Project assessment and implementation status

Kartonath Project Review

How Kartonath can deliver the dangerous, social feel of classic EverQuest while building something Project 1999 cannot: a modern native client and a carefully governed population of clearly identified AI adventurers.

Implemented September 4, 2026 · Server, native client, AI adventurers, and classic-content recovery

The short answer

Kartonath should not try to out-P99 Project 1999 as a stricter museum. Its strongest position is: “A dangerous, social Classic–Velious EverQuest world with a modern native client and clearly identified AI adventurers that keep the world alive without replacing humans.”

Jump to a section

1. Where Kartonath fits 2. What exists now 3. What must match P99 4. Where Kartonath can win 5. AI adventurer rules 6. Player and GM clients 7. Current gaps 8. Recommended order 9. Sources

1. Where Kartonath fits

Project 1999 has operated since 2009 and has a deep classic-research culture, a large established community, and written permission from Daybreak. It targets the original game through Scars of Velious. Trying to beat it only on “more classic” is a very difficult and unhelpful fight.

EverQuest Legends launched July 28, 2026 with a solo/casual and multiclass direction. That opens a clearer lane for Kartonath: remain social, dangerous, slow enough to make cooperation meaningful, and technically accessible without making the game convenient.

P99's durable advantage

Trust earned through longevity, era research, community memory, recognizable rules, and a known progression target.

Kartonath's opening

A fresh world, native rendering, simpler onboarding, visible content certification, and AI residents that support social play during low population hours.

2. What I found

The native EverQuest client exists

The custom client is not merely an idea. The working tree at /home/tonberry/eq-client contains a standalone Rust/wgpu client of roughly 16,700 lines. It already includes a protocol stack, rendering, UI, login and character flows, world interaction, audio, collision work, and updater code. Its expanded suite now passes 52 tests.

The login presentation is polished. The saved in-world view still looks like an engineering prototype, which is normal for this phase but important when setting public expectations.

Two-platform release path now working: tested Windows and Linux x86-64 bundles carry the same complete asset set. Both use separate HTTPS update channels, verify SHA-256 and file size before an atomic swap, and reject cross-platform executables. Linux includes native ALSA audio and preserves executable permissions during updates; the final package passed a real Vulkan/X11 startup test.

The AI adventurer project exists too

The /home/tonberry/kartonath-bot project has substantial recent work covering movement updates, memory, world interaction, and other resident behavior. It is a real differentiator—not marketing vapor.

The collision tuple mismatch is repaired. The bot now passes 87 tests, builds in release mode, and six residents have completed real login, world, and zone handshakes. Mentor, Rurik, Brak, Saga, Korin, and Yorvik are visibly surnamed AI Adventurer; a persistent watchdog keeps the small cohort alive after a process failure or reboot. EQEmu's separate player-commanded stock-bot subsystem is disabled and remains disabled across world restarts.

3. What must match the Project 1999 classic feel

These are not optional flavor details. Together they create classic EverQuest's social pressure and sense of place:

Classic → Kunark → Velious progression with explicit era boundaries.
Era-correct NPCs, spawn tables, loot, quests, factions, spells, boats, zone geometry, item changes, and patch timing.
One human, one client; no multibox advantage, real-money trading, or paid power.
Meaningful travel, downtime, corpse recovery, open-world danger, and class interdependence.
No quest GPS, entity radar, automatic navigation, fast travel, or invisible convenience systems that reveal knowledge players did not have.
A fresh economy protected from bot farming, dupes, excessive faucets, and staff favoritism.

Modern graphics, UI scaling, patching, security, and accessibility can improve. The friction that causes players to learn the world and rely on each other should remain.

