The Habr platform enforces a ban on AI-generated content under point 4 of its rules, which was implemented approximately three months ago. The author of the article notes that this policy leads to the blocking of users who utilize AI tools for editing or proofreading, regardless of whether the content was originally written by a human.
The author argues that the current moderation approach fails to distinguish between low-quality, automated 'slop' and the legitimate use of AI as a professional tool for structure and fact-checking. This creates an asymmetry of effort where moderators, unable to verify content in detail, resort to banning based on suspicion rather than actual quality assessment.
Beyond the platform's internal policies, the author discusses the efficiency of AI models in programming tasks. Specifically, the GPT-5.6 Luna (max) model is highlighted for its cost-effectiveness, with a 'Cost per Task' of $0.18 compared to $3.26 for the GPT-6 Astra (max) model. The Artificial Analysis Intelligence Index for these models is 38 and 53, respectively.
