Each project housed within Belmont Labs is classified as an ongoing experiment — a living, breathing, occasionally sparking thing. Observe them. Learn from them. Do not, under any circumstances, feed them after midnight or expose them to the wrong kind of electromagnetic field. The Lab makes no warranty, express or implied, regarding temporal side effects.
WhoDAT is the engine. The foundation. The silent recording apparatus humming beneath your game session, watching everything you do and writing it down with the dispassionate efficiency of a very thorough research assistant who never sleeps. Experiment 01 is the data backbone of the Belmont Labs ecosystem — vast, interconnected, and accumulating evidence at a rate that should probably concern you more than it does.
The core function is deceptively straightforward: track everything your character does, store it in a structured database, and export it cleanly at logout. Money earned and spent. Gear equipped and lost. Combat performance, per-encounter. Quests completed. Deaths suffered, and the precise circumstances thereof. Auction transactions. Guild bank activity. Raid lockouts. WhoDAT records it all with the calm authority of a system that has seen things you couldn't dream of and has a timestamp for every one of them.
Maintenance logs suggest the system has developed what researchers tentatively describe as \"opinions\" about certain data formats. These opinions have, so far, proven correct. We are documenting this.
Experiment 02 — henceforth designated WhoDASH — represents the laboratory's first foray into the realm of high-velocity identity resolution and dashboard presentation. It moves quickly. That is, in fact, the point.
The apparatus was designed to answer a deceptively simple question: who is responsible for what? It then renders that answer with the speed and visual clarity that the modern operational environment demands. No dithering. No fog. Just data, surfaced at the moment you need it, in a format your tired eyes can actually parse.
Initial testing revealed a curious phenomenon: operators who had previously spent considerable time hunting for information reported a sudden, unnerving sense of competence. Follow-up studies are ongoing. Side effects appear to be permanent.
Every engine requires a sync bridge. Experiment 03 — SyncDAT — is that bridge. The two-way conduit through which character data travels up to the dashboard and synchronized addon files travel back down again. It does not ask permission. It does not require reminders. It runs quietly in your system tray and handles both directions without complaint.
Configure it once: set your WoW installation path, enter your API key, and step away. It uploads, it downloads, it backs up, it watches — with the thorough, methodical patience of a researcher who has been told, many times, that data does not move itself.
Early prototype testing involved an incident with an improperly formatted CSV that shall not be discussed in detail. The resulting system improvements were considerable. The resulting facial expression of the lead researcher was, according to contemporaneous accounts, "transcendent."
If the other experiments are the what, the Build Utility is the how. Experiment 04 is the foundry. The forge. The crackling, humming, occasionally ominous machine room from which all other Belmont Labs experiments emerge, blinking, into the world.
The Belmont Labs Build Utility standardizes, automates, and coordinates the construction of the lab's software inventory. It is the thing that builds the things. This sounds straightforward. It is not straightforward. Anyone who has stood at the blast shield and watched a multi-stage build pipeline run from scratch will tell you: there are moments. There are always moments.
The utility currently governs all packaging, compilation, and deployment workflows for the laboratory's active experiments. It knows what everything needs before they know they need it. This is either elegant foresight or something slightly more unsettling. The team is divided on which.
Experiment 05 began as a perfectly reasonable grievance with the default WoW chat interface and escalated, as all reasonable grievances eventually do, into a complete replacement. WhoCHAT is a custom multi-panel chat system for WoW 3.3.5a that consolidates Whispers, Guild, Officer, Battleground, Party, Raid, Nearby, and dynamic channel tabs into a unified, tab-navigated frame.
The Glance dashboard surfaces your most recent mentions, incoming whispers, and raid warnings in one consolidated view. Dynamic channel detection registers your active channels automatically and creates navigation tabs without requiring manual configuration. The whole system can optionally intercept the Enter key and suppress the default chat frames entirely. It is, on paper, a very good idea.
In practice, WhoCHAT is functional and genuinely useful — up to a point. That point is, regrettably, somewhere between sixty and ninety minutes of continuous gameplay, at which time a persistent memory leak makes its presence aggressively known. A memory manager was developed and deployed. It helps. It does not solve the problem. The source of the leak remains under active investigation. The lab has opinions about where it lives. The lab has been wrong before.
/wcstats to monitor memory. Use /wcclean to
mitigate. Use /reload to reset. Do not use it in a progression raid without a plan."
Experiment 06 — the Belmont Labs Interactive Storage System — began as a perfectly reasonable desire to browse files through a browser and escalated, with the velocity the lab has come to accept as its natural operating mode, into a full windowed desktop environment running inside a Docker container. The lab is aware. The lab has made peace with this. And then the lab kept going.
B.L.I.S.S. now presents as a desktop-style file manager in your browser: draggable and resizable windows, snap-assist layout management, a taskbar with buttons for every open window, three view modes, a sidebar with pinnable locations, context menus, keyboard shortcuts, bulk operations, ZIP support, file preview, wallpaper, two themes — and a fully integrated text editor with live syntax highlighting for over 30 languages. All configuration persists server-side. Your files never leave your machine.
