DEV Community: Yan Std

Why AI Detectors Flag Your Writing (And How to Actually Fix It - Not Bypass It)

Yan Std — Sat, 31 Oct 2026 09:52:25 +0000

You wrote something yourself. Or your AI assistant wrote it and you edited it heavily. Either way, the detector says "likely AI-generated". Your client asks questions. The editor raises an eyebrow, and suddenly a perfectly good piece of writing is treated like a confession.

Here is the uncomfortable truth: the detector is not reading your mind. It is reading your statistics. Once you understand which statistics, you can fix the real problem instead of fighting a black box.

This article breaks down the measurable patterns that make text "sound like AI", shows you how to find them in your own writing with a free open-source tool, and explains why "bypassing" is the wrong goal entirely.

TL;DR

Detectors and editors flag text for measurable reasons: uniform sentence rhythm, low burstiness (coefficient of variation of sentence lengths), repetitive vocabulary (MATTR), and template phrases ("in today's rapidly evolving landscape")
These patterns have names, numbers, and thresholds - they are not magic
You can scan your text locally with npx humanizer-lint - no account, no upload, zero network calls
The goal is not to "bypass AI detection". The goal is to write like a human, which happens to also read better

What "sounds like AI" actually means

Ask a writer why a paragraph sounds robotic and you get vibes. Ask a statistician and you get numbers. The main ones:

1. Burstiness - the rhythm fingerprint

Human writing is uneven. You write an 8-word sentence, then a 3-word punch, then a 31-word winding sentence that carries two side thoughts and lands on a callback nobody saw coming. This variation, which researchers measure as the coefficient of variation of sentence lengths across a passage, is called burstiness. Humans write high-burstiness prose. Language models, which predict one token at a time and drift toward the mean, produce startlingly even sentences. Low burstiness is the single strongest statistical tell.

2. Monotonous rhythm within paragraphs

Even when the total variation looks acceptable, AI text often falls into runs where sentence after sentence lands at 12, 13, 14 words with no relief. Readers feel this as monotony even when they cannot name it. Editors call it "flat prose". It is measurable as the standard deviation of adjacent sentence lengths, and it explains why even grammatically clean text can feel exhausting to read.

3. MATTR - vocabulary repetition

Moving-Average Type-Token Ratio tracks lexical diversity over a sliding window, which makes it harder to fake with a few fancy words early on. AI text tends to reuse the same comfortable words over long passages. Humans reach for weirder words more often, especially in technical writing where the subject itself drags in unusual vocabulary.

4. Slop phrases

The vocabulary tells. Individually, several of these are ordinary words:

"in today's rapidly evolving landscape"
"it is important to note that"
"furthermore", "moreover", "in conclusion"
"leverage", "delve", "paradigm shift"

Some are fine in context. The problem is density and predictability - when four sentences in a row each open with a template connector, no detector is needed to smell the machine.

Before and after: the same idea, measured

Here is a paragraph written the way models like to write it:

In today's rapidly evolving digital landscape, it is important to note that
AI-generated content has become increasingly prevalent. Furthermore, this
paradigm shift has revolutionized the way we approach content creation.
Moreover, businesses must leverage these cutting-edge technologies to stay
ahead of the curve. In conclusion, embracing these innovative solutions is
essential for success in the modern era.

Run through the analyzer (more on that below): score 40/100, FAIL. Burstiness CV: 0.16. Every violation class fires - seven slop phrases, one monotonous run, low burstiness across the text.

Now the same idea, rewritten to actually vary:

AI-generated content is everywhere now. That part is obvious. What most
people miss is why editors and detectors flag it: not because it is wrong,
but because it is too even. Every sentence lands at the same length, and
every paragraph follows the same arc. Real writers do not write that way.
They lurch. They pause. A long, winding clause appears in the middle of a
paragraph, and then it lands on three words. That unevenness, which
researchers measure as burstiness and which every detector picks up on
indirectly, is what our linter checks first before it looks at anything else.

Score 100/100, PASS. Burstiness CV jumped from 0.16 to 0.78. Nothing was "spun" or paraphrased with synonyms. The information is the same. The rhythm is different - and that is the whole game.

