170 Applicants, 3 Offers: HN "Who Is Hiring" Broke in August 2026

The August 2026 "Who is Hiring" thread on Hacker News went up on Aug 1, and buried in the replies is the sentence that ends the debate: a founder at Prophet Town LLC reports 170 applicants from July converted to 3 offers. That is 1.76%, on the channel dev-friendly founders have defended for a decade as the last high-signal inbound funnel. Meanwhile, job-board operators are posting scraped ATS roll-ups directly into the thread, and nobody is downvoting them off.

The 1.76% number that killed the "HN is different" argument

Hacker News' hiring thread now converts like a generic job board, and the founders posting there know it. Prophet Town LLC's self-report in the August 2026 thread is the cleanest data point yet: "We got 170 responses to our July HN Who's Hiring, and made 3 offers." That is a 1.76% applicant-to-offer rate on a channel that has, for years, been the exception cited when someone argues that inbound still works for technical roles.

To be fair to HN, 1.76% is roughly 3.5x better than the industry inbound benchmark. Gem's 2026 report on 165 million applications puts generic inbound at 0.5% applicant-to-offer, or about 1 hire per 200 applications. So HN is still, technically, "above average." The problem is that the ceiling has moved. Outbound-sourced candidates convert at 6% in the same Gem dataset, which makes them roughly 12x more likely than generic inbound to result in a hire, and about 3.4x more likely than even Prophet Town's above-average HN experience.

1.76%

The read is not "HN is broken and LinkedIn is fine." The read is that even the best remaining inbound channel now sits below the outbound floor. If you are still calibrating channel choice by which inbound is least bad, you are optimizing the wrong axis.

What actually happened to the thread's signal

The signal collapsed because the thread's social contract collapsed. The August 2026 thread still opens with the same instruction it has carried for years: "Please only post if you personally are part of the hiring company - no recruiting firms or job boards." That rule was the entire mechanism. It kept the thread a direct channel between founders and candidates, which is why the responses used to be senior and self-selected.

Then, in the same August thread, an operator posted this: "I run CareerJumpShip.com, a free job board that surfaces active roles from Greenhouse / Lever / Ashby ATSes." A roll-up of scraped ATS listings, sitting in the thread, upvoted, not flagged into oblivion. That is the visible moment the norm died. Dead norms do not come back on volunteer moderation.

The candidate side degraded in parallel. Three specific mechanics are doing the damage:

• Auto-apply went mainstream. AIHawk, an open-source bot, sits at roughly 30,100 GitHub stars. A single 404 Media reporter fired off about 2,843 applications with it. This is no longer a fringe hack.
• Candidate bots are the median, not the outlier. Greenhouse's 2025 workforce research found 22% of active job seekers admit to using bots to auto-apply.
• The macro flood is measurable. LinkedIn applications hit 14,200 per minute in February 2026, up 58% from 2024. US applicants per open role have doubled since 2022.

The polished cover letter used to be a proxy for effort. That proxy is now fully controlled by whichever LLM the applicant piped their resume through. Any triage step that leans on writing quality is now noise.

The outbound math is not close

Outbound sourcing converts at roughly 12x the rate of generic inbound and about 3.4x the rate of a well-run HN post. The numbers are not subtle, and they compound with role scarcity.

Here is the comparison the research supports, laid out end to end:

The mechanism behind the outbound advantage is straightforward: outbound filters before the funnel, not during it. When you name the candidates you want and then reach out, the noisy population never enters your pipeline in the first place. Every inbound-first strategy is trying to filter noise after it has already arrived, which is why moderation cost keeps climbing: 67% of US HR leaders say reviewing AI-generated applications has slowed hiring, and one in five report delays of more than two weeks.

The polished cover letter is now a metric AI fully controls. Triage on it and you triage on nothing.

The scarcer the skill, the worse HN gets

For any specialized stack, HN inbound is mathematically absurd, because the global candidate pool is often smaller than a single month's applicant pile. This is where Refolk's index makes the point sharpest.

In Refolk's index of professional profiles, there are approximately 100,000 US-based Software, Senior, and Staff Engineers with Python listed as a skill. In the same index, there are 1,095 US SWEs with Rust. That is a ratio of roughly 91 to 1. A Rust role posted on HN and answered by 170 applicants is drawing from a broader auto-apply population many multiples larger than the actual target pool.

The named Rust employers in Refolk's index tell you exactly where to look instead:

• Oxide Computer
• Figure
• Shopify
• Uniswap
• Helius

If you are hiring a Rust engineer in 2026, the correct question is not "how do I get more HN applicants?" It is "which of the 1,095 US Rust engineers should I contact this week?" That is the exact gap Refolk closes: you describe the person in plain English ("US-based Rust engineer with distributed-systems experience, ex-Oxide or ex-Figure preferred") and get a ranked shortlist across GitHub, LinkedIn, and the open web.

