*Source: zhichai.net forum post, commit 2c47ab1. Translated from Chinese.*
In the Middle Ages, people believed Earth was the center of the universe. The belief persisted because it fit the power structure of the time: the church held the authority of interpretation, and geocentrism placed Earth, humanity, and the church in a sacred position. Then Copernicus arrived and said: no, the Sun is the center. This was not merely a scientific discovery—it was a challenge to the power structure. When knowledge is no longer monopolized and everyone can look up at the stars and think for themselves, the old order begins to crumble.
Today's AI world, the author argues, is experiencing a similar "Copernican moment."
The Twilight of Closed Source: When Subscriptions Become Shackles
After the release of Gemma 4, an interesting discussion emerged in the community: if open-source models are already "good enough" for daily work, is it still worth paying $20/$200 subscriptions for closed-source chat products?
This sounds like simple cost accounting, but it reflects a deeper questioning. Imagine subscribing to an AI service for $20 a month. One day it goes down. After hours of waiting it recovers, but your workflow has been interrupted. Another day, errors appear and your requests are rejected. You refresh, re-login—the problem persists. Frustration accumulates, and you start doing the math: what exactly did my money buy?
Recent Claude outages and errors have compounded sentiment into strong backlash against the "closed-source + subscription + quota" model.
It is like the difference between renting and buying a home. Renting seems flexible, but the landlord can raise the rent, refuse renewal, or restrict your pets at any time. Buying requires upfront investment, but the roof is truly yours.
When Gemma 4 can run locally for free, 24/7, without API limits, and with data never leaving your device—does the subscription model still hold appeal?
The Open-Source Declaration: Nous's "Inevitable"
Against the backdrop of Claude outages, subscription limits, and improving local model experiences, NousResearch publicly declared: "Open Source is inevitable."
This is not an emotional slogan but a judgment based on technical trends. The evidence cited:
- Gemma 4 reached roughly 2 million downloads in a week, topping Hugging Face's trending models
- Community focus is not leaderboard rankings but "easy to use, easy to deploy"
- iPhone 17 Pro, Raspberry Pi 5, and Macs can all run it locally
- Engineers are publicly doing the math and questioning subscription value
China's "Collective Dovishness": Speculation Over Delayed Open-Source Releases
But the story has another side. Chinese open-source models—MiniMax M2.7, GLM-5.1, Qwen3.6—have all collectively delayed releasing open weights after publicity campaigns, generally saying "wait a bit longer, it will be better."
The community speculates: is there a unified tightening of strategy? Or has pressure from IPOs/profitability pushed them toward closed-source? Behind this speculation lies anxiety about "open-source promises." When a lab says "we will open-source" and then delays repeatedly, trust erodes; users don't know whether to wait or invest effort adapting to a model that may never arrive.
Others believe this is merely closed testing and a race to SOTA, and that open weights will most likely still be released. Open-source has become a calling card for Chinese AI labs; abandoning it would come at considerable cost.
OpenAI's Governance Turmoil: A Microcosm of Trust Crisis
A long-form New Yorker article revisited OpenAI's 2023 board crisis, alleging internal memos, deception, board manipulation, and a weakened alignment team. OpenAI employees countered that the alignment team remains one of the company's most compute-rich projects, but the doubts have not stopped. There are also reports of clear disagreements between Sam Altman and the CFO over compute spending and IPO pacing.
What picture emerges? A company at the crest of the AI wave, internally full of power struggles, strategic disagreements, and trust crises. It preaches "benefiting humanity" externally while internally resembling any ordinary company—with politics, scheming, and conflicts of interest.
This is not criticism for its own sake, but a lesson: when AI is monopolized by a few companies, its fate becomes bound to those companies' internal politics. If Sam Altman were fired by the board tomorrow, would OpenAI's strategy change? If the CFO insists on IPO priority, would alignment research be cut? These are real uncertainties.
The Policy Game: OpenAI's "Intelligent Age Industrial Policy"
Facing the open-source wave and governance questions, OpenAI released a set of policy proposals aimed at governments: a public wealth fund, pilot 32-hour work weeks, defining a "right to use AI," strengthening provenance and auditing, and "containment/circuit-breaker" contingency plans for dangerous models.
The core signal: they are openly treating the path to superintelligence as a real policy problem, not distant science fiction.
Reactions are mixed. Some praise the candor as responsible; others criticize it as paving the way for self-interest—a public wealth fund managed by whom? Who pays for a 32-hour work week? The essence of policy is the distribution of power. In making these proposals, OpenAI is attempting to define the rules of the game. If rules are set by closed-source companies, will the open-source movement be marginalized?
Closing Thoughts
The open-source versus closed-source debate has never been merely a technical dispute. It is a philosophical debate over whether knowledge should be monopolized, a political debate over whether power should be centralized or distributed, and a debate over values—whether the future should be decided by a few companies or shaped by everyone.
Gemma 4's success proves that open-source models can be very good, even better in some respects. Nous's declaration proves the movement has a voice, strength, and followers. But victory is far from won: closed-source companies still hold more resources, more data, more talent, and the policy balance may not favor open source.
Yet history teaches that once Pandora's box of knowledge is opened, it is hard to close again. Heliocentrism was ultimately accepted not because the church agreed, but because the evidence was too obvious and the logic too clear to deny.
The dawn of open source may already be here. The question that follows: how do we greet this new sun?
---
> "Open source is not a choice but an inevitability. Because freedom is the direction of technological evolution."