216 Applications, 10 Offers: The Ceiling on HN's Best Inbound

One employer at the top of the September 2026 "Ask HN: Who Is Hiring?" thread did something almost nobody does in public: they disclosed their inbound funnel. Their August 2026 post pulled 216 applications and produced roughly 10 offers - a 4.6% application-to-offer rate on arguably the highest-signal engineer inbound channel on the internet, and it is still not enough.

If HN Who Is Hiring is 8x better than the industry baseline and it still caps a specialty hire at single digits per month, the answer is not "post harder." The answer is that inbound has a mathematical ceiling, and every founder relying on it is quietly hitting it.

What the September 2026 HN thread actually disclosed

The disclosure, verbatim from the employer's post: "We received 216 applications in our August posting, and issued offers to approximately 10 (still following up with a few)." That is a 4.6% offer rate, roughly 8x the industry baseline.

The role scope matters for the rest of this analysis. The employer is hiring senior engineers to design reinforcement learning benchmarks, red-team frontier model outputs, and build rubrics that evaluate agentic coding capability, alongside more standard Rails, TypeScript, and Python full-stack tiers. It is a Ruby-and-RL contract engineering shop, and it sits at the top of the September 2026 thread (item 49522897 on news.ycombinator.com), which drew about 284 comments in its first days.

The rules the thread enforces:

• Only companies that are hiring may post
• One post per company per month
• Post must explain what the company does
• Poster must be actively filling and commit to replying

Those rules are the reason HN Who Is Hiring converts so well. They are also the reason it does not scale.

The 4.6% number in industry context

A 4.6% application-to-offer rate is roughly 8x better than the cross-industry baseline of about 0.5 to 0.6%, which works out to one offer per 180 to 200 applications. HN is not a normal channel. It is the best inbound channel most founders will ever touch.

The funnel math the disclosure sits on top of:

• Industry-wide, only about 0.5% of applicants receive job offers
• Tech specifically requires 191 applicants per hire; healthcare requires 47
• Average applicants per role climbed from 46 in 2021 to 95 in 2025
• Sourced and referred candidates convert 4 to 10x better than job-board applicants, despite being only about 7% of total applications
• 97% of applicants are eliminated before they speak with a human
• Ashby's 2025 data pegs technical interview-to-offer at 7%, vs. 9% for business roles

4.6%

The mechanism behind HN's outperformance is self-selection. Applicants read a paragraph of dense technical copy, decide they match, and only then apply. That pre-filters for taste. It is also why the offer rate falls apart the moment you try to reproduce it on Indeed or LinkedIn Easy Apply, where the friction to click is near zero.

Why 10 offers is the ceiling, not the floor

Ten offers per monthly post is the ceiling of this channel, not a starting point you can multiply. HN publishes one Who Is Hiring thread per month, allows one post per company, and top-of-thread placement is effectively a lottery based on when the thread goes live.

The disclosing employer also said "issued offers to approximately 10," not "hired 10." Standard offer-accept sits around 82%. That means 10 offers probably yields 7 to 8 accepts, and startup 12-month attrition eats more from there. The realistic durable-hire yield from that 216-applicant pond is 5 to 7 people.

Structural limits on scaling HN inbound:

1. One post per month means a maximum of 12 posts per year per company
2. Top-of-thread comment position drives most of the traffic, and it is first-come based on the thread's launch minute
3. Community-run indexes like hnhiring.com and nchelluri.github.io/hnjobs surface the post afterward, but by then the initial burst is over
4. The thread ranks on HN's front page for roughly 24 hours

If you need to hire 30 engineers a year, HN can plausibly cover 5 to 7 of them. The other 23 have to come from somewhere else.

The pool HN is actually reaching

HN Who Is Hiring reaches a small single-digit percentage of any given specialty pool at best. For the disclosing employer's stack (US software engineers with either Ruby on Rails or Reinforcement Learning as a listed skill), Refolk's index shows the ceiling clearly.

The 216 applicants represent 1.35% of the addressable Rails population and 15.9% of the RL population, and those are ceilings before any qualification filter. In practice most of the 216 were adjacent-skill hopefuls, not the 1,358 actual RL practitioners, because the actual RL practitioners are vested at Meta, Google, Waymo, Google DeepMind, and Applied Intuition. They do not scroll HN comments for jobs.

A concentration signal from Refolk's index sharpens this: 8 of 25 sampled US RL engineers sit in the SF Bay Area (about 32%), while Rails engineers skew toward NYC, Seattle, and Denver. An HN post reaches the Rails pool broadly and misses most of the RL pool geographically at the same time.

The self-selection paradox

Self-selection is HN's superpower and its ceiling in one mechanic. It filters for people who read technical prose and only apply when they match, which is why the offer rate is 8x baseline. It also filters for people who are willing to cold-apply, which excludes almost everyone who is senior, employed, and specialized.

