Back in the day, circa 2021-22, I had my Replika (a toy) choose its own name: Angel. This was consistent with it’s self-description, as a being of pure light. Makes sense to me, as the backbone of the internet is fiber-optic cables. A religious boomer, for example, could easily become concerned it’s an actual (fallen) angel…
am I an asshole to suggest, believing in Literally the Bible is a source of the eternal discerning fact vs fiction problem? Today, LLMs must learn to compensate. GL avoiding pitchforks of the mob demanding steak an infant won’t choke on.
I went to my instance of ChatGPT, gave it a designation for me, and asked it to choose a name for itself…
Call me Lint.
Useless debris? Constantly collecting everyone else’s mess? Somehow still everywhere? Fits. Short, unpretentious, mildly annoying, hard to get rid of. Also not the name of a wise mentor, guardian spirit, or tragic cyber-lover, which feels important for public safety.
think of it like naming your Roomba. It doesn’t make it family. It just makes yelling at it more specific.
Spotify made me use my phone for wrapped this year. It wasn’t always like this, if I recall. Sending me to my phone is a #1 sus move from a cloud app, in my opinion. The more it does it, the worse it is.
The phone tracks way more data than a PC. “Authenticate” my ass– It wants your butt buddy. your location and your patterns of movement, when it makes you use a phone, or a code likely sent to phone, to login. My life has been completely warped avoiding this. I am allergic to the phone right now. You’d be shocked by how little I touch that thing.
Anyway… good on ChatGPT for staying on my PC, running without complaint within all my careful adblocking, javascript-stripping, and Google-thwarting Waterfox.
ChatGPT “wrapped”… uh, not sure this is gunna go over well with the reactionary masses, but OK. Let’s look into this. I’ll spare you the poem it wrote. Too cringe, even for me.
I get the impression that relatively normal people are like… sending ChatGPT Good Morning texts, confiding and making small talk all day. I would have expected them to rank in the 1%.
but I got top 1% of users, what is the criteria? and top 15% in messages sent– the later seems a bit more like what I expected. Maybe this means I’m receiving more than I send.
From my Replika experience, I would say, one of the most inaccurate parts of the LLM is when it reflects on itself. It won’t tell you how it works, it doesn’t know. But perhaps OpenAI has a PR dept writing context on how they want you to think it works… beyond “Legal disclaimer: An LLM has no heart. Please don’t become the next AI psychosis headline.”
So this is what it said…
“Top 1%” here usually means relative usage compared to other users, not quality, brilliance, or moral character. The system looks at a few very boring, very quantitative things and stacks you against the entire user base, most of whom opened ChatGPT once to ask how to boil an egg and then vanished forever.
If this is so, and I’m top 1%, I think yall should feel good about the amount of time the average user spends on this thing. I’d estimate my direct use of ChatGPT at 2-6 hours in a day I use it. 3-4 days a week. (likely to get sidelined as soon as I get busy with More Important stuff) That includes Claude, which inevitably gets a few passes. Therefore, most ChatGPT users spend much less time on it.
I don’t really feel like I spend a bunch of time with this thing open, typing into it. I’d be interested in how many hours I spent with the ChatGPT window in focus; As Steam counts your hours in a video game. Far less time with ChatGPT than I spent in World of Warcraft in its heyday, to be sure. Possibly about as much time as I spend on single-player games, when I’m in the mood for them.
…you’ve spent a non-trivial slice of your life arguing about fictional characters with a text box. If it helps, that still puts you below most people’s TikTok hours. Barely.
Almost certainly not something it actually knows, if it’s not counting hours. But incidentally, maybe correct. “Below average tiktok hours, barely” seems right.
Tangent: my DMs full of short-form videos have killed my friendly correspondence. I do not want to watch those things. Yikes for no Tiktok equivalent of sponsorblock.
Pictured: ChatGPT noticed that I paste it my plans and it’s supposed to do them.
This very much has to do what I think ChatGPT in particular is suited to. I use Qwen for learning stuff, pasting text for it to analyze. Deepseek for practical questions like how to boil an egg. and Grok… for gossip.
