FAQ

Each idle gap that exceeds the earned context window closes the billing cycle. A new cycle means a new first hour — and the first hour absorbs the full context reload: your topology, current state, constraints. Sprinting amortizes that cost across more hours.

No. The formula is deterministic — same inputs always produce the same price. This protects both sides: you know exactly what you will pay before the engagement starts, and I do not guess what the market will bear. The only discount is −15% for USDT payment.

Productive hours only, at 0.5h granularity. Waiting for your response, reading documentation, or troubleshooting my own tools — not billed. The first 30 minutes of any engagement are free.

USDT payments receive a 15% discount on the total. Lower transaction costs and faster settlement. Applied after all other calculations (multiplier, formula).

100% remote. You handle the physical layer (cable, power, rack), I handle the logical layer (config, routing, security, monitoring). This model works worldwide — I connect via screen sharing and remote access.

10:00–20:00 GMT+7 (Vietnam). Emergency tier is available outside these hours at ×2.0 rate. 1–2 days advance booking appreciated.

100% post-delivery. No deposits, no milestones. You see the completed work before paying. Billing is per cycle — hours accumulate within a billing cycle, and the cycle closes when the engagement is complete or idle time exceeds the earned context window.

Yes — with grounding infrastructure. AI generates RouterOS configurations from verified facts (your topology, firmware version, hardware model), not from hallucinated documentation. I authored routeros-7-llm-safe-reference specifically to prevent AI errors in network config.

Yes. Ungrounded AI generates plausible-looking but conflicting rules — I see this regularly. I audit the full configuration, identify contradictions, rebuild from scratch, and teach you how to use AI tools safely going forward.

A curated, version-specific dataset of valid RouterOS commands, syntax, and constraints. When an LLM generates configuration against this grounding infrastructure, it produces commands that actually exist in your firmware version. Without grounding, the model emits syntactically plausible commands absent from your firmware.