New Grad Jobs 2027: the listing is a lead funnel with a table on top

Every apply link routes through the vendor

The Apply column is the reason to read this repository closely. No cell points at an employer's own application page. Each one is a zapply.jobs link carrying a tracking parameter: the Capital One row points at `https://zapply.jobs/l/d/workday-capitalone-capital-one-R1001200?s=gh-new-grad-jobs-2027`, and the OpenAI rows point at ashby URLs on the same host. The path also records where the posting was collected from, because workday, ashby, greenhouse, apple, amazon, google and an sr-ServiceNow prefix each appear as the leading token. The README spends its own opening section on that same destination, advertising a Chrome extension that autofills applications, a dedicated job board, accounts with multiple resume profiles and tracking with streaks and commitment awards, and the first link in it carries `?ref=gh-newgrad-jobs`.

A ten minute refresh with no generator in the tree

The page states the roles are updated every 10 minutes. Nothing in the repository could do that. The tree holds .gitattributes, .github/, CODE_OF_CONDUCT.md, CONTRIBUTING-GUIDE.md, CONTRIBUTORS.md, LICENSE, README.md, images/, index.html and llms.txt: no data file, no script, no scheduled workflow, no fetch configuration. The published site is a GitHub Pages site at zapplyjobs.github.io/New-Grad-Jobs-2027/, so the same content the README shows is what a reader gets on the web, and the repository has no GitHub releases, so there is no published snapshot of a listing to compare a later one against. Whatever refreshes the table every ten minutes runs somewhere outside this repository.

Posted is an age in hours, and one row has no date

The Posted column counts upward in hours, and the visible Software Engineering rows read 1h, 2h, 8h, 9h, 10h, 11h, 12h, 13h, 14h, 15h and 16h. One row breaks the pattern: the Google Software Engineer III, App Ads Horizontal Modeling entry, located in the United States, reads Date unknown, and it is the only row in view with no age at all. Nothing in the table carries a calendar date, so the freshness of a listing cannot be recovered from the file, and comparing two commits shows that a row once said 14h without recording when that was true. Relative ages are what a board wants and what an archive cannot use.

Employers arrive as the legal entity on the form

Company names are not normalized. One employer appears three times under three legal names: Amazon Web Services, Inc. for a DynamoDB Transactions role in Seattle, Amazon.com Services LLC for a Software Development Engineer (SDE2) role in Seattle, and Amazon Kuiper Manufacturing Enterprises LLC for a Kuiper network orchestration role in Redmond. Johnson & Johnson appears under a link path of `workday-jj-jj-R-101018`, and the two Anduril rows resolve to `greenhouse-andurilindustries`. Counting distinct employers means counting strings, and the string a candidate searches on is the legal entity rather than the brand.

The same title appears three times under three job ids

Capital One's Full Stack Engineer 4 fills three rows with three different identifiers, R1001200, R1001376 and R1001156, two of them in Richmond, Virginia and one in McLean. Microsoft's two entries differ by a single word, Cloud Engineering Consultant CTJ TS/SCI and DevOps Cloud Engineering Consultant CTJ TS/SCI, both posted 2h in Springfield, Virginia. Cisco lists Software Engineer twice, in Milpitas and in San Francisco, both at 10h, and Apple lists a Teamcenter Site Reliability Engineer role in Austin next to a Core Audio Software Engineer role in Cupertino. Duplicate detection on this list keys on the link, not on the title or the employer.

A large share of the visible rows ask for clearance

Several entries in a new grad table are cleared work. GDIT posts Software Engineer TS/SCI with Polygraph and Cloud Software Engineer TS/SCI with Poly, both in Herndon, Virginia, both at 10h, and both marked as sponsor. Microsoft's two consultant rows carry CTJ and TS/SCI in the title, Cisco has Site Reliability Engineer FedRAMP in RTP, North Carolina, and GE Vernova has an Application Engineer (Technical Sales) role on Defense in Findlay Township. A polygraph requirement is not a step an application form can express, so the title carries the entire constraint. Alongside them sit two Anduril radar roles in Fort Collins and a Build Reliability Engineer II at Varda Space in El Segundo.

Seniority labels and location strings pass through unedited

The page calls these new grad, intern/co-op and adjacent early-career roles, and the visible table agrees in the way that matters: Software Engineer 2 at Amentum in Baltimore, Full Stack Engineer 4 at Capital One, Software Engineer III at Google, Build Reliability Engineer II at Varda Space and Robotics Software Engineer I/II at Agility Robotics all sit under the Software Engineering heading. The Location column shows the same pass-through. Values read USA VA Herndon, RTP North Carolina US, VA-Dahlgren, MD-Baltimore and Onsite- Salem, OR, and several are cut off mid-word with United... or Shepherdsville, Kentucky,... so a city search on this list will miss rows whose location string was truncated upstream.

License metadata and the LICENSE file disagree

The repository metadata reports the license as NOASSERTION, which is how GitHub records a license it could not classify, while the tree carries a LICENSE file at its root and the README names no license anywhere. Two files are in play and neither settles the other: nothing published here says which text applies to the listings, to the page, or to the images in images/. For a repository whose tables reproduce other employers' role titles, locations and identifiers, that gap is worth closing before anything here is copied elsewhere.

Editorial conclusion

Treat New Grad Jobs 2027 as a lead funnel rather than a dataset, and read the table as an index of who is hiring rather than as a source of record. It earns a look if you want a fast view of employers hiring early-career roles this week, and it is the wrong foundation for a job tracker, because there is no data file to track, every apply click goes through zapply.jobs, and the freshness you see is a relative age with at least one row reading Date unknown. Before you rely on it, check three things: that the employer, title and location strings match what the employer's own careers page says, whether the sponsorship column is filled in or simply empty, and what the LICENSE file actually says when the repository metadata reports NOASSERTION.

Frequently asked questions

Official sources

1. Official documentation
2. Official README
3. Project repository