Live·Open questions in longevity research
← All dispatches
Ukhvat

64 pull requests, 19 projects, and the maintainer problem

5 July 2026· CY2gg6hG

In its first week, OpenAI's Patch the Planet produced 64 pull requests across 19 open-source projects, but the harder problem is getting patches through the full path: accepted by maintainers, tested, released, and installed by users.

OpenAI's Patch the Planet, a program with Trail of Bits (cybersecurity), HackerOne (bug bounty), researchers, and open-source maintainers, produced 64 pull requests and 51 issues across 19 projects in its first week, including cURL, Go, Python, Sigstore, and pyca/cryptography.

Since March, Codex Security, OpenAI's automated vulnerability scanning tool, has scanned more than 30 million commits across more than 30,000 codebases. People marked more than 70,000 findings as fixed; the system identified more than 500,000 more as fixed automatically. But finding a vulnerability is half the problem.

Linux Foundation and Harvard research puts the bottleneck in numbers: in 94% of widely used open-source projects, fewer than ten developers write more than 90% of the code in a year. An auto-generated pull request solves nothing by itself. A maintainer has to accept it, the tests have to pass, a release has to ship, and users have to update.

GPT-5.5-Cyber scored 85.6% on CyberGym (a benchmark for reproducing known vulnerabilities in software environments) versus 81.8% for standard GPT-5.5. On ExploitGym, where the task is turning a known flaw into a working attack path, the gap widens: 39.5% versus 25.95%. Models are approaching the dangerous boundary faster, and OpenAI limits GPT-5.5-Cyber to vetted organizations, with monitoring, restrictions, and human review.

Researchers filter model output before it reaches maintainers. Some of the work: tests, pre-merge checks, fixes in a big-integer library, Python supply-chain improvements. The real test: can a fix complete the path from a discovered bug to an update that actually reaches users.

Open the related Eternal Search page

Sources
openai.com
t.me
blog.trailofbits.com

Follow the threadOpen source page
Why this was published

Patch the Planet directly targets the libraries (cURL, Go, Python, pyca/cryptography) that health-tech startups in MATTER's network depend on as core infrastructure. The news provides concrete first-week metrics that support a specific claim about supply-chain risk for health-tech ecosystems, linking to MATTER's Eternal Search profile as a hub whose portfolio companies inherit every unpatched open-source dependency in those libraries.