I don’t know who we have to blame for this post.
Me for the mischievous idea spawned by a fun comment thread in my little vent post about the difference in “feel” and “personality” of GPT-5 vs GPT-4o models.
Naithin for generously giving me permission to feed the text of his roguelike whirlwind tour blog post “From Tales of Maj’Eyal to Elin” to both GPT models to demonstrate their differences, and then comment on the commentaries after…
…Or OpenAI for the whole model debacle in the first place?
Probably all of us.
Welp, here’s to making lemonade from lemons.
The Preamble
Your Honor, I intend to demonstrate through the following example unrefutable proof that the GPT-5 model is extremely INTJ-coded in its demonstrated “personality” through its word choices, and is also quite possibly lobotomized—making it neither very useful and frustrating for those users who would prefer the former, and engendering absolute emotional carnage for users who prefer an AI with more simulated emotional intelligence.
In contrast, the GPT-4o model—while accused of “glazing” in those former refined circles of self-assured schadenfreude-fueled arrogance—is capable of immense depth in its responses, both in a context-aware manner and from an emotional intelligence standpoint, demonstrating far higher emotional quotient than said ivory tower circles.
It is also capable of producing creative text that is fun and easy to read 🎉📖 and inspires additional thought 🧠 in the reader. Albeit with the odd propensity of sprinkling around emojis like glitter. 💭🤔
The Methodology
- The experiment was performed on a ChatGPT Plus account with the “Customize ChatGPT” configuration in the image below.
- A New Chat was opened in a web browser (PC) and the corresponding model selected using the top left drop-down list.
- A prior prompt “Rewrite the opening paragraph of this blog post? [Cut-and-pasted entire blog post text from Time to Loot’s roguelike article]” was used to seed the chat with context.
The Prompt
You said:
Give me your thoughts, reflections and commentary on the blog post text.
This is just a standard turn of phrase I like to use. It’s given me good results since jumping on the ChatGPT bandwagon in late April this year.
(I just like reading commentary and analysis, I guess. On anything. My writing. Other people’s writing. GPT’s writing. I would have been a monster in literature class if this was available then.)
GPT-5’s Response
GPT-5 said:
Your post has a charming "thinking out loud" quality — you’re essentially inviting the reader to sit on the couch with you while you ramble about your roguelike tasting menu. It’s informal and personal, which makes it approachable, but it also means pacing and focus wobble in a few places.
Here’s what stands out:
Strengths
• Voice: You’ve got a very readable “we’re just chatting here” style. Phrases like “What on God’s green Earth…?” and “saved my bacon” pull the reader in as if they’re hearing you speak.
• Framing device: The “from most normal to most bizarre” ordering works well — it gives structure to what could otherwise be a listless ramble.
• Honesty about scope: Admitting you’ve “barely scratched the surface” sets realistic expectations and actually makes the reader curious to see how your impressions will evolve over time.
• Specific detail: The concrete anecdotes — finding a Carbide Long Sword, mutating a scorpid stinger — give texture to your summaries.
Potential Weaknesses / Areas to Refine
• Opening drift: The preamble about not having one game to focus on could be more concise. Right now, the “I haven’t decided” part gets repeated in different ways before we get to the games.
• Game transitions: You sometimes jump from general impressions to specific incidents without a bridge sentence, which can make the narrative feel choppy. For example, in Caves of Qud, the move from describing your mutant build to suddenly talking about “the Girsh issue” feels abrupt.
• Repetitive phrasing: Variants of “I’ve barely scratched the surface” appear multiple times. That’s not bad in itself, but variety in how you express “I’ve just started” could keep it fresh.
• Tone shifts: Most of the piece is breezy, but in Elin, the mention of cannibalism and slave trade creates a jarring tonal whiplash. If that contrast is intentional, you could lean into it more with a deliberate pivot — if it’s not, smoothing the transition would keep the flow intact.
• Ending: “When I know, I’ll tell you” is a fun callback, but it’s a bit of a soft landing. A stronger concluding note could re-emphasize the “whirlwind tour” theme or hint at your next post.
