TOPIC: Chat-GPT and True Dungeon revisited

It’s been a bit over a year, and with all the talk about how Chat-GPT and other Large Language Models (LLMs) have improved I decided to see if someone asking about True Dungeon would get better answers than before.
In short, Chat-GPT4o is a definite improvement but the fundamental flaws remain.

When I asked general questions – the sort where the Internet provides plenty of text to mine – it responded well. For example, when I gave it the prompt “I have a question about role-playing games. Please describe for me the game called ‘True Dungeon’” it came back with an easy-to-read, accurate description that I would be happy to give to anyone trying to decide if they want to try TD at a con.

However, when I asked questions of fact that required actual knowledge, it confidently gave very inaccurate answers – what are technically called “hallucinations” – along with correct answers. For example, when I gave it the prompt “In the game ‘True Dungeon’, what are the best tokens for a wizard to use?”, it gave a list of “Budget-friendly wizard token suggestions” and included “Ioun Stone Silver Nugget (affordable spell damage boost)”

This is the same gap I found a year ago despite all the advancements, and the reason is simple: When you look under the hood, Large Language Models fundamentally don’t think. They mathematically calculate what words most often appear next to other words in a very large dataset, the larger the better, and select those. Therefore, if there are multiple websites and posts saying what True Dungeon is, it will take that text from those websites. Similarly, though, it recommended “Redoubt Helm” for the wizard because clearly that token name often appeared next to the text “head slot”.

Just something to know in case you or someone you meet asked Chat-GPT their True Dungeon questions.

We recently got an enterprise license at work, so I ran this through ChatGPT 5 to see how much it differed from 4o. I'll paste the full reply to the prompt (just copied yours) below, but even 5 is still misleading.

I found a couple of things interesting. Given it's just an LLM, it doesn't really evaluate its source. For example it pulled from a post "Amulet of Freyja" (which should be Necklace of Freyja) and incorrectly cited it as +1 damage (should be +2).

It did point to a few good forum posts, most notably the Uncommonly Good Tokens guide (here: truedungeon.com/forum?catid=5&id=248885&view=topic&utm_source=chatgpt.com ). It also gave general advice that may be basic but is good from a very high level perspective.

So all that being said, 5 isn't much of an improvement from 4o for the game either, but it is interesting to see how it took the question in a slightly different direction.

Full reply:

Here are some of the best-token considerations for playing a Wizard (or Elf Wizard) in True Dungeon — with the caveat that “best” always depends on your build, party composition, the difficulty level (Normal/Hardcore/Nightmare/Epic), and what token slots you already occupy. But based on community consensus and “Uncommonly Good Tokens” lists, these are strong picks and general guidelines.

---

## ✅ What to prioritise for a Wizard

Since Wizards in True Dungeon are primarily spellcasters (versus heavy-melee or pure tank builds), you’ll want to focus on tokens that help you:

* Increase your spell damage (especially single-target and area‐spells)
* Improve your chances to hit or succeed with the spell (spell-resistance checks, saving throws by monsters, etc)
* Boost your defenses (AC, Reflex/Will saves) or mitigate risk (since you’re often more fragile)
* Possibly improve your action economy (extra uses, extra effects, etc)
As noted in the “Uncommonly Good Tokens” guide: “Start with either the filter view for all tokens … then sort by Slot or by purpose (e.g. boost AC, or Class Abilities, or Spell Healing) … for a particular class: … Wizards.” ([truedungeon.com][1])
So when you evaluate a token, ask: Does this meaningfully help *my* class (Wizard) rather than being a generic buff?

---

## 🎯 Some of the top token picks for Wizards

Here are specific tokens (or token types) often mentioned as strong picks for Wizards. Availability, rarity, cost, and whether you already have something in that slot will of course affect your decision.