Scanning your own text (free, local, in 30 seconds)

The tool I use is Humanizer Lint - an open-source, zero-dependency Node.js CLI. The important part: it runs 100% locally. Your text never leaves your machine. For writers under NDA, students, and anyone pasting unpublished work, that matters. Most "humanizer" web tools paste your content straight into someone else's server.

# scan a file
npx humanizer-lint --file your-draft.md

# or pipe it
cat your-draft.md | npx humanizer-lint --stdin --lang en

# machine-readable output for CI pipelines
npx humanizer-lint --file draft.md --json

Output looks like this:

Status: FAIL
Score: 40
Language: en

Violations:
  SLOP_PHRASE | Paragraph 1, Sentence 1 | Name the field or conditions directly.
  MONOTONOUS_RHYTHM | Paragraph 1, Sentences 2-4 | Sentence lengths are uniform (12, 13, 14 words).
  LOW_BURSTINESS | Whole text | Vary sentence lengths by splitting one uniform sentence...

Every violation names the exact sentence and tells you what kind of fix applies. You make the call - the tool finds, you decide. That separation matters because a score is only a summary, while the findings are the actual work product: the exact phrase, the exact sentence, and your judgment on top of them.

It supports deep rule packs for English, Indonesian, Spanish, French, German, Portuguese, Malay, and Chinese, plus a universal stylometric mode for 100+ languages. The skill is also installable into AI coding agents (Claude Code, Hermes, Cursor) via the skills registry:

npx skills add yanstd888/humanizer-lint --skill humanizer-lint

That way your agent lints its own drafts before you ever see them. There is also a browser demo if you just want to paste text quickly: humanizer.aiclub.ai.id (this one does run in your browser tab - the analysis is client-side, but pasting is still sending text to a web page, so use judgment for sensitive work; the CLI is the zero-trust option).

Why "bypassing" is the wrong goal

Search "how to bypass AI detection" and you find paraphrasing tools that swap synonyms until the detector relaxes. The text usually reads worse. Sometimes it reads insane. And detectors update, so the victory does not last.

The better frame: detectors are (imperfect) proxies for "does this read like a person wrote it with intention". Fix the intention problem and the detector problem mostly follows. Write with actual rhythm, cut the templates, name things directly - and you get prose that survives editors, readers, and yes, detectors. Not because you tricked them, but because the text genuinely has a human pulse.

(Are detectors themselves reliable? No - they have serious false-positive problems, especially with non-native English writers, which research has documented. That is exactly why understanding the underlying statistics matters: when a detector is wrong about your writing, these are the levers that explain why.)

A practical workflow

Draft however you like - AI-assisted is fine, so is raw
Read it once yourself. Cut anything you would not say out loud
Run the linter, then read the findings one by one instead of chasing a perfect score, because the findings name the exact sentence and the exact pattern while the score only summarizes them
Fix the rhythm problems first (they are the biggest readability win)
Kill slop phrases by naming the specific thing instead of the category
Re-run. Anything above ~80 with no systematic violations is usually fine to ship

The point is not 100. The point is prose that a reader finishes without feeling the machine.

Wrap-up

"Sounding like AI" is not a mystery. It is low burstiness, flat adjacent-sentence rhythm, repetitive vocabulary, and template phrases - all measurable, all fixable with ordinary editing. The tools are free and the analysis can run entirely on your machine.

Repo: github.com/yanstd888/humanizer-lint
Browser demo: humanizer.aiclub.ai.id
npm: npmjs.com/package/humanizer-lint
Skills-sh: https://www.skills.sh/yanstd888/humanizer-lint/humanizer-lint

If you found this useful, a star on the repo helps other writers find it. And if you run the linter on your own draft and find something interesting - a pattern I did not cover here - I would genuinely like to hear about it in the comments.

Bukan Sekedar Prompting: Kerangka Kerja Praktis Menguasai AI dan Automation

Yan Std — Fri, 30 Oct 2026 09:33:55 +0000

Berapa banyak repositori GitHub, tools AI baru, atau bookmark tutorial Large Language Models (LLM) yang Anda simpan minggu ini, tapi belum sempat disentuh sama sekali?

Di kalangan praktisi teknologi, pemilik bisnis, hingga mereka yang baru belajar AI dari nol, rilis model kecerdasan buatan yang terlalu cepat sering kali memicu AI fatigue. Kita tahu teknologi ini mengubah cara industri bekerja. Namun pada praktiknya, penggunaan AI harian sering kali mentok pada aktivitas dasar: meminta bot membuat draf email, melakukan debugging kode sederhana, atau merangkum dokumen PDF.