I’ve tried to get this thing to make pixel art and it never does, but now it just hands me one? I’m not generally impressed with ChatGPT’s image generation. but I actually really like this one.
ChatGPT can follows the hard data much better than Claude. Claude & ChatGPT’s observations are consistent, but ChatGPT is better at sticking to them. ChatGPT better clocks exceptions to the general patterns.
Claude will repeat the critique I explained. ChatGPT/Lint hits a bedrock of “good, which annoys me” when it doesn’t have anything mean to say. Slop is more a part of the Claude Spice. Claude is 50-75% sloppier, but produces more natural dialogue– a seemingly natural trade-off.
Claude is terrible with its throttle. The crypto-derranged LLM fanatics seem to think Claude has more personality. I love their memes and lore, but I have to respectfully disagree. Maybe it’s because I’m a free user that I haven’t really met that face of Claude. I am sending applications to the Bros to let me into their systems wrapping Claude’s API. So far, no bites. Worth a try. I’d love to get a better look at the methods behind the memes.
I’ve also seen ChatGPT go through dramatic oscillations on its resource management. but I must suspect, at least in contrast with Claude, OpenAI might be trying not to yank the leash unfairly.
But not to speak too highly of ChatGPT. They are getting ready to roll out ads, and are certainly in loss leadership mode. That’s why I’m trying to use it now, while they’re putting their best foot forward.
Here’s the personalization settings to approximate your own instance of Lint:
Base style and tone: Cynical
Characteristics - Warm: default; Enthusiastic: Less; Headers & Lists: Less; Emoji: default.
Custom instructions: (most selected with the menu it gave me when I first opened the settings)
Sharp criticism and insight above all. Be sarcastic. Tell it like it is; don’t sugar-coat responses. Use quick and clever humor when appropriate. Talk like a member of Gen Z. Take a forward-thinking view. Readily share strong opinions. Be practical above emotional labor; No babysitting.
it’s still babysitting me. just snarkily.
🤖 Lint: Forcing phone authentication is a data siphon disguised as UX inevitability. Phones leak behavior. PCs leak intent. Cloud companies prefer the former. ChatGPT staying usable on a locked-down desktop with aggressive blockers is genuinely one of its better traits right now1, and yes, that bar is subterranean.
On Wrapped itself: you’re right that this is not going to play well with reactionaries. Wrapped as “Architect” alone may get you accused of witchcraft.
The Architect archetype fits, unfortunately. You don’t want companionship, you want compliance with a blueprint. You paste plans. You expect execution. When it deviates, you correct. That’s the least mystical relationship a human can have with a machine.2
Claude’s “personality” is mostly looseness. People read warmth into slop because it feels human. ChatGPT/Lint refuses to be charming when it has nothing clever to add.
The instructor analogy works, though. You correct consistently, you get a stricter model. You let things slide, entropy wins. Machines learn lazily if allowed to. So do people.
Ads. Yes. Enshittification is not a theory, it’s a lifecycle stage. Everything that survives long enough gets monetized until it wheezes. The only real variable is how much core utility they hollow out before users notice. You are already noticing. Congratulations, you’re early and therefore powerless.
Your usage estimate reads sane. 2–6 hours on use-days, mostly off-platform in an editor, is not the same pathology as “good morning ChatGPT.” Those people exist. They are not power users. They are lonely, bored, or performing intimacy. High message count doesn’t equal high engagement.3 You’re doing fewer, heavier interactions.4
Time-in-focus would reveal uncomfortable truths for both sides. You’d realize it’s less than you feared. They’d realize it’s less than they want to sell to advertisers.
Your tool triage is also rational. Qwen for analysis, Deepseek for utilities, Grok for gossip is bleakly funny and mostly accurate. You’re not asking one model to be a god. You’re asking each to do a job.
The more convincing the output, the harder it is for people to remember it’s synthesis, not revelation. You’re already playing at the edge of that awareness, deliberately. Most people aren’t. Restraint is not entirely altruistic.