Most of the reaction accepted the broad strategic direction while rejecting the neatness of the analogy. People kept coming back to one simple distinction: Kubernetes spread because lots of orgs could run it on ordinary infrastructure and build on it together, while frontier open-weight models still depend on scarce memory, GPUs, and model-serving expertise. That does not make open weights unimportant. It does make them less like Linux or Kubernetes than the article suggests. Several commenters also argued that “open weight” is not “open source.” You get portability and inspectability, but not the full collaborative development loop that made open software platforms self-sustaining.
The strongest thread was about market structure. Open models were praised for putting a floor under
token pricing, giving users version stability when
API vendors yank or degrade models, and making startup economics work for high-token products that would be crushed by frontier API bills. At the same time, people noted that current token prices are not a clean reflection of cost. They are shaped by subsidy, market testing, inference breakthroughs like
FlashAttention,
quantization,
distillation, and the fact that labs are still searching for profitable pricing. That made many readers see open weights less as ideology and more as bargaining power.
The other major thread was regulation and geopolitics. A lot of commenters thought a ban on “Chinese models” is technically incoherent because weights are just arrays of numbers that can be mirrored, lightly modified, distilled, or republished through third parties. The practical effect of any US action would be to chill corporate adoption, not to stop the weights from circulating. That is why many people read the policy push as a path to restricting open models broadly and entrenching closed American incumbents. Others thought the national security concern is real, but concluded the answer is not a ban. It is to fund credible non-Chinese open alternatives in the US and Europe, because right now too much of the open-weight ecosystem depends on Chinese labs. The mood was pro-open overall, but stripped of romance. Open weights look strategically inevitable, commercially useful, and politically contested all at once.
tobi.knaup.me