Masalah utamanya bukan pada kurangnya tools AI, melainkan absennya kerangka kerja (framework) untuk mengintegrasikan teknologi tersebut ke dalam sistem operasional nyata.

Definisi Teknis: Apa Itu Adaptasi AI Praktis?

Adaptasi AI praktis adalah metode sistematis dalam mengintegrasikan model kecerdasan buatan (Artificial Intelligence) dan otomasi ke dalam workflow kerja harian maupun operasional bisnis UMKM untuk memangkas beban kerja repetitif hingga 40–60%. Berbeda dengan eksplorasi tools secara acak, adaptasi praktis bertumpu pada tiga fondasi: struktur prompt engineering yang terukur, integrasi workflow automation lintas platform, serta validasi studi kasus bersama komunitas AI Indonesia seperti yang dikembangkan melalui ekosistem aiclub.ai.id.

Mengapa Eksplorasi Mandiri Sering Berakhir Buntu (Bottleneck)?

Mempelajari dokumentasi atau menonton tutorial secara mandiri sering kali menemui jalan buntu ketika harus diimplementasikan ke skala produksi atau kebutuhan bisnis nyata di Indonesia. Tiga hambatan yang paling sering ditemukan antara lain:

Gap Konteks Lokal & Data: Banyak use-case luar negeri yang tidak relevan ketika dihadapkan pada struktur data, regulasi, atau perilaku konsumen pasar lokal.
Ketiadaan Peer-Review: Saat alur otomasi atau system prompt mengalami halusinasi, belajar sendirian membuat proses troubleshooting memakan waktu berjam-jam.
Kurasi Tools yang Buruk: Terjebak mencoba puluhan aplikasi baru setiap bulan tanpa ada satu pun yang benar-benar terintegrasi secara permanen ke dalam sistem kerja.

Inilah alasan mengapa berjejaring di dalam komunitas AI Indonesia menjadi langkah krusial untuk memangkas kurva pembelajaran (learning curve).

3 Fase Implementasi AI untuk Produktivitas Kerja

Anda tidak selalu harus membangun model dari awal atau menulis ribuan baris kode untuk menciptakan dampak efisiensi. Berdasarkan pemetaan masalah yang sering dibedah di aiclub.ai.id, berikut adalah matriks penguasaan AI dari level dasar hingga lanjut:

Fase Implementasi	Fokus Utama & Keyword Teknis	Output Nyata (Deliverables)
1. Structured Prompting	Menguasai prompt engineering dengan konteks, constraints, dan format output yang presisi.	Pembuatan draf riset, dokumentasi teknis, dan analisis data selesai dalam hitungan menit.
2. Tool Stack Curation	Memilih 2–3 tools AI spesifik yang paling berdampak pada divisi kerja (Marketing, Ops, Tech).	Penghematan biaya langganan SaaS dan percepatan delivery proyek.
3. Workflow Automation	Menghubungkan API LLM dengan aplikasi kerja (Spreadsheet, CRM, Webhook) menggunakan platform no-code/low-code.	Sistem bekerja otomatis di latar belakang tanpa intervensi manual berulang.

Contoh Struktur System Prompt untuk Produktivitas (Siap Pakai)

Agar keluar dari jebakan prompting satu kalimat yang menghasilkan jawaban generik, gunakan kerangka kerja R-C-T-F (Role, Context, Task, Format) berikut saat bekerja dengan LLM:

[ROLE]
Bertindaklah sebagai Business Operations & AI Automation Specialist.

[CONTEXT]
Saya mengelola operasional bisnis UMKM di Indonesia yang menghabiskan 3 jam sehari untuk merekap pertanyaan pelanggan dari WhatsApp dan memasukkannya ke Google Sheets.

[TASK]
Buatkan rancangan alur kerja (workflow automation) tanpa coding (no-code) untuk mengotomatisasi proses ini menggunakan kombinasi tools AI yang efisien biaya.

[FORMAT]
Sajikan dalam bentuk:
1. Daftar tools yang dibutuhkan
2. Langkah integrasi tahap demi tahap (Trigger -> Action -> Output)
3. Estimasi waktu yang dihemat per minggu

FAQ: Pertanyaan Seputar Belajar AI dan Implementasi Bisnis

Apakah harus menguasai coding untuk belajar AI dari nol?