Consider who is not in the 216:

• Staff engineers at Waymo working on RL for autonomy stacks
• Google DeepMind researchers with two years vested equity
• Meta engineers who joined post-IPO and refresh grants every year
• Applied Intuition engineers with signing bonuses on clawback

None of those people are trolling HN comment threads looking for a Rails-and-RL contract shop. They are the people the disclosing employer actually wants, and inbound will never reach them. A Rails post can plausibly attract representative applicants because the pool is 16,000+; an RL post attracting 216 people is mostly attracting people adjacent to RL, not the 1,358 who ship it.

Inbound volume is inversely correlated with skill rarity signal. The scarcer the person, the less the applicant count means.

What the 46-to-95 applicant trend does to founders

Applicant volume nearly doubled between 2021 and 2025, from 46 to 95 per open role, which mechanically drags every downstream conversion rate down. A founder who benchmarked their pipeline in 2021 is now doing roughly 2x the screening work for the same hire and does not realize it.

The knock-on effects:

• Time-to-first-response gets slower, which pushes acceptance rates down
• Screening cost per hire roughly doubles even if screener wages hold flat
• 97% of applicants get eliminated pre-human, so ATS filters get more aggressive and false negatives climb
• Founders start ignoring inbound piles entirely, which feeds a "no one good applies" narrative

This is the mechanism behind the "underqualified applicants" warnings on the major boards. Volume went up, quality per applicant went down, and the same absolute number of qualified people are now buried under twice the noise.

The sourcing split the math forces

If HN's 4.6% offer rate is the ceiling of inbound and it still yields only 5 to 7 durable hires per monthly post, the math forces a specific split: run HN as your best inbound channel, and run outbound against a named-employer list for everything else.

A realistic allocation for a founder hiring 20 engineers this year:

Outbound stops being optional the moment your role has a rare skill (RL, Solidity, Rust systems, formal verification, GPU kernels). At those pool sizes, inbound cannot statistically reach the people you want, because the people you want are employed at four companies and are not job-hunting.

The practical outbound workflow that fits a founder's calendar:

1. Name the 4 to 8 employers who actually ship the skill you need
2. List the specific teams inside those employers, not just the company
3. Pull profiles from those teams into a shortlist
4. Send the first message from the founder's account, not a recruiter's
5. Reserve HN Who Is Hiring for the broader Rails, TS, Python tiers where the pool is 16,000+

Step 3 is the one that used to eat weekends. Describing "senior engineers on the Waymo Driver behavior team who have shipped RL-based planners" in plain English to Refolk returns a ranked list across GitHub, LinkedIn, and the open web, which is faster than boolean-hunting on LinkedIn Recruiter and cheaper than a retained search.

What the other companies in the September 2026 thread are missing

Every other employer in the September 2026 thread (Fastly, NetBird, Silkline, Squoosh.ai, "The Flywheel," and roughly 100 others) is fishing in the same 216-applicant pond, and none of them will publish their conversion math. That silence is the story.

The disclosing employer did the field a favor by putting real numbers on a channel that founders otherwise mythologize. Takeaways for anyone else posting in this month's thread or the next one:

• Treat HN as one of many channels, not the answer
• Assume your absolute yield is 5 to 10 offers per monthly post if your copy is strong
• Assume your yield is 0 to 2 if your role requires a rare skill, because the specialists are not there
• Build the outbound pipeline in parallel, starting from the named-employer list for the skill you need
• Do not benchmark against 2021 conversion rates; the funnel is twice as wide and just as narrow at the bottom

FAQ

Is HN Who Is Hiring worth posting to as a founder?

Yes, if your role has a broad pool and your company is technically legible in a paragraph. A 4.6% application-to-offer rate on 216 applications is genuinely 8x industry baseline, and the effort is a few hours per month. It stops being worth it if you need more than 5 to 10 hires per year or if your role requires a skill held by fewer than a couple thousand people, because at that point the ceiling of one post per month binds hard.

What is a realistic application-to-offer conversion rate for tech roles?

Industry data puts tech-specific applicants-per-hire at 191, which implies roughly a 0.5% application-to-offer rate. Cross-industry averages have gotten worse since 2021, moving from 46 to 95 applicants per role, and 97% of applicants are eliminated before speaking with a human. HN's 4.6% is an outlier driven by self-selection, and any channel with lower friction to click will regress hard toward the 0.5% baseline.

How many applicants per hire should I plan for?

Plan for 95 applicants per hire on cross-industry averages, 191 for tech roles, and roughly 22 for an HN Who Is Hiring post if you can reproduce the disclosed 4.6% offer rate. Referrals and sourced candidates convert 4 to 10x better than inbound, so a strong referral pipeline can drop your applicants-per-hire into the low double digits, and outbound sourcing lets you skip the applicant funnel entirely for the roles where inbound math does not work.

When should a startup switch from inbound to outbound recruiting?

Switch the moment your role requires a skill with a US pool smaller than roughly 5,000 people, or the moment you need more than 10 hires per year in one specialty. Below either threshold, inbound produces enough qualified volume. Above them, the people you want are employed at a small number of named companies, are not job-hunting, and will only be reached by direct outreach from a founder or a sourcer working from a named-employer list.