The Great Retreat: Why Erdosproblems.com is Closing the Gates on AI Proof-Mining

The Abattoir Effect: When Mathematical Discourse Becomes AI Slop

Erdosproblems.com, once a vibrant digital agora for mathematicians, has effectively shuttered its doors to the public. The site, which served as a critical barometer for the intersection of human curiosity and machine intelligence, has been overwhelmed by a wave of unverified, AI-generated proofs. Much like the internal friction at major labs where the culture is broken, the Erdos community is fracturing under the weight of automated output.

Founder Thomas Bloom has taken the drastic step of freezing all new comments and proof claims. His reasoning is stark and uncompromising, highlighting the fundamental incompatibility between collaborative human inquiry and the transactional nature of AI-driven 'proof-mining.' As Bloom noted, 'Just as one does not open a restaurant in an abattoir, it is important that there be a separation between such repositories and a site which aims to promote the actual questions.'

The Death of the 'Open' Problem: Why Status Labels Are Now Meaningless

To combat the influx of 'glory-seeking' users who treat mathematical discovery as a gamified dopamine loop, the site has undergone a radical structural purge. The removal of 'solved' versus 'open' status labels is a defensive maneuver designed to strip away the incentive for automated, low-context submissions. By neutralizing the visual markers of progress, the site seeks to return to a state of pure inquiry.

Key Site Changes:

• Hiatus on comments: A total freeze on new community-driven proof claims.
• Removal of problem status: All problems are now displayed with neutral, non-hierarchical labeling.
• Shift to high-quality exposition: A new focus on curated, human-verified explanations rather than raw results.
• Rejection of credit-giving language: Moving away from ownership-based metrics to discourage the 'priority claim' culture.

Signal Decay in the Age of Automated Priority Claims

The mathematical community is now facing a backlink deficit as high-quality human discourse is drowned out by automated, low-value proof submissions. This phenomenon is not merely a nuisance; it is a fundamental degradation of the intellectual commons. When machine-generated noise replaces human-led iteration, the 'backlink' of shared understanding is severed, leaving researchers to navigate a landscape of disconnected, context-poor data.

Reclaiming the 'Brain is Open' Ethos in a Synthetic World

The closure of Erdosproblems.com is a warning shot for the future of digital knowledge. We are witnessing a transition where the 'open' internet is being cannibalized by agents that prioritize speed over substance. The challenge for the next generation of platforms is to build environments that are resilient to this 'abattoir effect'—spaces where the human element is not just a participant, but the primary filter for value.

True mathematical progress, as Paul Erdős practiced it, was a social, messy, and deeply human endeavor. It required the ability to be stuck, to be confused, and to find a flow state through shared struggle. If we allow our intellectual repositories to become mere dumping grounds for AI-generated tokens, we risk losing the very curiosity that drives discovery. The future of mathematics will likely reside in gated, high-signal enclaves where the 'brain is open'—but only to those who are willing to engage with the complexity of the problem, rather than just the output of the machine.