GPT-5 is the symptom, the era of easy scaling is over

The AI hype cycle just collided with reality.

It's called GPT-5.

For months, we heard whispers of a revolutionary leap towards AGI - the pursuit of a machine that thinks, not just computes. Sam Altman promised a model that felt like "talking to a legitimate PhD level expert."

The reality is a multi-billion dollar shrug.

To be clear, there are incremental gains. It's better at coding, and OpenAI reports fewer hallucinations in certain modes⁷. But these are minor tune-ups. They are a distraction from the fundamental story.

The fundamental story is this: the engine of AI progress has stalled. The era of easy scaling is over.

This warning doesn't come from a cynic. It comes from the heart of the industry. Ilya Sutskever, former co-founder of OpenAI and co-founder of the AI lab Safe Superintelligence (SSI), publicly declared this was happening back in late 2024.

He stated that the progress from simply making models bigger has plateaued, primarily because the internet - the "fossil fuel of AI" - is a finite resource that has been nearly exhausted³.

What we see with GPT-5 is not a clumsy launch. It is the predictable symptom of this deeper disease. The low performance gains are a direct consequence of this scaling wall.

These are not isolated problems. They are the cracks appearing in a foundation that can no longer support the weight of expectation. The path to AGI does not look like this.

The brute-force approach has failed. The investor dream of replacing all human expertise with a single, god-like machine has hit a wall.

And this presents a strategic inflection point for European industry.

The game is no longer about who has the most computing power or the largest trove of scraped data - a game the EU was set to lose. The end of scaling marks a new era.

For the near future, progress will not come from bigger models, but from customizing more reliable systems.

This is a game of ingenuity, precision, and domain-specific craftsmanship.

A game built on the very strengths that define European companies and especially the German Mittelstand.

Sources:¹ Gary Marcus, "GPT-5: Overdue, Overhyped, and Underwhelming," Substack, August 2025.² Ethan Mollick, LinkedIn Post on GPT-5 user experience issues, August 2025.³ Ilya Sutskever, Statements at NeurIPS 2024 and in Reuters interviews on scaling law limits.⁴ Andriy Burkov, LinkedIn Post on GPT-5 results, August 2025.⁵ Lasse Rindom, LinkedIn Post on GPT-5 technical regressions, August 2025.⁶ SplxAI, "GPT-5 Red Teaming Results," SplxAI Blog, August 2025.⁷ OpenAI, "Introducing GPT-5," OpenAI Blog, August 7, 2025.