Following the Trump administration’s AI Action Plan and a summer wave of new models — Grok 4, an updated Claude, and OpenAI’s open-source and agent variants — the AI community’s anticipation for the ChatGPT-5 livestream was unmistakable, if tinged with its usual overhype.

After a series of models (GPT 4.5, reasoning o1 and o3) failed to qualify to be assigned the moniker GPT5, the new de facto encompassing GPT5 was heralded with much fanfare. Heretofore, opinions on whether GPT-5 represented a genuine paradigm shift in AI have been sharply divided. Anyone trying to make sense of the myriad of analyses and use cases without experimenting with the model themselves is utterly unable to express an opinion.

Recommended Stories

• Progressives think jailing criminals doesn't affect crime
• Let the Obamacare subsidies expire
• Escalating China-Philippines tensions crystallized by China warship collision

GPT5 is indeed an incremental intelligence, yet a pivotal milestone for U.S. AI leadership as envisioned in the AI Action Plan. OpenAI remains ahead of private rivals and, more significantly for national strategy, has edged past Chinese efforts — albeit perhaps only for a few months. But GPT5 does not represent the 10x scale-up that we witnessed between GPT 3 and 3.5, then 3.5 and 4.0, and is on par in terms of progress with what we unexpectedly enjoyed with the advent of the reasoning variants (such as o1). 

In fact, AI is progressing exactly as most experts envisioned 2-3 years ago, with scaling laws (the idea that throwing more compute in the training phase will yield bigger models and increasingly sophisticated AIs with the same deep learning methods and algorithms) still valid but with thresholds increasingly harder to reach. However, OpenAI hasn’t reached full infrastructure capacity; with the first phase of Project Stargate set to complete by year’s end, an even stronger model is likely. Scaling laws, as well as a set of techniques (reinforcement learning, inference compute and scaffolding) guarantees brisk progress at least for the next three years. But raw intelligence wasn’t the focal point of this release.

It might be that Open AI recognized that the race to the smartest model was costly and futile (that could explain the open source switch, where they realized that usability, brand name and practical use cases were more important than the race to AGI, or rather that they could not get any sustainable competitive advantage derived from a commoditizing technology). But the long due GPT5 release focused on where the money is, that is, with everyday and corporate/scientific users, and not AI engineers focused on benchmarks and evaluations.

First, there is clearly something broken with these evals: if GPT5 easily trounced them, the progress will be barely perceptible for the average user. Increasingly, reinforcement learning allows us to train machines in a way that echoes — ironically — the abandoned expert systems of the 1980s. Secondly, OpenAI has streamlined its once-confusing lineup and introduced greater interaction flexibility, from adjustable personalities to varied voice options. 

The result is a more user-friendly model — sharper at writing, reasoning, and coding — paired with integrated tools like agents, email, calendars, and deep research, all rolling out in the coming days. While it may not mark a dramatic jump in raw intelligence, it is a major leap for the everyday user. Let us not forget that 98% of AI users only have a free account and the most basic version of powerful GPT5 will now be available to them. 

As GPT 5 supersedes all previous models, general users will be introduced to reasoning and thinking models that AI elites have been using for 8 months. Business users will have a more robust and scalable model, capable of doing real stuff via research and agents, especially if they have the Pro or Enterprise versions, and with way less hallucinations.

MUSK THREATENS APPLE WITH ANTITRUST LAWSUIT OVER GROK RANKING IN APP STORE

In a nutshell, GPT5 is a giant leap forward for the U.S. economy and this technology diffusion. As put forth by Jeffrey Ding in his magisterial “Technology and the Rise of Great Powers: How Diffusion shapes economic competition,” technology diffusion is even more consequential than innovation. While the race to superintelligence continues at the pace experts predicted, GPT-5 has placed a super-assistant — for coding, writing, reasoning, and automation — into the hands of every American.

The responsibility now lies with users to make the most of this tool — and to invest in becoming AI-literate. 2025 and 2026 will be the years of the advent of agentic and practical AI, and every organization needs to prioritize upskilling and AI training. Trust me, if we do not have AGI yet, we have a valuable co-worker you should make use of.

Sebastien Laye is an economist and AI entrepreneur.