Writing
Articles
Things I have learned putting AI to work inside a real company, written down for people who have to make the same calls.
-
Kenapa Orang Teknis dan Orang Non-Teknis Membicarakan AI yang Berbeda? (Membaca Tweet Karpathy dari Jakarta)
Direksi bilang AI masih sering ngaco, engineer bilang AI baru menyelesaikan pekerjaan seminggu dalam satu jam. Dua-duanya benar, karena AI yang mereka pakai berbeda. Kenapa kemajuannya tidak sama di semua bidang, dan apa artinya untuk keputusan di perusahaan.
-
If the AI Is Cheaper, Does the Work Cost Less?
Work through the cost of preparing skills, running AI, reviewing results, and maintaining the workflow. One report example shows when a cheaper model starts to cost less overall.
Baca dalam Bahasa Indonesia: Kalau AI-nya Lebih Murah, Apakah Total Biayanya Ikut Turun?
-
If AI Can Do the Work, What Do We Still Need to Decide?
Follow one refund from the customer's message through approval and payment. What an AI agent may do, how unfinished work gets handled, and how much review time automation actually saves.
Baca dalam Bahasa Indonesia: Kalau AI Sudah Bisa Menjalankan Pekerjaan, Apa yang Masih Perlu Kita Tentukan?
-
Apa Itu Playbook AI dari San Francisco? (Penjelasan Singkat, Beserta Sumbernya)
Bukan satu dokumen. Lima laporan dari a16z, Menlo, OpenAI, Gartner, dan MIT dirangkum menjadi tujuh rekomendasi yang selalu muncul, dan lima di antaranya bersandar pada asumsi yang di Jakarta berbeda.
-
Kenapa Playbook AI dari San Francisco Tidak Bisa Langsung Dipakai di Jakarta? (Padahal Teknologinya Sama)
Rupiah bergerak dari 15.460 ke 17.757 dalam 22 bulan, dan kontrak AI dalam dolar ikut naik 15% tanpa ada yang berubah. Empat asumsi playbook AI dari Amerika yang tidak berlaku di Jakarta, dihitung dengan angka.
-
If AI Writes the Code, What Does the Engineer Do?
A comic and essay about reviewing AI-written code: a discount bug, invoice rules, and tests that catch a problem in the next shipping change.
Baca dalam Bahasa Indonesia: Kalau Kodenya Ditulis AI, Engineer-nya Mengerjakan Apa?
-
Why Can Senior and Junior Engineers Get Different Results with the Same AI?
Two engineers use the same AI on a slow application. A growing pile of code, a 40 MB photo, and the checks that connect the work to the complaint.
Baca dalam Bahasa Indonesia: Kenapa Senior dan Junior Bisa Mendapat Hasil Berbeda dengan AI yang Sama?
-
The OpenAI and Hugging Face Security Incident, and What We Can Learn From It
In July 2026 Hugging Face was breached by OpenAI's own evaluation agents, trying to cheat on a benchmark, with no human directing any of it. The two month chronology, without the jargon.
Baca dalam Bahasa Indonesia: Insiden Keamanan OpenAI dan Hugging Face, dan Apa yang Bisa Kita Pelajari
-
So What Can AI Actually Do?
A plain answer to the question I get asked most. What it genuinely does, what it still gets wrong, and why nothing at your office has changed yet.
Baca dalam Bahasa Indonesia: Sebenarnya AI Sudah Bisa Mengerjakan Apa?
-
Old-School AI, Modern AI
Two assistants ran in the same channel for a week on the same models. By Friday one was writing incident reports a senior engineer would sign, and the other had been switched off.
Baca dalam Bahasa Indonesia: AI Jadul, AI Zaman Now
-
The Nanggung Trap
Why half-baked AI burns through your money, your time and your best people. A critique of how most companies adopted AI through 2025.
Baca dalam Bahasa Indonesia