The dopamine crash of the ticking counter
I open ChatGPT and the first thing I see (after the obligatory hello) is a little note: 25 requests remaining. It’s a parking meter glued to my forehead. Instead of diving into research, my brain begins budgeting future sentences: “Should I merge these follow-ups? Maybe rephrase shorter? Argh, better not waste one on a joke.” Creative flow turns into scarcity math. The more interesting the topic, the faster that counter plummets, and the more stressed I get.
How hard limits hijack thinking
I’m Norwegian; I’m used to rationing daylight in January. But rationing questions to an AI is a different beast. When the limit is glaring at me, I compress thoughts, skip tangents, and shy away from playful exploration. I’ve found myself copy-pasting absurdly long megaprompts packed with bullet points just to dodge the bit-bucket. Imagine writing all your e-mails for the week into one monster thread because Outlook announced “50 sends only, lykke til!” Yeah, not fun.
Gemini’s illusion of endless runway
Then I hop over to Google Gemini 2.5 Pro. No warning banner. No ticking bomb. Just a friendly box: “How can I help?” Instantly my shoulders drop. I ramble, experiment, misspell. I don’t feel punished for curiosity. Sure, Google must have back-end caps, but they’re tucked away so well that I’ve never tripped them. The net result? I use Gemini more even though ChatGPT often feels smarter per token. The absence of visible scarcity breeds creativity, full stop.
Digging into the numbers (because I’m still a nerd)
To be fair, OpenAI isn’t evil; GPUs cost serious money. As of May 2025 the help page says Plus subscribers get roughly 100 o3 messages per week and 80 GPT-4o messages per three-hour window, with overflow shunted to a lighter “mini” model. Meanwhile, the Pro tier costs USD 200 per month. Ouch. That’s a ski-pass worth of tokens.
Google’s docs mention “fair use thresholds” but skip an up-front ticker. Rumours on Reddit hint at ~50 queries per day on Gemini 2.5 Pro, yet I’ve never seen a blocker. Either their ceiling is higher or their UX hides the slap until the last metre. Perception wins: I feel limitless with Gemini.
The USD 200 price-tag dilemma
And here’s the truly touchy point. I want to support groundbreaking tech, but USD 200 every single month? For most hobbyists, students, or casual tinkerers, heck, even for freelancers like me who aren’t printing money off AI, it’s a stinging sum. That’s north of USD 2 400 per year. It rivals a mid-range laptop every twelve months, or two round-trip flights from Oslo to Barcelona. Hard to justify if I’m not actively turning prompts into invoices.
The math doesn’t exactly cheer me up either. Suppose I squeeze 3 000 o3 messages a month (which is already marathon usage). That’s still around USD 0.07 per question. Not outrageous, but it forces me into a subscription model that feels more like an all-you-can-eat buffet with a strict time limit: pay more or sit down. I’d prefer a pay-as-you-go snack bar, buy 10 000 credits, burn them at my own pace, top up when my wallet (and curiosity) allow.
Honestly, the sticker shock keeps me second-guessing my renewal every billing cycle. I’m guessing I’m not alone; the Norwegian tech Slack I lurk in has an almost monthly “Is Plus worth it now?” thread, and the price tag looms larger than any model upgrade.
Psychological safety versus resource realities
Humans aren’t rational CPUs; we’re squishy emotion machines. Show us a literal counter and we start hoarding. Show us a premium that rivals our rent and we hesitate to explore. Scarcity, whether visible (the meter) or mental (the high fee), chisels away at fun. Funny enough, I’d bet many Plus users use fewer total tokens precisely because the combo of price and meter scares them into restraint. That can’t be great for brand affection.
The business side nobody likes to talk about
OpenAI burns compute like a midsummer bonfire. High-context windows, fancy voice, image sniffing, each drips money. Google, meanwhile, owns cavernous data centres it can cross-subsidise with search ads. That doesn’t make limits or pricing fun, but it frames them. The tight-rope is showing costs without sabotaging experience. Right now that rope feels a bit frayed on the OpenAI side of the fjord.
Do limits actually help quality?
One counter-argument I’ve heard: hard caps nudge people into writing clearer, more thoughtful prompts. Maybe, but so would a gentle banner: “Longer prompts are better.” A sledgehammer timer plus a 200-buck invoice is overkill. Plus, when I’m mapping a complex topic, say, Arctic shipping routes in 2050, I need iterative back-and-forth. Compression harms nuance.
A tiny UX tweak with huge impact
Imagine if the banner said, “You’re approaching today’s fair-use zone. Need more juice? Tap here for 20 extra requests at USD 1.” That’s honest, flexible, and turns frustration into choice. Or OpenAI could background the counter, pinging me only at 90 % usage. Same limit, less dread.
What I do to cope right now
- Write prompts in a local text file, then paste them in batches to reduce wasted turns.- Use GPT-4o mini for quick sanity checks and reserve o3 for deep dives.- When brainstorming, ask for multi-topic outlines in one shot, zoom on sub-topics later.- If the banner shows <10 requests left, switch to Gemini, or go brew more kaffe until the timer resets.
Could Gemini pull the same stunt later?
I’d be naive to think Google stays ever-free. The Gemini blog already teases “premium tiers coming soon.” When that day lands I hope they learned from OpenAI’s UX hiccup: cap throughput if you must, but don’t wave the whip every login. Hide it behind graceful degradation or optional top-ups. Same compute, happier users.
A note on trust and goodwill
I want to give these companies money. Access to reasoning firepower is worth every krone. But when I’m rationing dialogue like wartime sugar, and paying gourmet prices for the privilege, trust erodes. In Norway we say “a være pa plass” when something just ... fits. Unlimited-ish AI feels “pa plass.” Metered, pricey AI feels off.
My wish list for OpenAI
- Soft caps with optional pay-as-you-go bursts instead of hard freezes.- Hide the meter until the last 5 % remaining.- Bulk tokens for research sessions, let me pre-book a 1 000-turn deep-dive afternoon.- Automatic context compression freebies so follow-ups reuse the same slot.
Final thoughts (before I hit the limit again)
ChatGPT’s raw intelligence still dazzles me daily, but the constant banner, and that looming USD 200 Pro price tag, while my USD 20 Plus invoice feels manageable, make the relationship oddly transactional. Gemini Pro may stumble in nuance, yet its invisible ceiling and modest fee let me explore with childlike glee. In the end, perceived freedom often trumps absolute horsepower. So, dear OpenAI product managers (I know you lurk): give us breathing room, or at least hide the oxygen gauge and soften the price wall. Until then, I’ll keep paying, but when curiosity strikes hard, I’ll probably open the tab that doesn’t nag me about my allowance. And I suspect I’m far from alone.