ChatGPT, Claude & Cara Guna AI Dengan Betul — Panduan Prompt Engineering

Ramai yang dah cuba ChatGPT tapi dapat hasil yang hambar. Masalahnya bukan AI tu — masalahnya cara kita bagi arahan. Dalam post ni, kita belajar cara "bercakap" dengan AI supaya dapat hasil yang betul-betul berguna.

Apa Itu Model Bahasa Besar (LLM)?

Large Language Model (LLM) atau Model Bahasa Besar adalah jenis AI yang dilatih khas untuk memahami dan menghasilkan teks dalam bahasa manusia. Ia boleh menulis esei, menjawab soalan, meringkaskan artikel, menterjemah bahasa, dan banyak lagi.

LLM berfungsi dengan cara yang menarik — ia meramalkan perkataan seterusnya berdasarkan konteks. Macam siapkan ayat: "Saya minum ______" → AI ramalkan "air" atau "teh" berdasarkan konteks sebelumnya. Prinsip mudah ini, apabila dibesarkan dengan data trilion patah perkataan, menghasilkan AI yang kelihatan "faham" bahasa manusia.

LLM Popular Hari Ini

💬 ChatGPT (OpenAI)

Yang paling terkenal. GPT-4o boleh faham teks, gambar dan audio. Sesuai untuk penulisan, analisis dan coding. Tersedia di chatgpt.com.

🧩 Claude (Anthropic)

Dikenali selamat dan teliti. Sangat bagus untuk dokumen panjang, analisis mendalam dan penulisan kreatif. Tersedia di claude.ai.

🔷 Gemini (Google)

Terintegrasi dengan Google Search dan produk Google. Bagus untuk maklumat semasa dan pengguna ekosistem Google.

🦙 Llama (Meta)

Model open-source percuma. Boleh dijalankan sendiri tanpa internet. Pilihan terbaik untuk pembangun yang nak kawalan penuh.

Apa Itu Prompt Engineering?

Prompt adalah arahan atau soalan yang anda berikan kepada AI. Kualiti prompt anda menentukan kualiti jawapan AI.

LLM seperti pekerja yang sangat pandai tapi perlu arahan yang jelas. Berikan arahan kabur → dapat hasil kabur. Berikan arahan tepat → dapat hasil cemerlang!

Formula Prompt Berkesan — RCTF

Gunakan formula RCTF untuk prompt yang mantap:

--	Maksud	Contoh
R	Role (Peranan)	"Anda adalah guru Bahasa Melayu berpengalaman..."
C	Context (Konteks)	"Saya pelajar tingkatan 5 yang nak..."
T	Task (Tugas)	"Tulis / Analisa / Ringkaskan / Cipta..."
F	Format	"Dalam bentuk senarai / 5 poin / jadual / 300 patah perkataan"

Contoh Nyata: Arahan Lemah vs Arahan Kuat

Arahan Lemah ❌	Arahan Kuat ✅
Tulis esei tentang AI	Tulis esei 500 patah perkataan tentang kesan AI pada pendidikan di Malaysia, gaya bahasa formal, untuk pelajar universiti
Ringkaskan ini	Ringkaskan artikel ini dalam 3 mata utama, guna bahasa mudah untuk orang awam
Semak tatabahasa	Semak tatabahasa teks ini, kenal pasti kesilapan dan beri cadangan pembetulan

Contoh Prompt Untuk Kehidupan Seharian

Kerja Pejabat

Tulis emel profesional kepada pelanggan yang terlambat bayar invois. Nada sopan tapi tegas. Dalam Bahasa Melayu.

Belajar

Terangkan konsep faedah kompaun seperti saya berumur 15 tahun. Guna analogi mudah dan contoh wang RM100.

Perniagaan 

Saya jual nasi lemak di Johor Bahru. Buat 5 idea promosi media sosial untuk bulan Ramadan dengan bajet rendah.

Kesihatan

Buat jadual makan 7 hari untuk pesakit diabetes jenis 2. Fokus pada makanan tempatan Malaysia yang rendah glisemik.

Hiburan & Pendidikan 

Cipta 10 soalan kuiz tentang sejarah Malaysia untuk pelajar sekolah rendah. Pilihan jawapan A-D untuk setiap soalan.

Teknik Lanjutan Yang Perlu Cuba

1. Berikan AI "Peranan"

Hasil lebih baik apabila AI ada persona yang jelas:

• "Bertindak sebagai doktor pakar..."
• "Anda adalah jurulatih kecergasan..."
• "Bertindak sebagai peguam yang menerangkan kepada orang awam..."

2. Chain of Thought

Minta AI berfikir langkah demi langkah:

"Fikir perlahan-lahan dan terangkan setiap langkah pemikiran anda sebelum bagi jawapan."

Teknik ini sangat berguna untuk soalan matematik, logik, atau analisis kompleks.

3. Berikan Contoh

Tunjukkan format yang anda mahu:

"Tulis dalam gaya seperti contoh ini: [masukkan contoh]"

4. Iterasi dan Perbaiki

Jangan berhenti di jawapan pertama. Teruskan dengan:

• "Buat lebih pendek"
• "Guna bahasa lebih formal"
• "Tambah lebih banyak contoh tempatan"

Peringatan Penting

⚠️ Jangan masukkan maklumat sensitif seperti nombor IC, nombor akaun bank, atau kata laluan ke dalam mana-mana AI awam. Data anda mungkin digunakan untuk latihan model.

Selain itu, AI boleh silap — dipanggil "halusinasi AI". Untuk maklumat penting seperti perubatan, undang-undang, atau kewangan, sentiasa sahkan dengan sumber rasmi.