The Python situation is different in size but similar in shape. A pool of ~100,000 is large, but it is still finite, still nameable, and still worth attacking with a targeted query rather than a public post that any auto-apply bot can spray.

What to do outbound instead, this week

Replace the HN post with a five-step outbound loop that filters before candidates enter your funnel. This is a concrete week-one playbook, not a philosophical shift.

1. Name the target pool. Write the role description as a query, not a JD. "Staff SWE, Rust, distributed systems, US, has shipped at Oxide or Figure" is a query. "Passionate about cutting-edge infra" is not.
2. Pull a ranked list, not a search dump. Refolk's plain-English sourcing across GitHub, LinkedIn, and the open web is designed for this exact step. Give it the query, get the shortlist. Skip the Boolean gymnastics.
3. Read the top 20 profiles yourself. Cut anyone whose GitHub or employer history contradicts the query. A 20-profile human read takes 40 minutes and beats any AI screener at reproducibility.
4. Send 20 personalized first messages. Reference one specific repo, patent, talk, or shipped product per candidate. If you cannot find one specific thing to reference, they were not in your real pool.
5. Measure reply rate weekly. Outbound benchmarks put you at roughly 6% offer conversion on replies. If you are well below that, the query was too broad. Rewrite it.

The founders still posting in the August 2026 HN thread (Bucket Robotics, Estuary, Langfuse, and FusionAuth among them) are betting that their brand pull will outrun the auto-apply flood. For a small subset of them, at very specific moments, it will. For everyone else, the arithmetic says a targeted outbound loop against a named pool will produce more hires, faster, with less moderation cost.

The one number that should change your Q4 plan

If you take one number from this piece, take the ratio: outbound converts at approximately 12x generic inbound, and approximately 3.4x an above-average HN inbound month. Any hiring plan for Q4 2026 that leaves inbound as the primary channel is planning to hit its ceiling.

91x

The specific action for founders and engineering leaders reading this in August 2026:

• Stop measuring channels against each other. Measure them against the outbound floor of ~6%.
• Name your pool before you write a JD. If you cannot name it, you are not hiring, you are wishing.
• Treat the HN thread as a brand post, not a sourcing channel. Post there because it signals culture, not because it produces hires.
• Use a plain-English sourcing tool for the actual work. Refolk exists because "find me the 40 people who match this" is a query, not a Boolean puzzle.

Prophet Town's 170-to-3 is not an outlier. It is a fair sample of what a well-run inbound month looks like on the best inbound channel technical hiring has left. The channel is not coming back. The math already moved.

FAQ

Is Hacker News' "Who is Hiring" thread actually dead for sourcing engineers in 2026?

Not dead, but decisively below the outbound floor. Prophet Town's self-reported 1.76% applicant-to-offer rate in the August 2026 thread is roughly 3.5x better than generic inbound (0.5% per Gem's 2026 benchmark on 165M applications), yet still about 3.4x worse than outbound sourcing at 6%. Treat the thread as a low-cost brand signal, not a primary channel, and put your hiring hours into targeted outbound against a named pool.

Why did the thread's signal degrade so fast?

Two mechanisms compounded. On the poster side, the "no recruiting firms or job boards" rule broke visibly in August 2026 when CareerJumpShip openly posted a scraped Greenhouse/Lever/Ashby roll-up. On the applicant side, auto-apply went mainstream: AIHawk sits at ~30,100 GitHub stars, 22% of active job seekers admit to using bots (Greenhouse 2025), and LinkedIn hit 14,200 applications per minute in February 2026. When the norm dies and the flood arrives at the same time, signal collapses.

What is the fastest way to switch from HN inbound to outbound sourcing?

Write the role as a plain-English query, not a JD. Then pull a ranked shortlist from a sourcing tool built for that: Refolk is designed for exactly this, describe the person and get the right people across GitHub, LinkedIn, and the open web. Read the top 20 profiles yourself, send 20 personalized first messages with one specific reference each, and measure reply rate weekly against the 6% outbound benchmark.

Does outbound work for abundant skills like Python, or only for scarce ones like Rust?

It works for both, for different reasons. For scarce stacks, outbound is the only channel that mathematically intersects the pool: Refolk's index shows just 1,095 US Rust SWEs vs ~100,000 Python SWEs, a 91-to-1 ratio, so any Rust role is a naming exercise. For abundant stacks, outbound still wins because it filters before the funnel: the 6% conversion rate holds regardless of pool size, while inbound conversion degrades as auto-apply volume climbs.