No GitHub, No Read: What HN's September Hiring Manager Admitted

In the September 2026 HN "Who is hiring?" thread, a hiring manager posting as html5cat told applicant boltzmann-brain that his application was "definitely well written but didn't have GitHub so it got skipped." The framing was explicit: GitHub is the defense against "the sea of AI slop." That admission, said out loud in a public thread, is the clearest signal yet that inbound screening in 2026 has quietly stopped grading prose and started grading proof of humanity.

What actually happened in the HN thread

A hiring manager in the September 2026 HN "Who is hiring?" thread publicly skipped a well-written application because the candidate had no GitHub link, and named AI application slop as the reason. That is the entire mechanism in one sentence.

The exchange sat inside HN thread id 49522897, whose posting rules still require that you personally work at the hiring company, post once, and commit to replying to every applicant. html5cat was trying to honor that last rule ("my goal is to respond to every earnestly submitted application despite the sea of AI slop") and used GitHub presence as the tiebreaker for who got a human read.

Two things are worth noticing:

• The rejected application was not bad. html5cat called it "definitely well written." Under 2023 rules, it would have moved forward.
• The filter is not about skill. It is about whether a person or a model wrote the email. A commit history from 2022 is hard for a model to fabricate on the fly; a paragraph of prose is not.

Other posts in the same thread signaled the same thing without saying so. Close.io opened with "Our code: https://github.com/closeio" and listed TaskTiger, SocketShark, and ciso8601 as public artifacts. Lumen Research wrote, "If you can link to a time you got a trained policy running on real hardware, you go to the top of the pile." NetBird in Berlin led with its open-source posture. The thread is turning into an artifact-first channel, and the hiring managers are the ones enforcing it.

Why the "well-written resume" era just ended

The well-written resume stopped being a positive signal the moment writing one became free. Screeners in 2026 are pricing that in.

The volume side is the easier half to see. LinkedIn is reporting roughly 11,000 job applications per minute, a 45% year-on-year increase, while job postings dropped 10.6% over the same period. That works out to about 183 applications per second landing on a shrinking set of reqs. The average corporate opening now pulls 242 applications, a 0.4% success rate per applicant.

11,000

The quality side is the interesting half. Robert Half's March 10, 2026 release, polling 2,000 hiring managers in November 2025, found:

• 84% of HR teams report heavier workloads from AI-tailored applications
• 65% of hiring managers say AI-enhanced resumes make skills harder to verify
• 67% of U.S. HR leaders say reviewing AI-generated applications has slowed hiring
• 20% report delays of more than two weeks
• 67% now lean on staffing firms; 89% say those partners have been effective

Robert Half's Dawn Fay put the mechanism plainly: "Companies are looking to hire, but a surge in unverified applications is extending timelines and delaying critical work." The Willo Hiring Trends Report 2026, drawn from 100+ hiring pros and 2.5M candidate interviews, backs it up: only 37% of employers view resume-style credentials as reliable talent indicators, 41% are actively moving away from resume-first hiring, and 77% of hiring teams regularly encounter AI-generated or AI-assisted applications.

Read the two datasets together and the conclusion is uncomfortable: inbound volume is up 45%, and the trust the market places in that volume is down to about a third. Hiring managers are not being lazy when they skip a polished email with no GitHub. They are refusing to spend the next 40 minutes trying to figure out if a human wrote it.

GitHub as proof of humanity, not proof of skill

GitHub is being used in 2026 as a proof-of-humanity signal, not a skills signal. The distinction matters because it changes who you should be sourcing.

html5cat did not claim boltzmann-brain's code was better than the accepted candidates. The filter ran before any code got read. The logic is closer to a captcha than a technical screen: a three-year commit graph, a maintained side project, a linked PR to a known repo, these are things a language model cannot retroactively fabricate in the ten seconds it takes to generate a cover letter. Robert Half flagged this explicitly, noting generative AI tools are "fabricating or embellishing work history and skills." A GitHub link inoculates against that specific failure mode.

A commit graph is a captcha with a three-year cooldown. That is what hiring managers are actually buying.

There are two second-order effects sourcers should absorb:

1. The bar to pass the filter is low. You do not need a 5,000-star repo. A visible, dated, plausibly-yours account is enough to move out of the AI-slop bucket.
2. The bar to find candidates who clear it is high. Most engineers who commit publicly do not advertise it on their LinkedIn headline. The ones who do are a rounding error.

That second point is where sourcing strategy actually breaks.

How many engineers actually surface GitHub on their profile

Only about 0.12% of U.S. software engineers surface GitHub in their profile headline, which means waiting for GitHub-visible candidates to apply is fishing a pond of a few hundred people. The real population is larger, but you have to reach it through code, not through profiles.

Here is the arithmetic straight from the Refolk index.

