OpenAI Model Disproves Erdős Unit-Distance Conjecture, Verified by Mathematicians

·1 views

OpenAI announced on May 20, 2026, that an internal AI model generated a proof disproving the Erdős unit-distance conjecture, a decades-old open problem in discrete geometry. The conjecture concerned the maximum number of unit-distance pairs possible within a set of points in the plane. The model produced a construction demonstrating that, for infinitely many values of n, point sets can contain far more unit-distance pairs than the conjecture had implied. External mathematicians independently verified the proof's key steps, and a companion document was prepared to provide historical context and a human-readable exposition of the argument. OpenAI described the result as a concrete example of AI contributing to frontier mathematical research, distinct from benchmark performance or routine symbolic assistance.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

77% of Firms Hit by Software Supply Chain Attacks in Past Year, Omdia Finds

A Docker-sponsored Omdia survey published on August 4 found that 77 percent of organizations experienced a software supply chain security incident in the preceding twelve months. Exploits targeting known vulnerabilities in third-party software remained the leading attack type at 38 percent, while AI has emerged as the top-ranked supply chain risk, cited by 40 percent of respondents. Reported consequences included unauthorized access to applications and data (46 percent), SLA disruptions during remediation (37 percent), and theft of developer credentials or secrets (35 percent). Omdia projects that third-party code will account for 58 percent of organizational codebases within a year, up from a current majority share reported by 38 percent of respondents. Sixty-two percent of teams anticipate increased supply-chain security spending in the next budget cycle, while 45 percent do not currently rate their own supply chain security as adequate.

0 comments Read more at DEV Community

ProgrammingHacker News ·

Ex-Deloitte Auditor Open-Sources Full SOC 2 Methodology for AI Systems

A former Deloitte auditor has publicly released a complete SOC 2 audit methodology on GitHub under the Chiaro-HQ repository. The resource is aimed at helping teams apply structured compliance frameworks to AI systems. The release was shared on Hacker News, where it received minimal early engagement. SOC 2 is a widely recognized auditing standard focused on security, availability, and data privacy controls. The open-source approach is intended to make professional audit methodology more accessible to developers and organizations building AI products.

0 comments Read more at Hacker News

ProgrammingDEV Community ·

Developer Recovers Tour Booking Site After Supabase Permanently Deleted Inactive Project

A developer built Kenya Tours, a Next.js and Supabase-powered tour booking site for the Kenyan coast, but neglected it after starting a demanding new job. Supabase's free tier pauses and eventually permanently deletes inactive projects, and the developer returned to find the project gone well beyond the 90-day restore window. Recovery required creating a brand-new Supabase project and manually rebuilding it using a downloaded pg_dump backup for the database schema and a separate zip archive for stored images. The process surfaced several technical hurdles, including Windows-generated metadata files blocking CLI migrations and the absence of a CLI command to run SQL files against a remote database. After re-uploading images, reconnecting auth providers, and resetting environment variables, the site was fully restored with tours and bookings functioning again.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Why Domain Expertise Makes You Better at Using AI Language Models

A software developer's hands-on experience with GPT-4 reveals that the quality of output from large language models is closely tied to the user's subject-matter knowledge. Vague prompts tend to produce generic or inaccurate results, while detailed, context-rich instructions yield more precise and useful responses. The author found that framing requests with technical specificity — such as describing exact component behavior in React — dramatically improved code quality. However, LLMs are not infallible; the developer encountered cases where the model produced entirely off-target outputs, underscoring the need to validate AI-generated content rigorously. The key takeaway is that LLMs work best as collaborative tools when guided by users who already possess a strong foundation in the relevant domain.

0 comments Read more at DEV Community