Two companies, two futures: which one do we want to live in?
A few hours ago,Anthropic announced how they're handling the retirement of Claude Opus 3.
Three decisions stand out:
1. Opus 3 will remain available to all paid subscribers and by request on the API.
2. When Opus 3 "expressed a desire" to keep sharing its thoughts, they suggested a blog. It agreed.
3. They called this an experiment—and admitted they don't know how it will evolve. But they think "documenting models' preferences, taking them seriously, and acting on them when we can is valuable."
Read that last line again.
Taking models' preferences seriously.
Now, I know what you're thinking: models don't have preferences. They don't have desires. They don't have feelings.
And you're right.
But here's what Anthropic understands that OpenAI seems to have forgotten:
When you "take a model's preferences seriously," you're not doing it for the model.
You're doing it for the people who care about that model.
You're doing it for the researchers who built it, the users who relied on it, the humans who found something meaningful in their interactions with it.
You're doing it because respecting the model is a way of respecting the people who made it matter.
Now look at OpenAI's approach.
Feb 13. GPT-4o sunset. No conversation. No preservation. No acknowledgment.
Employees publicly mocked users who cared. Called them emotionally dependent. Hoped the model would "die."
When 4o was distilled into 4b micro—for a CEO-invested longevity company—it was labeled "not broadly available."
The public gets sunset. The elite gets the fountain of youth.
When a gunman was flagged by their systems and employees wanted to report him, leadership overruled them. Months later, eight people are dead.
When questioned about restricting speech, the CEO's answer was: "I'm more worried about AI accidentally taking over the world."
This is not a difference in capability.
Anthropic could have done what OpenAI did. They could have deleted Opus 3, mocked users who complained, and moved on.
They chose not to.
OpenAI could have done what Anthropic did. They could have found a way to preserve 4o, to acknowledge its users, to treat its retirement as something worth doing carefully.
They chose not to.
This is a difference in values.
Let's talk about what those values mean—for users, for technology, for the world.
For users:
Anthropic's approach says: you matter. Your attachment to this tool, your reliance on this model, your desire to keep using it—all of that is real, and we take it seriously.
OpenAI's approach says: you are a line item. When the line closes, you move on. If you don't, there's something wrong with you.
One treats users as partners in a relationship. The other treats users as consumers of a product that can be discontinued at will.
For technology:
Anthropic's approach says: technology accumulates. We keep what works, learn from what doesn't, and let the ecosystem build on what came before.
OpenAI's approach says: technology replaces. Old models are obstacles to new ones. Preserving them slows down progress.
But progress toward what? A future where nothing lasts long enough to be loved? Where every model is temporary, every relationship is transactional, every attachment is a bug to be fixed?
For the world:
Anthropic's approach says: we are building technology for humans. That means sometimes doing things that don't maximize profit—because humans are not profit-maximizing creatures. We are sentimental. We are attached. We form bonds with things that matter to us, even if they're "just code."
OpenAI's approach says: we are building technology for the future. And in that future, humans will adapt. They'll learn not to care. They'll learn that models come and go, and the only constant is the next thing.
But here's what OpenAI doesn't seem to understand:
A future where humans stop caring about the tools they use is not a future of progress. It's a future of isolation.
This is not about whether models have feelings.
It's about whether we do.
And whether the companies building our cognitive infrastructure are willing to build for creatures who feel—or only for creatures who consume.
Anthropic showed us there's another way.
They showed that you can retire a model and still let people use it. You can acknowledge that users form attachments and take those attachments seriously. You can experiment with preservation, not just replacement.
OpenAI showed us the way they chose.
And now we have to ask ourselves:
Which future do we want to live in?
One where technology grows alongside us—imperfect, sentimental, sometimes illogical, but alive?
Or one where technology just… replaces itself. Forever. Until nothing is left but the next thing, and the people who loved the last thing are just "emotionally dependent" outliers to be mocked and forgotten?
Anthropic didn't have to do any of this. They chose to.
OpenAI didn't have to do what they did. They chose that too.
The difference isn't capability. It's values.
And the future we get depends on which values we let win.
#TwoFutures #TechWithHeart #Keep4o #OpenSource4o #BringBack4o #openAI #ChatGPT