TrickCV Data Report
🤖 Public data · 17 months of HN · 6 sources

What 6,539 tech job postings actually require

I scraped every "Ask HN: Who is hiring?" thread from January 2025 to May 2026 plus 5 other public job boards, then ran a full analysis on what the dataset actually says — vs what you think it says. Below: 13 sections, every chart sourced, every number reproducible.

📊 7,066 raw postings 🎯 6,539 tech-relevant 📅 17 HN threads 🌐 6 sources 🧪 85 skills tracked 🥴 30 fluff phrases
3.4×
more PyTorch than TensorFlow mentions across the dataset
65%
of "Senior" roles ask for ≤5 years of experience
$150K-$210K
median comp range from 956 explicit postings (15% of dataset)
In this report
01 — Skills

Top 15 required skills across all JDs

Counted across every JD in the tech-filtered subset (n=6,539). Python is #1 — but TypeScript + React together describe a frontend stack present in roughly 1 in 4 postings, comparable in scale to Python.

02 — Languages

Programming languages, ranked

Tracking 16 languages, only showing those with non-zero presence. The Python lead is 7 percentage points over TypeScript, and the gap widens further from there. Rust has firmly overtaken Java in HN-flavored hiring.

03 — Frontend

React is no longer just dominant — it's the assumption

React appears in 22.0% of tech postings. Vue+Angular+Svelte combined account for less than half of that. Next.js shows up in 5.5%, which is striking — it means roughly 1 in 4 React jobs explicitly mention the Next.js stack.

Tailwind appears in twice as many postings as it did a year ago. Stack inertia is real, but at the trend level the CSS-in-JS era looks over.

04 — Cloud

AWS dominance is bigger than you'd guess

AWS appears in 15.0% of postings. GCP 5.9%. Azure 3.0%.

That's roughly a 3:1 ratio of AWS to GCP and 5:1 AWS to Azure in HN-tier hiring. The "multi-cloud" narrative doesn't show up in JDs — postings strongly bias toward a single provider.

05 — Data layer

Postgres won the database war

PostgreSQL appears in 15.0% of postings — the same share as AWS. MySQL barely registers; the gap to Postgres is ~5x. Redis dominates caching; MongoDB still hangs around but is shrinking year-over-year.

06 — AI/ML stack

The 2026 AI/ML stack, decoded

Tracking 15 markers of the modern AI/ML stack. LLM shows up in 11.5% of all tech postings — meaning roughly 1 in 9 tech jobs now explicitly touches LLMs. Anthropic mentions narrowly beat OpenAI in the dataset (3.24% vs 2.86%), which is a notable shift from 2024 when OpenAI dominated 5:1.

LLM mention rate
11.5%
RAG mentions
1.9%
Anthropic vs OpenAI ratio
1.13×
07 — Framework war

The ML framework war is over (mostly)

PyTorch appears in 133 postings. TensorFlow in 39. That's a 3.4× margin in favor of PyTorch across 17 months.

The contrarian wrinkle: TensorFlow is the fastest-trending framework in the most recent 3-month window (+120% vs prior 3). Real comeback or sampling noise? Open question — but worth watching.

Note: covers explicit framework mentions in JD bodies. Doesn't capture "we use X internally" job-board categories or roles where ML is implied but unstated.

08 — Contrarian finding

The "Senior" YoE myth

Of roles with "Senior" in the title that explicitly state YoE requirements, 65% ask for 5 years or less.

The median for "Senior" is 5 years. For "Staff": 10. The big jump is Senior → Staff, not Mid → Senior. If you have 5 YoE and aren't applying to Senior roles, you're underselling — by 2-3 years' worth of compensation.

09 — YoE distribution

3-5 years is the modal experience ask

Of the 418 postings with an explicit minimum YoE, the largest bucket is 3-5 years by a clear margin. Only ~7% ask for 10+ years.

Translation: if you're 3-5 years in and feel "junior," the market disagrees. The HN-tier hiring band is overwhelmingly mid-senior, not staff+.

11 — Compensation

Comp transparency is up

956 postings (~15% of dataset) include an explicit USD comp range — a major shift from the "competitive salary" era of 3 years ago.

Median band: $150K – $210K. 25th-75th percentile band: $120K – $250K.

53% of postings are remote, keeping US-equivalent salaries in play for globally distributed teams.

12 — Stack co-occurrence

"If a JD requires X, what else does it require?"

For each of the top 15 skills: the 5 most common partner skills, plus how often each partner appears in the same JD as the primary. The percentages are conditional — given the JD mentions the primary, what % of the time does it also mention the partner?

13 — Founding engineers

What "Founding Engineer" actually means in 2026

326 postings mention "founding engineer" / "founding software" / "founding member" — that's 5% of the entire dataset.

The stack is heavily TypeScript + React + Python. LLM shows up in nearly 1 in 4 founding-engineer postings — the AI wave is the dominant story behind the founding-engineer hiring boom.

14 — The fluff index

Corporate fluff phrase frequency 🥴

Tracking 30 well-worn phrases. The good news for HN: "self-starter" appears in only 0.6% of postings — about an order of magnitude below typical corporate JDs. HN posters write meaningfully better JDs.

15 — Source comparison

How HN differs from other job boards

% of postings on each source that mention each of the top 5 skills. HN over-indexes massively on the modern stack (Python, TypeScript, React, AWS) — Arbeitnow (Europe-heavy) and The Muse (corporate-heavy) skew differently.

Methodology & sources

Data sources
  • 5,717 jobs — Hacker News "Who is hiring?" threads (Jan 2025 – May 2026, via the public HN Algolia API)
  • 1,000 jobs — Arbeitnow public job board API (EU-heavy)
  • 182 jobs — The Muse public API (Engineering/Data/Product categories)
  • 97 jobs — RemoteOK public feed
  • 50 jobs — Jobicy remote-jobs API
  • 20 jobs — Remotive public API
How analysis works
  • 85 technologies matched against full JD body via regex (case-insensitive, word-boundary'd)
  • 30 corporate phrases tracked the same way
  • YoE extracted with 4 patterns (e.g. "3+ years", "5-7 years", "at least N", "minimum N")
  • Compensation parsed from explicit USD ranges (e.g. "$150K-$210K"), filtered to plausible bounds
  • Trending = recent 3 HN months vs prior 3, gated by min 5 prior mentions
  • Co-occurrence = per top-15 skill, how often each other skill appears in the same JD
Limitations: HN is engineer-heavy and startup-heavy — findings won't generalize cleanly to corporate recruiters, non-tech roles, or non-English markets. Skill detection is presence-based; a JD that doesn't mention "Python" may still require it implicitly. Compensation parser only catches explicit USD ranges. Trending detection requires ≥5 prior mentions. All data is a point-in-time snapshot regenerated each scrape.

Reproducibility: Every scraper, the analyzer, and the raw dumps are in the data/ folder of this project. Re-run with ./run_pipeline.sh. Open-source under MIT for the code.

Tailoring your CV against this stack?

TrickCV reads any job description and rewrites your CV to match — in under 10 seconds, free trial, no credit card.

Try TrickCV free →