My AI Learning Journey – Part 10 – AI Assisted Coding – VSCodium and Continue

Key Takeaways

• VSCodium + Continue enables LLM coding without telemetry
• GPT‑5.3‑codex via OpenRouter delivered full‑stack Python script
• Project finished in 3 hours, costing ~€4 (≈$4.30) tokens
• LLM refactored 600‑line script, adding TLS and fingerprint checks
• Privacy‑first toolchain avoids cloud‑based profiling while using frontier models

Pulse Analysis

The rise of large language models (LLMs) has transformed how developers write code, but most mainstream solutions embed telemetry that can expose proprietary logic. Open‑source editors like VSCodium, which strips Microsoft’s data‑collection layers, paired with privacy‑focused extensions such as Continue, offer a compelling alternative. By routing requests through OpenRouter, users gain access to cutting‑edge models—including GPT‑5.3‑codex—while retaining anonymity and control over token billing. This architecture bridges the gap between enterprise‑grade AI capabilities and the open‑source ethos that many developers demand.

In a practical demonstration, the author built a Fritzbox monitoring utility entirely through conversational prompts. The LLM supplied a functional Python skeleton, suggested a suitable third‑party library, and iteratively refined the code to incorporate configuration files, graceful shutdown logic, and TLS fingerprint verification. Over three hours, the script grew to roughly 600 lines, yet required only about €4 (≈$4.30) in token costs. Compared with a manual approach that would have taken days, the AI‑assisted workflow delivered a high‑quality, well‑documented solution at a fraction of the time and expense.

Beyond the single project, this workflow signals a broader shift in software development. Developers can now embed powerful LLMs directly into their IDEs without surrendering code privacy, enabling rapid prototyping, automated refactoring, and security hardening on demand. As open‑source registries expand and more models become accessible via platforms like OpenRouter, the barrier to adopting AI‑augmented development will continue to fall, reshaping productivity expectations across startups and large enterprises alike.