4. Where Kartonath can be better

A reliable native client: true high-resolution output, readable scaling, modern GPU support, stable windowing, and fewer legacy compatibility problems.
Easy installation: a small launcher, signed releases, HTTPS manifests, incremental patches, rollback, and file verification instead of a giant ZIP.
A fresh world: no decade-old economy or permanently entrenched progression.
Evidence instead of promises: publish an era manifest, a deviation ledger, and automated zone-certification results.
Clearly marked AI residents: adventurers that converse, group, travel, and make quiet zones feel inhabited while remaining subordinate to human play.
Fairer scarcity: preserve competition and rare rewards while designing rules that reduce permanent raid and camp monopolization.

The key design distinction: modernize access to the world, not the challenge inside the world.

5. Rules for AI adventurers

AI residents can become Kartonath's signature feature only if players trust them. The rules should be public, enforced in code, and visible in logs.

They are always clearly identified as AI and never masquerade as humans.
They yield contested camps and do not monopolize spawns.
They do not farm named loot, raid targets, rare materials, or currency for the economy.
They do not compete for server-firsts or replace human guild progression.
They cannot receive privileged spawn, map, loot, or player information.
Their purpose is social atmosphere, emergency group support, travel, conversation, and teaching—not optimization.

A useful default is “human-first scarcity”: if an AI and a player want the same limited opportunity, the AI disengages.

Alpha state: the live cohort is labeled, Mentor is under a human-first WAIT command, and the launch soak recorded no attack or loot actions.

6. Split the player and GM clients

Player client

No radar, entity map, free camera, arbitrary-character viewpoints, hidden spawn information, or staff observation tools. It should expose only what a legitimate character can perceive.

GM observer client

Keep operational powers in a separately built client protected by server-side authorization, auditable staff accounts, and an explicit privilege boundary.

Hiding GM controls in a normal client UI is not a sufficient security boundary. The server must refuse unauthorized actions.

7. Current gaps to close

Critical recovery completed: the production login, world, chat, database, and zone services are online. Every active Classic–Kunark zone has a version-0 spawn population, and the global active-era audit reports zero dangling NPC, loot-table, loot-drop, or item references. Nektulos now has 447 valid spawn points, 112 groups, 109 referenced NPC types, 2 merchants, 9 zone points, and 25 doors. The automated certification runs daily, and the public status is generated from live processes instead of a stale snapshot.

Extend the live Classic–Kunark spawn-and-loot certification into deeper checks for merchants, bankers, doors, zone lines, factions, quests, and later Velious activation.
Recheck and repair the earlier banker findings, including West Freeport and the intended open-field service coverage.
Clean and commit the heavily modified AI bot working tree, then run both Rust suites in continuous integration.
Extend the updater from executable-only replacement to signed, incremental asset manifests; make the game endpoint release-configurable.
Complete the separate player/GM build boundary and enforce every staff capability on the server.
Build discoverability and population trust through published certification results and small, measured alpha sessions.

The public server-list snapshot at review time showed Project 1999 Green at 210 players, Blue at 79, Red at 12, and Quarm at 112. Those figures change constantly; the meaningful point is that discoverability and population trust must be built deliberately.

8. Recommended order of work

Keep the recovered service observable. Preserve live status, tested backups, and restartable game/AI services as the baseline.
Finish protecting the work. The native client is committed; clean the bot repository and make both passing test suites continuous.
Deepen world certification. The daily Classic–Kunark spawn-and-loot gate is live; extend it to merchants, bankers, doors, zone lines, factions, quests, and later Velious activation.
Enforce the client boundary. Produce separate player and GM builds with server-side permissions.
Finish the trustworthy launcher. HTTPS integrity checking and rollback are live; add signed manifests and incremental content-addressed asset patches.
Run a focused closed alpha. The small labeled AI cohort is online; certify the Qeynos → Blackburrow → Karanas level 1–20 journey, then invite classic veterans.

That alpha should measure grouping, travel, recovery, economy, AI camp yielding, crash rate, and whether veterans say it feels like classic EverQuest—not merely whether quests complete.

9. Public sources

Repository and database observations are point-in-time findings from the September 4, 2026 review unless another date is stated.

↑ BACK TO TOP