Case study / 2025 — now
FiveMesh
Infrastructure for FiveM communities. One interface, one API, one ecosystem — without asking server owners to become infrastructure specialists first.
What it is
FiveMesh started as a much smaller idea. It did not remain a small idea.
It’s now a platform of focused services for FiveM communities: media delivery, asset caching, managed voice, log ingestion and an account-aware assistant. The common thread is simple: keep the complicated infrastructure underneath, not in the customer’s way.
The problem
FiveM communities need the same unglamorous things as every online product: fast delivery, reliable servers, observability, sensible billing and tooling that doesn’t create a second job. The difference is that the people who need them are often busy running a community, not an infrastructure department.
FiveMesh is my attempt to meet them in the middle: powerful primitives, presented as products with clear defaults and fewer knobs.
The product
CDN and Cache
FiveMesh CDN delivers media at the edge, with HTTP Range support, ETags and caching logic built around Cloudflare Workers, R2, KV and Analytics Engine. Cache takes a similar approach for FiveM assets: origin retrieval, hashed keys, edge caching and usage tracking.
Both are increasingly measured through unified concepts like global requests and bandwidth. It’s more useful than exposing every internal implementation detail just because it exists.
Voice
FiveMesh Voice manages voice infrastructure compatible with the FiveM/pma-voice ecosystem. It has operated nodes in Europe and the United States, with automated provisioning, telemetry and health monitoring around TypeScript, Docker and rust-mumble.
Logs
FiveMesh Logs ingests and queries server action logs with Pipelines, R2, R2 SQL, Data Catalog and ClickHouse-powered querying. Authentication and ingestion Workers sit at the edge of the system.
Load-test results — not production traffic
One 100k-event test reached roughly 2.3k events/sec. A 1 million-event test completed with zero failed requests.
Agent
FiveMesh Agent is intentionally not a generic chatbot pasted into a dashboard. It can inspect a user’s services and servers, understand configuration, retrieve official documentation and help diagnose a problem in context. The initial version is read-only.
Selected decisions
A large portion of FiveMesh is moving toward Cloudflare-native, serverless infrastructure: Workers, R2, KV, D1, Analytics Engine, Pipelines, Workflows, Durable Objects and Workers AI. The goal is not to collect services. It’s to use direct primitives where they remove an unnecessary intermediary.
For future Agent write actions, authorization stays server-side. Actions need explicit confirmation, short-lived single-use identifiers and auditability. Secrets do not enter model context. Being helpful should not be an excuse to become careless.
Beyond the infrastructure
FiveMesh has also meant pricing, metering, subscriptions, onboarding, localization, accessibility, documentation, dashboard UX, lifecycle automation and service provisioning. I rebuilt parts of the web presence around server-rendered Workers, internationalized routes and responsive, accessible interfaces.
That product work matters as much as the edge code. A fast system is only useful when people can understand what it does.
What I’m learning
Simple products often sit on complex systems. The work is deciding which complexity deserves to be visible, and which complexity should become a good default. I’m still learning that every day — preferably before 3 AM.
TypeScript / Cloudflare Workers / R2 / KV / D1 / Analytics Engine / Pipelines / Workflows / Durable Objects / Workers AI / Docker / rust-mumble