Two things jump out. First, the GitHub-visible LinkedIn population is 625 people out of half a million. Second, when you look at the top employers inside those 625, GitHub itself accounts for 10 of the top self-signalers, with Microsoft next at 3. The people who put "GitHub" in their headline mostly work at GitHub. That is not the population html5cat wants. That is a rounding error dominated by one employer's brand affinity.

The engineers who actually clear the AI-slop filter, the ones with a real commit trail on a real repo, are almost never the same people who advertise it in a LinkedIn headline. They are discoverable through their code, their PRs, their maintainer status on a specific library. You cannot filter a resume database for them. You have to search the artifact.

That is the exact gap Refolk closes: describe the person in plain English ("senior backend engineers in the US who maintain Python packages with more than 200 stars, not currently at FAANG") and get a ranked shortlist pulled across GitHub, LinkedIn, and the open web, not just whichever headline mentioned the word "github."

Outbound vs inbound recruiting, repriced for 2026

Outbound is now cheaper per verified hire than inbound, because inbound's per-application review cost has quietly exploded while outbound's cost per contact has not moved. That is the reprice most teams have not made yet.

Do the math with the numbers the market has already published:

• Inbound: 242 applications per opening, 77% of which look AI-assisted, 20% of hiring teams delayed 2+ weeks. Even at 30 seconds per resume, one req burns two hours before any human conversation happens, and the yield is a fraction of a percent.
• Outbound to a GitHub-verified cohort: you start with candidates who have already cleared html5cat's filter. The screening cost is front-loaded into the search query, not the review pile.

The Willo finding that 41% of employers are "actively moving away from resume-first hiring" is the same trend from the buyer's side. So is the Robert Half data that 67% of teams are now paying staffing firms and 89% call them effective. The market is voting with its budget: people will pay a premium for candidates whose humanity and skill are already verified before contact.

The founders posting on HN "Who is hiring" have figured this out and hacked their way to it manually. Close.io, Lumen Research, and NetBird are writing job copy that pre-filters for GitHub-visible engineers, because they know the applicants who self-select in have already crossed the artifact line. The scaled version of that instinct is to run the same filter proactively, on the whole open-source graph, rather than hoping the right 45 senior open-source contributors happen to see your HN post this month.

What to change on Monday

Rebuild your sourcing top-of-funnel around three moves that treat GitHub as a signal rather than a keyword. That is the practical version of the argument.

1. Stop keyword-matching "github" in LinkedIn headlines. You are fishing 625 people, most of whom work at GitHub. It is a bad pond.
2. Source from the artifact directly. Query the commit graph, filter for maintainers of libraries in your stack, then reverse-resolve to LinkedIn and email.
3. Rewrite your outbound opener to name the artifact. "Saw your work on <repo>, specifically the <PR> from March" is a first line no language model can bulk-generate against your prospect list. It is also the same signal html5cat is looking for, just delivered by you instead of extracted from an application.

For inbound that you cannot kill, borrow the HN thread's design. Require a link. Make the field mandatory. Read the linked artifacts first and the prose second. You will process the same volume in a fraction of the time and produce a shortlist that would survive an html5cat skim.

The hiring manager on HN was blunt about a filter most teams are running silently. The winning move is not to argue with the filter. It is to source the population the filter selects for, before that population ever sees your job post.

FAQ

Is skipping applications without a GitHub link legally risky?

Skipping applications for lacking a specific voluntary artifact is generally lower-risk than screening on protected characteristics, but it is not risk-free if the criterion produces disparate impact on a protected group. The safer path most teams are taking is to use GitHub as a positive signal that moves candidates up the pile, not as a hard reject. That is closer to what html5cat actually did: the pile was too big to read fully, so artifacts got read first. Talk to counsel before writing "no GitHub, auto-reject" into your ATS logic.

Does this only apply to software engineers?

No. The mechanism generalizes to any role where the candidate can produce a public, dated artifact that is hard to fabricate. For designers it is Dribbble, Figma community files, or shipped product screens with attribution. For writers it is bylines. For data scientists it is Kaggle placements or published notebooks. Any role where a language model can generate a plausible resume but not a plausible three-year public trail is subject to the same shift.

How do I write an outbound message that does not itself look like AI slop?

Name a specific artifact in the first line and reference something only a human reader of that artifact would notice. "Saw your rewrite of the auth middleware in <repo> last April; the decision to drop the session cache was the interesting one" is a message a language model cannot cheaply generate at list scale. Generic praise ("impressive work on your open source contributions") reads as slop even when a human writes it. The specificity is the signal.

Where does HN "Who is hiring" fit in a 2026 sourcing stack?

Treat the HN thread as a monthly market-read, not a channel. The posts tell you which companies are hiring, at what seniority, and increasingly what artifacts they are gating on. That is useful intelligence for outbound. As an inbound channel, HN produces a small volume of high-signal candidates precisely because it forces plain-text email, no ATS, and rewards linkable work. Founders running seed-stage teams should post. Series-B and later teams should read the thread and source against the companies posting in it.