Before You Complain About GPT-5, Read This
It’s Easy to Criticize. It’s Harder to Build.
The morning GPT-5 launched, I woke to the sound of my phone buzzing. Before my coffee had even cooled, my news feed was full of headlines, screenshots and hot takes from every corner of the internet. Many were quick to highlight what the model still could not do, sharing clips of its occasional missteps as though they defined the entire release.
It reminded me of something that happens whenever innovation crosses a certain threshold. The closer something gets to extraordinary, the more we stop seeing the miracle in front of us and start searching for the flaw.
A few years ago, the idea of an AI that could reason deeply, interpret visuals, switch between formats, and write fully functional code would have been dismissed as fantasy. Today, it is available in a browser tab, and yet it is often measured against an impossible perfection rather than the progress it represents.
Seeing the 99 Percent for What It Is
In 2015, I sat in a cramped meeting room with a team of engineers, debating whether AI could ever build a complete website from a single sentence. The idea was so far fetched it drew laughter. The thought of explaining quantum physics to a ten-year-old in a way they could actually understand seemed even more absurd.
Now, GPT-5 can do both before lunch. This is not a staged demo or a controlled experiment. It happens in real-world scenarios, with real people using it for real work. And yet, somewhere along the way, such achievements have started to feel ordinary.
When Possibilities Become Habits
The most profound revolutions are rarely loud. They do not arrive with fireworks. They arrive quietly, weaving themselves into the daily routines of people who are focused not on hype, but on results. GPT-5’s greatest impact is not in staged demonstrations, but in the way it is already changing workflows across industries.
At Microsoft, GPT-5 now powers Microsoft 365 Copilot, GitHub Copilot, and Azure AI Foundry. Project managers can draft complex proposals in Word with structured recommendations, designers in PowerPoint can generate visual concepts in minutes, and developers in Visual Studio Code can debug problems instantly. Microsoft’s new “smart mode” decides in real time whether a smaller model or GPT-5 itself should handle the request, balancing speed, cost, and accuracy.
At PwC, GPT-5 is in the hands of 100,000 employees across the U.S. and U.K. Legal teams feed it hundreds of pages of contracts and compliance documents, and within minutes it produces detailed summaries that highlight risks and negotiation points. What once took days now takes hours, allowing consultants to advise clients faster and with greater clarity.
In healthcare, Abridge is using GPT-5 to automatically generate detailed visit summaries for doctors. By listening and structuring the conversation in real time, it frees clinicians to focus entirely on their patients rather than dividing attention between the person and the screen.
In creative spaces, GPT-5 has shown its range by building entire apps on the spot. In one demonstration, it created two French learning applications, complete with flashcards, quizzes, and progress tracking in seconds. Studios in Los Angeles are feeding it fragments of ideas, from half-finished scripts to mood-board sketches, and getting back storyboards, dialogue, and production notes that accelerate the creative process without replacing the human vision at its core.
These are not experiments or speculative ideas. They are today’s workflows, quietly becoming the new normal. And they prove something important: transformative technology rarely announces itself. It simply becomes part of the way we work.
Criticism Is Easy. Creation Is Not.
It costs almost nothing to say, “It still makes mistakes,” or, “It is not as creative as I hoped.” But building something like GPT-5 requires years of research, billions in investment, and relentless determination from teams across the globe.
The temptation is to compare it to a fictional perfect AI that never falters. The fairer comparison is to look at what came before. By that measure, GPT-5 is not a small step forward, it is a leap.
There Is No Final Finish Line
Artificial intelligence is not a product that will ever be “done.” There will always be another version, another leap, another set of possibilities. GPT-5 is not the conclusion of AI’s story. It is a milestone, and like all milestones, it opens doors to opportunities that were previously out of reach.
Dismissing a milestone because it is not the destination is to misunderstand the journey entirely.
A Better Way to Look at It
I have started asking myself a different question. Instead of “What can it not do?” I now ask, “What can I build with what it can do today?”
The people and organizations asking that question are the ones already shaping the future.
Educators are redesigning classrooms around new ways of learning.
Developers are delivering projects in days instead of weeks.
Businesses are making strategic decisions with richer, faster insights.
Healthcare providers are reclaiming valuable minutes with patients because the paperwork writes itself.
“Those who wait for perfection will always follow. Those who work with progress will lead.”
My thoughts: Innovation does not owe us perfection. It owes us momentum. GPT-5 has delivered that quietly, powerfully, and in ways we will only fully recognize when we look back years from now and see how much changed while we were busy pointing out its flaws.