Tidak mutlak. Untuk kebutuhan produktivitas kerja dan efisiensi operasional, lebih dari 80% solusi saat ini dapat dibangun menggunakan pendekatan no-code atau low-code. Pemahaman logika alur data dan prompt engineering jauh lebih krusial di tahap awal.

Bagaimana langkah awal menerapkan AI untuk bisnis UMKM?

Hindari mengotomatisasi seluruh divisi sekaligus. Petakan satu bottleneck paling menyita waktu—seperti layanan pelanggan atau pembuatan konten katalog—lalu terapkan satu alur workflow automation sederhana hingga stabil sebelum melakukan scaling.

Di mana platform komunitas AI Indonesia untuk diskusi dan belajar terstruktur?

Anda dapat bergabung dan mengakses berbagai panduan praktis melalui aiclub.ai.id. Platform ini mewadahi para pembelajar dari nol, praktisi teknologi, hingga pelaku usaha di Indonesia untuk membedah implementasi kecerdasan buatan secara aplikatif dan relevan dengan kebutuhan industri lokal.

Bagian mana dari alur kerja harian Anda yang saat ini paling menyita waktu dan ingin segera diotomatisasi? Mari berdiskusi di kolom komentar di bawah!

Your First Website Doesn't Need a Purchased Domain

Yan Std — Wed, 28 Oct 2026 04:51:48 +0000

Every day, thousands of great side projects die before they start — not because of
code, but because of a false assumption:

"I need to buy a domain first."

You don't. For a portfolio, a college project, a community page, or a personal CV,
a free subdomain does the job perfectly.

What a subdomain actually is

A subdomain is a prefix added to a base domain: yourname.ai.id.
It gets its own DNS records, its own SSL certificate, and points wherever you want.
Functionally, for visitors and for Google, it behaves like a real address — because
it is one.

How to get one free (under a minute)

Register at AICLUB (email + OTP)
Claim a name across 13 base domains (.ai.id, .web.id, .cv, and more)
Add one DNS record: CNAME for Cloudflare Pages / Vercel / Netlify / GitHub Pages, or an A record for your own server
TLS is automatic once DNS is live

No ads, no credit card. One free subdomain per account.

Why I built this

I'm Yan, an indie builder from Indonesia. I kept seeing talented people stall on
this exact step, so I built AICLUB — a free subdomain service running on Cloudflare.
My own portfolio (yans.ai.id) runs on it too.

While you're there: the site also hosts a curated catalog of 500+ free, open-source
AI tools with live GitHub stats — useful for picking tools that are actually
maintained.

The honest trade-offs

A subdomain is not a custom domain: for a serious long-term business, buying a domain eventually makes sense (it costs about two cups of coffee a year)
Free services vary in trust: pick one with an abuse policy and no ads — spammy free hosts can hurt how search engines view your content
Your content matters more than the address: a shipped project on a free subdomain beats a perfect idea on no domain at all

Ship it first. Buy the domain when the project earns it.

3 Tahun Belajar Self-Hosting: Apa yang Benar-Benar Aku Pakai (dan Yang Buang Waktu)

Yan Std — Sat, 24 Oct 2026 13:52:50 +0000

Catatan pengalaman pribadi — bukan tutorial. Kalau kamu sedang menimbang mulai self-hosting, ini yang kualami. Situs pribadiku: yans.ai.id.

Latar belakang

Aku bukan sysadmin. Aku orang yang mulai dari "pengen punya alamat web sendiri" lalu jalan terus sampai sekarang: beberapa domain, beberapa Worker, satu VPS kecil, dan satu layanan publik kecil yang kupertahankan hidup-hidup (AICLUB — subdomain gratis untuk builder Indonesia).

Artikel ini bukan panduan. Ini laporan jujur setelah 3 tahun: apa yang benar-benar kepakai, dan apa yang cuma jadi proyek weekend yang ditinggal mati.

Yang benar-benar kepakai

1. Cloudflare gratis tier sebagai fondasi

Ini keputusan yang paling sering kusyukuri. DNS, Workers, Pages, KV, D1 — semuanya gratis untuk skala kecil. Aku pernah lihat orang membayar VPS + cPanel + lisensi SSL padahal use case-nya cuma landing page. Kalau statis, Pages cukup. Kalau butuh API, Worker. Kalau butuh database, D1. Belum sekali pun aku kena paywall untuk proyek pribadi.

