In recent days, an article by Adam Khaled has been making the rounds, titled: “Antigravity: The Inevitable Evolution of the Developer.”
Central thesis: November 18, 2025—with the launch of Gemini 3.0 and the Antigravity platform—might be remembered as the day software development changed its nature forever.
Not because of yet another “more powerful model,” but because of something deeper: 👉 the combination of advanced AI reasoning and a new way of developing: agent-first.
A 45.1% That Is Worth More Than a 100
Khaled’s article starts with a number that, at first glance, seems mediocre: 45.1%. This is Gemini 3.0’s score on the ARC-AGI benchmark, created by François Chollet.
If you think of a school exam, 45% is a failing grade. But the context here is different:
ARC-AGI doesn’t measure knowledge,
You cannot “study” for the test,
It is a set of abstract puzzles never seen before,
Humans score practically 100%,
Previous models were stuck near 0%.
In this scenario, going from 0% to 45.1% is a quantum leap. It is the transition from a stochastic parrot to something that begins to resemble a mechanical thought process.
Gemini 3’s Deep Think mode doesn’t just “think out loud” step-by-step. It does something more unsettling and interesting:
It chooses a path,
It realizes it is in a dead end,
It backtracks,
It erases that line of reasoning,
It tries another.
It is the shift from a confident improviser to a skeptical scientist.
From Chat to Control Center: What Antigravity Really Is
If Gemini 3 is the brain, Antigravity is the body. It is the platform that allows that brain to act on code, systems, and products.
Khaled describes it as the end of Vibe Coding: No more loop of: prompt → code → error → apologies → new code → another error.
Instead of the usual chat interface, Antigravity is described as a “mission control room”: 🎯 You don’t ask: “Write me a function for X.” You ask: “Restructure the authentication flow to support passkeys.”
From there:
An analyst agent scans the codebase,
A researcher agent reads documentation and APIs,
A developer agent writes the code,
A tester agent runs the tests,
A debugger agent analyzes logs, fixes, and relaunches the cycle.
All automatically. The human is no longer at the center of the action, but at the center of supervision. Khaled calls this Agent-First Development.
Extinction or Metamorphosis? From Writer to Orchestrator
Here the article becomes brutal. “If your core competency is knowing the syntax of a language, you are unemployed.”
The logic is ruthless but linear:
If the marginal cost of code goes to zero,
Then the value of someone manually writing code goes near zero.
But this isn’t just a story of extinction. It is a story of transformation.
Khaled draws a clear line between:
Developer-Writer (2015–2024) – knows libraries and syntax – spends hours on bugs – is proud to say: “I built this feature myself.”
Developer-Orchestrator (2025+) – designs systems, flows, contracts between services – defines tests, constraints, acceptance criteria – says: “I manage the swarm of agents that builds the feature.”
In between, new roles are emerging:
Verification Engineer – stress-tests AI-generated code.
Agent Architect – designs the “team” of agents and their coordination.
Model Evaluator – assesses output, risks, bias, and robustness.
The real question becomes: 👉 “Do I know how to lead a team of AI agents?” If yes, your value grows. If not, the curve gets steep.
The Future of Programming Is Not a New Language
The article’s conclusion is almost a provocation:
“The future of programming is not about learning a new language. The future of programming is English. And your new compiler is an alien super-intelligence that costs a few dollars per million tokens.”
Behind this spectacular image lies a bigger question:
If tools like Antigravity make development almost frictionless, almost free, and accessible to anyone who can describe an idea… who will truly benefit?
Perhaps not just the developers who manage to reinvent themselves as orchestrators. But the hundreds of millions of people who, for the first time, will be able to build products, services, and tools without knowing how to program.
And, paradoxically, the real event won’t be the extinction of developers. It will be the explosion of creators.
