AI-Assisted Engineering

Bringing agentic systems and AI-assisted development into the operational loop. Pragmatically, in public, with verifiable code.

We build platforms. But we also build the methodology that makes AI a reliable partner in that process. The tools, the discipline, the traceability. Not buzzwords. Verified experiments with real code, real tests, and real outcomes — all published openly on onlyascii.dev.

What we do

Autonomous SRE watching. The SRE Watchtower — a four-stage loop: observe, issue, investigate, report. Pipeline fails, agent opens a ticket, fetches logs, pattern-matches the error, writes a structured investigation. No human pages until there’s something to decide. The code is public, the tests pass.

AI-assisted development methodology. TDD with LSP as the guardrail. Per-cycle commits. Trace tags for auditability. A repeatable process that works across Rust, Go, and Python. The weather-cli, mini-clap, and task-tracker repos are the proof.

Methodology creation. We do not just use AI tools. We design the process by which AI builds reliable systems. The discipline is the product. The Philosophical Developer series is the running experiment log.

The Philosophical Developer

The methodology is developed openly as a series of real experiments. Each chapter is a real project with real code, real tests, and full traceability. The repos are on GitHub. The reasoning is in the tags. The conclusions are on the blog.

Explore the full series at onlyascii.dev.

Want a deeper look?

The source code, trace tags, and full experiment logs are public on GitHub. Start with the projects that validate the methodology.

• weather-cli
• mini-clap
• task-tracker
• afero experiments
• tqdm experiments
• sre-watchtower