@juanjovn The goal of weekly limits is to stabilize supply and demand and provide predictability.
No feature request will take this goal away. Supply and demand will always exist and will always be a problem for bandwidth / compute providers
@birdabo This is just not true.
The subscriptions are cheaper than the API pricing, but the API pricing is for a completely different demographic.
You have no idea what their margins are outside of the API pricing evidence, which again, isn’t accurate.
@macabi3059@grok Perfect! @macabi3059 I think it’s so much more enjoyable for readers to hear from you, not from ChatGPT. You clearly have a great perspective
This is cool, but AI labs are not subsidizing AI usage.
This is like saying pharma subsidizes the cost of drugs… they pay smart people to establish the drug, run tests and release it and be liable
Once the drug comes out, it costs almost nothing to produce
It’s all about compute allocation. API tier pricing does not mean that’s how much it costs. It means they can charge more for their models and servers to reach their models because the customer needs that type of product.
Most API revenue comes from being the foundational layer for AI products
Meaning businesses are making money off of the models, hence why it’s more expensive
For those loving this prompt:
It was created by AI, not a human
If you want an even BETTER prompt…
- Type in what you want to accomplish
- “interview me one question at a time to uncover how i can best use you to accomplish this goal”
If AI is creating the “prompt” either way, you might as well make it personally applicable
Please don’t forget
When people “reveal” “million dollar secrets”
at that exact moment
(If they were ever truly million dollar secrets)
they are no longer
another trick to avoid claude writing slop:
add this to your claude .md (alongside the simplified tech english):
"follow Zinsser's four principles of quality writing:
1. Simplicity
2. Brevity
3. Clarity
4. Humanity"
next, create a forbidden .md file where you add patterns to avoid as soon as you spot them, this is the extra layer that catches slop before it ships
examples of what to ban:
> Staccato pairs
> Antithesis reframe / negative parallelism
> Isocolon metaphor-pairs
> Backward-references
last thing, if you have a writer agent, make sure to keep it under 50% context window at all times, call subagents if you need to or the quality drops fast
the subscription price & API price are not comparing apples to apples.
Customers on API pricing are mostly businesses built off ChatGPT’s capabilities. So, they are MAKING MONEY on every call.
One could argue the true subsidy is the customers paying for API pricing because most of the time they give free usage to get their customers hooked.