The editor opens as a proper window: draggable, resizable, snap-assist compatible, taskbar-buttoned, and correctly layered in the z-index stack. It highlights code as you type using a contenteditable approach that eliminates cursor drift entirely. Deployment remains one command. The experiment is, structurally, as low-friction as an infrastructure apparatus can reasonably be. Functionally, it is now something considerably more.
Certain individuals in Azeroth operate with a pattern: they find you, they kill you, they move on — confident you won't remember them. Experiment 07 exists because that confidence is misplaced. The Grudge maintains a persistent, cross-character kill list and watches for those names across seven detection layers simultaneously. When a match surfaces, an alert fires. Colour, sound, and intensity scale with kill count. The operator is informed. What happens next is outside the experiment's scope.
The visual popup pins to screen while the cursor rests on the target. The settings window offers per-severity sound previews, custom MP3 slots, and a one-click grudge list broadcast to party, raid, guild, or anyone in earshot. The list does not forget. Neither do you.
Experiment 08 was initiated from a garage containing two projectors and a wall. The lab wished to watch YouTube on that wall — specifically, spanned across both projectors as a single continuous image — and discovered that no existing tool addressed this configuration without requiring a degree of manual intervention the lab found professionally unsatisfying. S.P.A.N. is the corrective apparatus. It is a single Python script that resolves YouTube streams through yt-dlp and fires mpv with a geometry argument that covers exactly the display area you specify.
Stream URLs are pre-resolved using yt-dlp twice — video and audio separately — and passed as bare CDN URLs to mpv. This sidesteps the HTTP 403 errors that result from mpv's internal ytdl hook probing DASH fragment manifests. The lab discovered this through observation. The observation was not pleasant but the fix was clean. Negative X and Y display offsets are supported for non-standard monitor configurations. The lab has a non-standard monitor configuration.
Experiment 09 began as a support tool and, as many things in this lab, promptly exceeded its original remit. SaveState is a self-hosted contact and support intelligence system — a full-stack web application for logging tickets, tracking contacts, managing known issues, and building a searchable vault of the people and problems your team encounters repeatedly. It runs in Docker. It persists to MySQL. It has opinions about data, and the data has, largely, accepted this.
The interface runs in three modes: Laboratory for methodical data entry, Business for professional deployment, and Arcade for operators who require gamified XP feedback and a level system to remain engaged with their ticketing obligations. The lab acknowledges this is unusual. The lab does not consider it a problem. Multiple visual themes are provided. The glassmorphic aesthetic is non-negotiable.
A first-run setup experience guides the server administrator through network mode selection. Local Network mode allows one-click profile login — frictionless, no passwords. Public Network mode enforces full password authentication on every account. This choice is written to the database. It does not ask twice.
Experiment 10 began as an act of self-preservation. The lab needed peace. The lab built peace. C.A.R.I.L.L.O.N. is a single-file interactive wind chime visualizer — seven chimes on full 3D pendulum physics, real-time procedural audio synthesis with ten world-matched synthesis architectures, and ten fully realized visual environments drawn entirely in code. No images. No server. No accounts. It runs offline. It remembers your preferences. It asks nothing of you in return.
Experiment 11 began as an act of protest. The Lab wanted a way to schedule micro-workouts throughout the day. The good applications that could accomplish this were, without exception, barricaded behind subscription paywalls that the research staff found personally offensive. We free the 1's and 0's here. This has been policy since before it was written down.
FitNES is a free Android fitness tracker built around a full RPG progression system. Log sets, earn XP, level up, unlock WoW-style Feats of Strength, and monitor your adherence with real progress charts. The gamification layer was initially proposed as a joke. Internal testing confirmed it was more effective than the non-joke version. The chemicals may have accelerated this conclusion.
The audio engine deserves special mention: every boop, ping, and level-up fanfare is synthesized in real time — no MP3 assets, no bloated audio bundles. Pure, procedurally-generated NES-style tones, because the Lab believes the medium and the message should be thematically aligned.
Experiment 12 emerged from a straightforward grievance: traditional speed tests measure bandwidth, but they do not measure suitability for streaming. A connection testing at 100 Mbps can still experience buffering, dropouts, and failures due to jitter, latency variance, CDN accessibility issues, or packet loss under sustained load. The lab found this gap unacceptable.
ekoh — Externally Known Obstruction Heuristics — is a comprehensive client-side diagnostic tool that measures the conditions that actually matter when streaming audio content. Seventeen distinct tests evaluate stability, latency, jitter, throughput consistency, DNS resolution, CDN reachability, and real-time protocol support. The result is a single composite score (0-100) that reflects actual streaming reliability.
Results are stored with human-readable three-word passphrases that expire in 72 hours. The data is shareable. The data is exportable. Support teams can collect standardized diagnostics. Network engineers can benchmark CDN accessibility. End users can verify VPN impact before filing a complaint. The apparatus does not guess. It measures.