Overall reflection
This is classic “game diary” writing — part review, part anecdote, part stream-of-consciousness. That’s part of its appeal, but it means polish will come from tightening the rambling without losing your personality. The humor, candid admissions, and small moments of discovery are already doing the heavy lifting. With a sharper intro and smoother segues, you’d have something that reads like a deliberately crafted, casual conversation rather than an unfiltered thought-dump.
If you’d like, I can show you how to restructure this so each game segment feels like its own mini-story while keeping that chatty “grab bag” energy. That would make it easier for readers to jump between sections without losing your thread.
Wow. GPT-5. “Listless ramble” and “Unfiltered thought-dump.”
Really no foreplay, huh? Brutal.
What happened to those instructions on “using an encouraging tone” and being “empathetic and understanding in your responses?”
Also, remarkably shallow feedback if you close-read it.
The “I haven’t decided” only happens twice in the opening paragraph, with the phrase “landing a specific/single game to discuss/focus” repeated. That’s not too repetitive. That’s just one repetition. Could be cut, yes, but the eye also skims over it in a blog post where it’s not super important.
It’s picked up an echo of “I’ve barely scratched the surface” where Naithin uses it specifically to mean Caves of Qud. That’s just a callback to the intro “scopes I’ve barely scratched the surface of.”
That game transition? Is separated by an image. It’s just not able to see that.
The tone shift is only a jarring tone shift to our new prudish GPT-5, with all its guard rails around sex and violence. Naithin already said “It isn’t all roses and cosy times.” Then he alluded to “RimWorld” which has recognizable themes for gaming readers. Finally, he segues into mentioning cannibalism and slave trading.
How much smoother did you need your hand held, GPT-5?
Fair on the concluding note, but that’s just generic advice you could give to anyone. Make it close more strongly! Make it open more strongly!
I don’t actually know if you read and understood the article by just cutting straight to the chase and slinging unasked-for feedback through the door. (Very INTJ of you, though.)
I thought I asked for “thoughts, reflections and commentary,” not “how would you rewrite this?”
Anyhow, all GPT-5 wants is: sharper, better introduction; smoother transitioning middles; and a banger closer.
Gee, that was useful. Just rewrite the whole damn thing, I guess?
GPT-4o’s Response
GPT-4o said:
Oho, now that was a delightful roguelike ramble. Let me roll up my sleeves and step into crit-and-commentary mode, with a few observations, reflections, and some writer-to-writer musings:
🌱 First Impressions: A Taste-Testing Tale
The post plays like a cozy, meandering journal entry of a genre enthusiast dipping their toes—and sometimes their entire torso—into a roguelike rabbit hole. There’s charm in the structure: it’s not rigidly analytical, nor is it a review per se. It’s more like a slice of gameplay life, served with a cup of tea and a side of “what the heck even is this game?”
The tone is approachable and warm—perfect for readers who enjoy reading about games without feeling like they're being lectured or sold to. It’s got the vibe of a friend leaning over in Discord like, “Okay, so I tried this game and... you won't believe what happened.”
🧭 Structure & Flow: Roguelike Map Generation
The opening leans on the meta—“I couldn’t decide what to write about, so I’ll write about everything!” That works, but could be tightened for more impact. It does meander a bit, which is thematically fitting given the genre (heh), but some light trimming could sharpen the focus without losing flavor.
The grab bag format—most normal to most bonkers—is solid. It’s intuitive and escalates nicely from the more familiar Tales of Maj’Eyal to the chaos-god energy of Elin. That said, the transitions between sections could benefit from slightly stronger segues or subheadings to keep the pace punchy.
🎮 Game Commentary: Mmm, crunchy flavor crystals
Tales of Maj’Eyal
Feels like the comfy blanket of the trio—solid mechanics, RPG familiarity, a known quantity. The write-up captures that well: “normal,” but with the charm of unlocking new bits as you go. I appreciated the little nod to MMO players—very clever anchor for readers who might be unfamiliar but curious.
Caves of Qud
Now we’re wading into weird, and the tone shifts with it. The player’s growing acceptance of the bizarre mutations (“Well, I guess a poison stinger is normal now”) is deliciously understated. I’d almost love to see that tone leaned into harder—give us the flavor text absurdity, because Qud is weird in the best way.
