rust for the systems that need speed, correctness, and explicit ownership; PostgreSQL for durable state; Python for automation, prototyping, data work, bots, and glue. Shell and PowerShell still matter because real IT work needs practical tools that run where the machines are.
The domains pulling me hardest are high-frequency trading infrastructure, crypto markets, cryptography, encryption, secure networking, local-first software, and private web-native products. LoveWard is the current primary product focus: a self-hostable practice PWA for Meditate, Breath, Yoga, Love, grounding, and private practice history.
The work I like best has a clear owner, a clear data model, and an operating story that can be understood without a vendor diagram. A small program on infrastructure you control is still one of the fastest ways to turn an idea into something durable.
What I care about
Performance with receipts. Fast code should come with benchmarks, traces, and a clear reason for every optimization.
Security as architecture. Cryptography, encryption, key handling, identity, retention, logging, and trust boundaries shape the design from the start.
Ownership. I like software that can run on hardware you control, with data you can inspect, back up, restore, and move when you need to.
Explicit data. Schemas, migrations, records, validation, and SQL should make the system easier to reason about, not harder. If you cannot trace a value through the system, the system is too clever.
Operational discipline. I value iteration, but not chaos. The goal is useful software with complexity that has earned its place.
The stack
Rust is the center: CLIs, network services, protocol work, secure transport, local tools, trading experiments, and performance-sensitive paths. The defaults are small crates, explicit APIs, real error types, structured logs, benchmarks before claims, and the normal Cargo quality gate. For web apps and self-hosted dashboards, the default lane is Leptos, Axum, Tokio, SQLx, and PostgreSQL.
PostgreSQL is the primary data layer. Relational state, JSONB documents when they fit, search, audit trails, job tables, reporting, and operational history live in one inspectable system before another moving part earns its place.
Interfaces should match the job. Leptos is the default for Rust web surfaces, Ratatui for terminal applications, clap for serious CLIs, and Tauri, egui, iced, or Dioxus only when the product genuinely needs desktop-native or cross-platform app UI.
Product web apps use the boring mature web stack when that is the right shape: Next.js App Router, React, TypeScript, Tailwind CSS, shadcn/ui, Radix, Better Auth, Drizzle, PostgreSQL, pgvector, Dexie, Stripe, Docker Compose, Caddy, OpenAI provider abstractions, Vitest, Testing Library, and Playwright.
Python is for leverage: automation, prototypes, data
scripts, bots, operational glue, and one-off tooling. New Python work
uses uv, project-local virtual environments, Ruff,
pytest, and a checked-in pyproject.toml.
Operations stay practical: macOS for local work, Ubuntu LTS for servers, Caddy for TLS, systemd for long-running processes, SSH deploys, PostgreSQL backups, restore drills, Bash/zsh on Unix, and PowerShell where Windows administration is the right surface.