Andrew Marble stated that while adopting Linux once carried significant risks, that gap has narrowed due to the proliferation of web applications.
However, he points out that clear penalties still remain for open LLM users. As of 2026-06-21, Claude and GPT maintain the top positions on the Artificial Analysis leaderboard.
Furthermore, he noted that while the APIs of the two major companies excel in compatibility and reliability, the distribution channels for open models still leave concerns regarding privacy.
Self-hosting solves privacy issues but is disadvantageous in terms of cost, complexity, and speed. Marble analyzes that the use of open models, which had previously been largely a hobby, is beginning to change due to the widespread adoption of ID verification for Claude.
He believes that open models are currently trailing the leaders by only a few months and that the gap is not as wide as the difference between Linux and Windows in 2008. While he expects a short-term decline in productivity, he expressed the view that it will not become a decisive problem.
Sources: There is minimal downside to switching to open models (HN 407pt, 330 comments) (HN Search (backfill), 2026-06-22)