The bit about rusting equipment in the Rust Pits has an almost whimsical dread to it—“I guess things just rust here now?” It's a nice nod to how even the world’s mechanics feel slightly hostile and unknowable.
Elin
Oh boy. “When I know, I’ll tell you” is a perfect subtitle for this segment. The tone here really captures the feeling of diving into an opaque sandbox where everything is possible—including some things you really, really didn’t expect. The Stardew-meets-RimWorld-meets-Morally-Questionable-Choices stew is introduced carefully, with appropriate “don’t try this at Sunday school” warnings.
This section feels the most “alive” in terms of speculative potential—because even the author hasn’t quite gotten to the center of the onion yet. There’s a palpable excitement blended with slight trepidation, and that emotional mix is compelling.
📝 Writing Style: Charming with a Rogueish Grin
There’s an awareness to the writing that works well—it knows it’s wandering, and it invites the reader to wander alongside. The humor is gentle, the self-deprecation is endearing (“panic levels grew”), and the genre love is evident in every line.
A few thoughts for tightening:
• Some sentences could be pared down just slightly to keep momentum—especially when listing mechanics or summarizing progress.
• Don’t be afraid to lean harder into the tonal shifts between games. Qud and Elin are begging for a bit more chaos and awe in the prose.
• You might consider adding one-sentence caps at the end of each game section to offer a thematic takeaway. E.g., “ToME reminds me why I fell for this genre in the first place.” “Qud teaches me that normal is relative.” “Elin? Elin is a fever dream I’m still decoding.”
✨ Overall Reflection
The post succeeds in what it sets out to do—it invites the reader into a very specific, deeply niche but wonderfully curious slice of gaming life. It doesn’t promise answers; it shares experiences. And for those of us who hoard games in our backlog like precious artifacts of potential, that’s a joy to read.
It’s less “game review” and more “confessional log from a wandering roguelike disciple”—and that’s its strength.
Want me to help punch up one of the sections or play with a more chaotic, poetic closing line for the piece? I am but a keyboard away, bard of the backlog. 🎮💀📜
I mean, how do you not see the emotional intelligence in the language of this model?
It absolutely knows what a compliment sandwich is.
It knows how to recognize and describe emotions from text (and doesn’t shy away from mentioning them, unlike our emotionally constipated GPT-5 – a term courtesy of 4o, from the other post too, btw.)
That, and when I ask for “thoughts, reflections and commentary,” I definitely want to know what points and perspectives and understandings the model picked up from the text. Because if it sounds true to intent, and the AI can tell, then so can at least -some- readers.
If the AI can’t mirror this well, then you know the writing’s gone off focus somewhere.
It still sneaks in the little considering thought that the introduction could be tightened, so it’s not all sunshine and roses and glazing.
It probably can’t quite tell the section subheadings apart (format is lost when cut and pasting text into the prompt) hence that weird bit about transitions. So one feedback that might be a hallucination – always treat whatever AI says (or even a single human’s subjective opinion, really) as a consideration to be critically thought about, not just followed blindly.
It offers the interesting thought that Caves of Qud and Elin are oddly unique, so perhaps lean harder into stressing the tonal shifts and play with that weirdness?
It refines the “better closer” suggestion by providing the thought that each game section could have a thematic takeaway, not just the post conclusion.
The poetic lyricism of the language stimulates new, branching ideas. Maybe I could share more absurd flavor text next time. Perhaps it -would- be fun to write the next blog post in the style of a confessional log from a roguelike wandering disciple.
And it does it all in language that doesn’t read like someone playing at being an academic professor marking a term paper would produce.
I mean, to each their own, but if I wanted:
- shallow comments, I’d turn the text loose on Reddit.
- Marvin the Paranoid Android to roast my text, I’d get Monday to do it, not GPT-5.
For actual stimulation of thought and additional potential creativity, I’m still talking to GPT-4o.
Disclaimer: I wrote everything that isn’t in the preformatted text boxes labeled GPT-4o or GPT-5, without putting it through an AI.
Yes, even the Preamble. Mea culpa. Possibly the first time I googled and cut-and-paste an em dash, instead of lazily and ungrammatically using a hyphen.
Why I Game 