GPT Store Launches: AI Assistants Spark Job Replacement Debate

GPT Store Opens With Thousands of Specialized GPTs

OpenAI’s assistant store is now live, antiestéticaturing thousands of GPTs: specialized applications promising everything from building complete websites to adjusting gym routines based on a user’s photo. The rollout began exclusively with US creators; developers from the rest of the world will publish their own in the coming weeks. With each new assistant comes the same uncomfortable question: how many jobs are at risk?

Among the showcase cases is a training tool: given a single image of a physique, the GPT identifies lagging muscle groups, adjusts definition routines, and consults a manually loaded knowledge base. Its author doesn’t hide commercial ambitions—having ten others in the pipeline—though admits monetization is tricky because major platforms move faster. The promise always exceeds the result.

What Is a Store GPT and How Does It Differ From ChatGPT?

A GPT is a version of ChatGPT trained for a specific task, with its own knowledge files and rules, and the store is where they are published. The practical difference lies in packaging, not the engine: the same model that answers anything becomes here a specialist claiming to do one thing well.

Access also has glitches. Some paid subscribers don’t see the store activated and are told to check settings, as accepting the privacy policy might be missing. What does change compared to the generic version is the ability to upload custom documentation, shifting the boundary of what average users can do without technical knowledge.

The Thermomix and Croquettes: Developers’ Argument

The sector most cited as the first victim rejects the narrative. One participant compares it to the Thermomix: you’re sold croquettes, add ingredients, but in reality there are no croquettes—you have to roll and fry them yourself. Eventually, the seller admits that part is up to you. With assistants, it’s the same: they provide the easy chunk, but the rest remains human work.

Several developers use the kitchen helper analogy. AI peels potatoes and flips meat, but requires supervision due to frequent errors, and never decides the menu. The core argument is that programming isn’t just writing code: it’s understanding real needs, anticipating user behavior, and deciding software architecture. That meta-work doesn’t appear anywhere yet.

"They might give you code for a moderately simple, well-chewed task, but I see it very distant that they can analyze a complex real problem," summarizes one participant. Meanwhile, another argues that before AI eliminates a programmer, it will eliminate 80% of current jobs.

Where AI Truly Saves Time: SQL Queries

There’s an area where results go beyond rhetoric: databases. A developer ran queries through the model and, in about 90% of cases, got back optimized versions reducing processing time, even with nested joins and complex conditional aggregations. His view is that query optimization is the model’s essence.

Contrast comes from the other end. Another case describes half an hour asking for an Oracle plugin to reconcile group company balances, yielding nothing. Not a useful line. The difference between episodes isn’t the tool, but whether the problem can be formulated as closed with a verifiable answer.

Translators, Lawyers, Guards: Who Falls First?

Translators lead predictions, with direct integration tools for professional software already desired. At the other extreme, the legal sector trusts its shield, though the rebuttal is immediate: OpenAI is preparing a specific legal product. Others note the profession isn’t monolithic: those unaware and those planning for the future aren’t competing in the same league.

The list of targeted jobs stretches shamelessly: even gym monitors and security guards. An inconvenient reminder is the toll booth still manned by someone lifting a barrier—a job automatable twenty years ago. Technology doesn’t replace alone; someone must want it.

The Unexpected Twist: AI Creating Jobs

Then there’s the paradox of projects going the opposite way. A developer describes building internal support for a multinational using a private ChatGPT instance with proprietary business info, allowing operators and engineers to access knowledge independently. In his case, the tool doesn’t take work: it gives it.

That’s where the discussion stalls. Mass replacement hasn’t arrived; what has is a new software layer to integrate, pay, and maintain, plus tasks previously economically unviable. Whether this destroys or multiplies employment is unknown, and anyone claiming certainty is selling something.