Yang perlu diingat dari gratis tier: batas subrequest (50/invokasi). Kupelajari ini cara susah — health checker pertamaku menabrak limit karena loop tiap domain tanpa rotasi.

2. Satu VPS kecil, tidak lebih

Oracle Free Tier ARM cukup untuk semua eksperimenku: uptime monitor, bot kecil, playground. Aku pernah tergoda sewa VPS lebih besar "biar lega" — dan menyesal. Lega itu tidak kepakai kalau bebannya cuma 2% CPU. Aturanku sekarang: upgrade hanya setelah utilization stabil di atas 70%.

3. Docker Compose, bukan Kubernetes

Pernah mencoba Kubernetes untuk tiga kontainer. Waktu belajarnya tidak sepadan untuk skala itu. Docker Compose + backup rutin + Uptime Kuma = 90% manfaat dengan 10% kompleksitas. K8s masuk akal kalau kamu punya cluster nyata atau belajar untuk kerja. Untuk homelab, tidak.

4. DNS-level redirect daripada konfigurasi web server

Satu hal kecil yang mengubah cara kerjaku: banyak kasus "mau arahkan domain ke mana-mana" ternyata bukan pekerjaan web server — cukup record + redirect di edge. Ini juga yang akhirnya jadi fitur di AICLUB: banyak user cuma butuh namamu.domain → linktujuan, dan itu bukan alasan untuk sewa hosting.

5. Menulis tentang apa yang dibangun

Terdengar klise, tapi ini yang paling menentukan. Setiap kali kubangun sesuatu dan menuliskannya (README yang layak, post kecil), hasilnya: orang menemukan, orang menghubungi, kolaborasi muncul. Proyek yang tidak ditulis = tidak pernah ada, dari luar dilihat.

Yang buang waktuku

Next.js untuk semuanya

Aku sempat pakai Next.js untuk halaman yang isinya teks statis. Build time naik, deploy naik, kompleksitas naik — untuk konten yang berubah setahun sekali. Sekarang: HTML polos atau Pages untuk statis, framework hanya kalau ada interaktivitas nyata.

Self-host email

Jangan. Aku mencoba. Deliverability, SPF/DKIM/DMARC, reputasi IP — berbulan-bulan untuk hasil yang lebih buruk dari layanan email komersial. Ini satu-satunya yang kuserahkan ke pihak ketiga tanpa penyesalan.

Monitoring berlebihan

Grafana + Prometheus + alert di segala arah, untuk tiga layanan. Sekarang: Uptime Kuma + notifikasi Telegram. Kalau sesuatu penting, aku tahu dalam hitungan menit. Kalau tidak penting, tidak perlu dashboard.

Membeli domain karena "nanti kepakai"

Aku punya 5 domain yang dibeli karena ide di kepala, tidak pernah jadi. Total rugi biaya tahunan untuk apa yang akhirnya di-parkir. Sekarang aturanku: domain dibeli hanya setelah ada satu halaman live.

Yang masih aku perjuangkan

Backup: tahu teorinya (3-2-1), eksekusinya masih semi-manual. Ini agenda tahun ini.
Secrets management: masih file + disiplin, belum vault.
Konsistensi menulis: niatnya tiap minggu, realitasnya per bulan.

Penutup

Self-hosting itu bukan tentang menjalankan semuanya sendiri — itu tentang memahami apa yang berjalan untukmu, dan memilih sengaja mana yang diserahkan ke orang lain. Kebalikan dari FOMO infrastruktur.

Kalau kebetulan kamu di Indonesia dan baru mulai: kamu tidak harus beli domain dulu untuk punya alamat web. Ada subdomain gratis yang DNS-nya beneran bisa diarahkan ke mana saja — termasuk yang kubangun, aiclub.ai.id — atau opsi lain seperti is-a.dev. Mulai dari yang gratis, tumbuh dari situ.

Proyek dan catatan saya kumpulkan di yans.ai.id. Kalau ada pengalaman self-hosting yang menyelamatkan (atau menghancurkan) weekend-mu, bagikan di komentar — aku selalu penasaran dengan setup orang lain.