13 comments

[ 3.5 ms ] story [ 35.8 ms ] thread
Right, ChatGPT now with GPT-5 has become some sort of router to different suitable models
ChatGPT is a wrapper of GPT
“This stateless nature isn’t a bug - it’s a feature. It makes LLMs:

Predictable: Same input always produces consistent output”

There is no always.

You forgot to mention that this post was written by Opus
This is just semantic pedantry.
> The key insight: when you interact with modern ChatGPT, you’re not just talking to a language model - you’re collaborating with an intelligent agent that uses language models as one component of a much more sophisticated cognitive architecture that includes memory, tools, and orchestration.

Fwiw, right at the end

“This isn’t just about keeping a chat history - it’s about building and maintaining a model of the user, the task, and the evolving context.”

Was this written by GPT? ;)

> This stateless nature isn’t a bug - it’s a feature

> This isn’t just about keeping a chat history - it’s about building and maintaining a model

> Understanding this distinction isn’t just about getting the terminology right. It’s about understanding the future of human-computer interaction

> When we say “ChatGPT” when we mean “LLM,” we’re not just being sloppy - we’re obscuring fundamental architectural and strategic decisions

> when you interact with modern ChatGPT, you’re not just talking to a language model - you’re collaborating with an intelligent agent

No comment.

And there goes the enshitification of all that was good about blog posts: human element.
AI sloppiness of this blog post aside, it's a reasonable observation.

If you're thinking about how to integrate AI into your system, it's worth asking the question of why your system isn't just ChatGPT.

- Do you have unique data you can pass as context?

- Do you have APIs or actions that are awkward to teach to other systems via MCP?

- Do you have a unique viewpoint that you are writing into your system prompt?

- Do you have a way to structure stored information that's more valuable than freeform text memories?

- etc.

For instance, we [0] are writing an agent that helps you plan migrations. You can do this with ChatGPT, but it hugely benefits from (in descending order of uniqueness) access to

1) a structured memory that's a cross between Asana and the output of `grep` in a spreadsheet,

2) a bunch of best-practice instructions on how to prep your codebase for a migration, and

3) the AI code assistant-style tools like ls, find, bash, etc.

So yeah, we're writing at agent, not building a model. And I'm not worried about ChatGPT doing this, despite the fact that GPT5 is pretty good at it.

[0] https://tern.sh

I like this post. I see this confusion all the time! What’s the difference between ChatGPT and gpt-5 or gpt-4o, and so on. OpenAI’s carefully crafted naming schemes don’t help. Though, I come from AWS so glass houses.

Anyway, agents are control systems that using planning, tools, and a collection of underlying models. ChatGPT is an agent. What kind? The kind optimized for the general user looking to do work with public knowledge. That’s the best definition I can come up with.

Anyway, let’s make sure people understand the difference between AI systems and AI models. The former is where a lot of startup activity will be for a decade. The latter will be in the hands of a few well funded behemoths.

This is becoming a problem for me: I like to use LibreChat to interface LLMs (Gemini, GPT-x), because it allows me to pay by usage (instead of 20+ USD per month) and keep a (tiny bit) more of the data under my control.

But the offerings of ChatGPT or Google's AI Studio surpass the feature set of LibreChat by a lot. It used to be just a "better" system prompt, but now it's a lot more.

Call it GenAI.

It’s not acting as an agent unless it does something in the world on your behalf.