I Classified 842 'Who Is Hiring' Posts Across Three Augusts. AI Mentions Doubled. AI Requirements Flatlined.

TL;DR: I pulled three August "Ask HN: Who is hiring?" threads (2023, 2025, 2026), took the 842 top-level postings that were actual job ads, and flagged each on four things. AI mentions went 27.7% to 52.9% to 63.4%. But the share that put AI experience under requirements went 2.7% to 8.0% to 7.1%: up sharply, then flat and slightly down. So the answer is that AI experience is still overwhelmingly a bullet point. Two thirds of these postings say AI; one in fourteen asks you to have done it. This is Hacker News, a self-selected corpus of mostly startups that read Hacker News, not the labor market, and three points is a shape, not a trend line.

Everybody has a story about AI showing up in job requirements. Almost nobody has counted. The counting is easy, which is the annoying part: Hacker News runs a hiring thread on the first of every month, and the Algolia API returns an entire thread, full comment text included, in one request with no key. I took three, all in the same month to keep seasonality out.

What exactly did I count?

Three threads: August 2023 (item 36956867), August 2025 (item 44757794), and August 2026 (item 49156683).

Top-level postings only. The comment counts you see on the thread include replies, which are complaints about the company, so counting those would be counting the wrong thing. The top-level counts are 344, 307 and 240. From those I dropped flagged comments, "Who wants to be hired" posts that landed in the wrong thread, sub-150-character stubs, and pure commentary, leaving 329, 289 and 224 valid job postings: 842 total.

Each posting got four flags: does it mention AI anywhere; is the company itself an AI product (measured by whether AI terminology appears in the first 260 characters, which on HN is the company header and pitch line); does it name Claude Code, Cursor or Copilot; and does AI appear inside a requirements context, meaning either under a requirements-style heading or in the same sentence as language like "experience with", "must have", "years of" or "proficiency in".

What did the four flags show?

Flag	Aug 2023	Aug 2025	Aug 2026
Valid top-level postings	329	289	224
Mentions AI anywhere	27.7% (91)	52.9% (153)	63.4% (142)
Company is itself an AI product	19.1% (63)	36.0% (104)	37.9% (85)
Names Claude Code, Cursor or Copilot	0.3% (1)	1.4% (4)	4.5% (10)
AI listed under requirements	2.7% (9)	8.0% (23)	7.1% (16)
AI only in a bonus / nice-to-have block	1.8% (6)	1.4% (4)	4.9% (11)

The first row and the fifth row are doing opposite things, and that gap is the whole story. In August 2026, 142 postings mention AI and 16 require it. For every posting that asks you to have AI experience, roughly nine mention AI without asking for anything.

Now the fair caveat, because a raw 7.1% understates it: HN postings are blurbs, not job descriptions. The median posting in my 2026 sample is 154 words, and only 46.9% of them contain requirement language at all (up from 33.4% in 2023). Condition on the postings that actually state requirements and the AI-requirement rate is 8.2% in 2023, 22.3% in 2025, and 15.2% in 2026. Same shape either way: a jump between 2023 and 2025, then a move down into 2026.

So is it a requirement or a bullet point?

A bullet point, mostly, and 2026 is the year that got more true rather than less. The bonus-only row is the cleanest evidence: postings where AI appears exclusively in a nice-to-have block went from 4 to 11 between 2025 and 2026, while hard requirements went from 23 to 16. Companies are moving AI out of the must-have column at the same time as more of them mention it.

You can hear it in the wording. AusRehab in the 2026 thread: "Prior experience with ML, NLP, and agentic AI systems is highly desirable, but not essential." Interview Resources: "Experience with LangGraph, LangChain, MCP, RAG, vector databases, or AI agents is a strong plus, but curiosity" matters more. Shepherd: "If you have experience building AI agents or multi-step reasoning systems, we'd love to hear from you." That is three companies describing AI as an upside, not a gate.

Compare 2023, where the AI requirements that did exist were narrow specialist asks. Trunk Tools wanted "Machine Learning / Search / LLM Engineers with degrees in AI/ML/CS/Math, and who have 2+ years professional experience." That is a research-adjacent role, not a general engineering job with an AI line. The category broadened and softened at once.

Who actually names a coding tool?

This is the row that moved most in relative terms: 1 posting in 2023, 4 in 2025, 10 in 2026. Still only 4.5% of the thread. Three years after these tools became ubiquitous among the exact demographic that posts here, 19 out of every 20 postings do not name one.

The ten that do are specific enough to be worth quoting. Aeqium asks for "proficiency in leveraging AI-assisted development tools (e.g., Claude Code, GitHub Copilot) to accelerate development, optimize workflows, and maintain high code quality." Flywheel Motion describes its whole delivery model as "agentic: Claude Code, Cursor and Aider doing the build work, humans holding editorial judgement." Klara Systems wants "daily experience using AI tools."

That is a genuinely new job requirement, and rare enough in August 2026 that you could read all ten in a coffee break. If you are deciding which of those tools to actually learn, the hiring market is not making the choice for you yet.

Why is the thread shrinking?

The secondary finding, and it holds up cleanly because it needs no classification at all. Total comments on the August thread, straight from the Algolia search index: 895 in 2021, 745 in 2022, 529 in 2023, 466 in 2024, 447 in 2025, 373 in 2026. Six consecutive Augusts of decline, a 58% drop from the 2021 peak. The 2026 months around it agree: 552 in May, 495 in June, 430 in July, 373 in August.

I am not going to tell you AI caused that. A single forum's hiring thread shrinking is consistent with an AI-driven collapse in junior hiring, and equally consistent with Hacker News mattering less as a recruiting channel than it did in 2021. I cannot separate those with this data. What matters is that the denominator is falling while the AI share of it rises, which is the easiest chart in the world to misread, and the reason I report counts next to every percentage above.

What would change my mind?

Three points is not a trend, and I will not claim causation from it. The 2025-to-2026 dip in AI requirements is 23 postings to 16, which is small enough that a couple of unusually verbose months could produce it; if I ran every August back to 2021 and the dip vanished, I would drop that half of the finding without much argument.

The corpus is the bigger limit. Hacker News hiring threads are self-selected, skewed toward small startups, and written by whoever remembered to post that day. Postings are a good corpus for what employers say and a poor one for what they run, which is why 954 NYC postings turned up two employers publishing the bias audit the law requires. Nothing here scales to the American labor market, and any sentence starting "X% of jobs now require AI" is not something this data can support. Indeed's Hiring Lab, which does have the national corpus, found that 37% of the increase in software development postings between May 2025 and May 2026 came from roles with AI in the title. That is growth in AI-titled roles, a different claim from AI becoming a requirement in ordinary ones.

What would move me is a requirements-level count on a real national corpus. Until someone runs it, the honest summary is the one in the table: on Hacker News, AI has won the pitch and lost the requirements section. It is what companies say they are. It is not yet what they ask you to be. If you want the flip side, I also read the requirements on 69 postings actually titled "AI Engineer", where the picture is much stricter, and the layoff narrative gets the same treatment here.