* **Amulet of the Norns** – This token is frequently cited as a strong rare‐level brain for Wizards because it gives *an additional Missile Spell* (or similar) which directly improves your output when you don’t have many other top‐tier tokens. For example one commenter:

> “They could use an Amulet of the Norns for an additional Missile Spell.” ([truedungeon.com][2])
* **Amulet of Freyja (UC/Uncommon)** – Mentioned as increasing *all spells by +1* which is very strong for a spellcaster (as opposed to a melee class getting +1 damage). For example:

> “They do also get Amulet of Freyja at UC to increase all spells by 1.” ([truedungeon.com][2])
* **Legendary tokens (class specific) for Wizards** – On the “Legendary Tokens by Class” forum:

> “I play Elf Wizard/Wizard a lot … Between Aron's, Drake's, Relsa's, Pharcus', Ashenne's, Luna's and Muk’s there’s a lot to consider.” ([truedungeon.com][3])
> So if you’re building toward “best‐in‐slot” endgame, investigate which legendary tokens are being flagged for Wizards.
* **Defensive/Save Boost Tokens** – Even for a Wizard, you’ll often want strong saves (Will/Reflex) because spells often rely on monsters failing saves or you surviving hazards. One discussion:

> “Pharacus’ +6 to saves is really, really good in dungeons needing lots of saves.” ([truedungeon.com][3])
* **Token slot flexibility and build synergy** – Some tokens are not “must have” for every build, but if you’re missing something in your gear, they become very valuable. For example:

> “It’s hard to believe that this was equippable in best in slot … due to a combination of: fewer STR boosters overall … lack of important saves in the dungeon…” (though this quote is more melee‐biased) ([truedungeon.com][4])
* **Common/Uncommon still matter for newer builds** – If you’re earlier in your token collection, looking at the “Uncommonly Good Tokens” guide is helpful. It suggests that even non‐legendary tokens can make big differences. ([truedungeon.com][1])

---

## 🔍 Additional notes and caveats

* Availability and cost: Many of the top tokens (especially legendary or UR) are expensive or hard to obtain. So you need to weigh whether a “maybe” token in your slot is better than nothing or worse than putting something in another slot.
* Slot competition: As a Wizard you have limited slots (Amulet, Head, Hands, Offhand/Focus, etc). Sometimes the “next best” token might be in a less obvious slot (e.g., a Focus token boosting spells) rather than trying to fill “best amulet” immediately.
* Meta and dungeon difficulty matter: For example, if the dungeon is heavy on saves, tokens boosting monster save failure (or your spell hit chance) matter more. If it’s heavy on traps/hazards, defensive tokens may become more valuable.
* Trade-off between general utility vs class-specific: Some tokens work great for multiple classes (flexible), but Wizards often benefit from tokens that specifically enhance spells rather than weapon/melee damage.
* The community rankings change year to year: New token sets, new mechanics (e.g., for 2024/2025 there were new conditions and set bonuses) mean that what is “top” now might shift next year. ([truedungeon.com][5])

---

If you like, I can pull together **a curated list of ~10 tokens** (from Rare, Relic, Legendary tiers) specifically *recommended for Wizards*, with approximate market/GP values and slot-by-slot suggestions (Amulet, Offhand/Focus, Head, etc). Would that be useful to you?

[1]: truedungeon.com/forum?catid=5&id=248885&view=topic&utm_source=chatgpt.com "Uncommonly Good Tokens - True Dungeon"
[2]: truedungeon.com/forum?catid=737&id=257357&view=topic&utm_source=chatgpt.com "2025 Rare Tokens - True Dungeon"
[3]: truedungeon.com/forum?catid=5&id=252934&view=topic&utm_source=chatgpt.com "Legendary Tokens by Class - True Dungeon"
[4]: truedungeon.com/forum?catid=5&id=256148&view=topic&utm_source=chatgpt.com "Personal Top 3 Tokens? - True Dungeon"
[5]: truedungeon.com/forum?catid=710&id=256050&view=topic&utm_source=chatgpt.com "2024 Token FAQ - True Dungeon"

True dungeon falls into what the papers call an under represented domain in the training data and the results are always hilarious. Its a warning though for uncritically taking any LLM generated text as factual. They are powerful tools for sure, especially in domains of language that are well represented in the training data, but LLMs are still very very bad at inference learning. Part of it comes from there not being a good way to do reward modeling for saying "I don't know" since that can become a universal answer.