---

Kesimpulan

Prompt Engineering bukan kemahiran teknikal yang susah — ia adalah kemahiran komunikasi. Semakin jelas anda berkomunikasi dengan AI, semakin berguna jawapan yang anda dapat. Mulakan dengan formula RCTF dan praktis setiap hari.

Dalam post seterusnya, kita akan bincang tentang AI Agentic — generasi baharu AI yang bukan sekadar menjawab soalan, tapi boleh bertindak dan menyelesaikan tugas secara automatik!

---

Ada prompt kegemaran anda? Kongsi dalam komen di bawah! 👇

 

Apa Itu AI dan Bagaimana Ia Berfungsi?

Pernah terfikir macam mana ChatGPT boleh jawab soalan anda dalam masa sesaat? Atau macam mana telefon anda boleh kenal wajah anda? Semua ini adalah kerja Kecerdasan Buatan, atau AI. Jom kita faham dari asas.

---

Apa Itu Kecerdasan Buatan (AI)?

Kecerdasan Buatan atau Artificial Intelligence (AI) adalah kemampuan komputer untuk melakukan tugas yang biasanya memerlukan kecerdasan manusia — seperti memahami bahasa, mengenali wajah, membuat keputusan, dan belajar daripada pengalaman.

Bayangkan anda mengajar anak kecil mengenali kucing. Anda tunjukkan beribu-ribu gambar kucing hingga dia boleh kenal sendiri. AI pun begitu — ia "belajar" daripada data yang banyak hingga boleh membuat sesuatu yang seolah-olah "pintar".

Ingat ini: AI bukan sihir. Ia adalah matematik, data, dan kod komputer yang bekerja bersama-sama untuk meniru cara manusia berfikir.

---

Sejarah Ringkas AI

AI bukan teknologi baru. Ia telah berkembang selama berdekad-dekad:

Tahun	Peristiwa
1950	Alan Turing cipta "Turing Test" — ujian adakah mesin boleh berfikir
1956	Istilah "Artificial Intelligence" dicipta di Dartmouth Conference
1997	Deep Blue (IBM) kalahkan juara catur dunia Garry Kasparov
2012	Deep Learning mula menguasai AI moden
2022	ChatGPT dilancarkan — AI jadi milik semua orang
2024–25	AI Agentic & multimodal menjadi standard baharu

---

Jenis-Jenis AI Yang Perlu Anda Tahu

🎯 AI Sempit (Narrow AI)

Hanya pandai dalam satu tugas sahaja. Contoh: penapis spam emel, sistem cadangan Netflix, atau Face ID pada telefon. Inilah AI yang wujud hari ini.

🧠 AI Am (General AI)

Boleh buat apa sahaja seperti manusia — dalam semua bidang. Masih dalam kajian dan belum wujud sepenuhnya lagi.

⚡ AI Super (Super AI)

Lebih pintar dari manusia dalam semua bidang. Masih teori. Ini yang ramai pakar bimbangkan.

🤖 AI Generatif

Boleh cipta kandungan baharu: teks, gambar, muzik, video. Contoh terbaik: ChatGPT, Midjourney, Suno AI.

---

AI vs Manusia — Apa Bezanya?

Aspek	Manusia	AI
Kelajuan	Lambat, perlu rehat	Sangat laju, 24/7
Kreativiti	Tinggi, inovatif	Terhad pada data latihan
Emosi	Ada perasaan	Tiada perasaan sebenar
Kos	Gaji bulanan	Bayar sekali/langganan
Belajar	Lambat tapi mendalam	Laju tapi perlu data banyak

---

Bagaimana AI Belajar? — Machine Learning

Machine Learning (ML) adalah cara utama AI "belajar". Ia bukan diprogramkan dengan peraturan tetap — sebaliknya, ia diajar dengan contoh dan data.

Contoh mudah: nak ajar AI bezakan kucing dan anjing? Berikan 1 juta gambar berlabel "kucing" dan "anjing". AI akan kesan corak — bentuk telinga, rupa muka, badan — dan belajar bezakan sendiri.

Ada 4 jenis pembelajaran dalam ML:

• Pembelajaran Berselia (Supervised) — AI belajar dari data berlabel. Seperti buku latihan berisi soalan DAN jawapan.
• Pembelajaran Tak Berselia (Unsupervised) — AI cari corak sendiri tanpa label. Seperti kanak-kanak menyusun mainan mengikut warna tanpa diajar.
• Pembelajaran Pengukuhan (Reinforcement) — AI belajar dari ganjaran dan hukuman. Seperti melatih anjing dengan biskut!
• Pembelajaran Pindahan (Transfer) — AI guna pengetahuan dari satu tugas untuk tugas lain. Lebih cepat dan cekap.

---

Neural Network — Otak Tiruan AI

Neural Network adalah struktur dalam AI yang terinspirasi daripada otak manusia. Otak kita ada bilion neuron yang bersambung — neural network ada "neuron buatan" yang buat perkara sama.

Maklumat masuk → diproses melalui lapisan-lapisan → keputusan keluar. Semakin banyak lapisan, semakin "dalam" (deep) rangkaian itu — inilah yang dipanggil Deep Learning.

Analogi mudah: Bayangkan saringan kopi berlapis-lapis. Air kopi (data) melalui setiap lapisan, dan setiap lapisan menapis sesuatu yang berbeza hingga hasilnya sempurna di bawah.

---

Data — Makanan AI

AI tidak boleh berfungsi tanpa data. Semakin banyak dan berkualiti data, semakin pandai AI tersebut. Data yang digunakan termasuk:

• Teks — artikel, buku, laman web, perbualan
• Gambar — foto, lukisan, diagram
• Suara — rakaman, muzik, perbualan
• Video — klip YouTube, filem, berita
• Angka — data jualan, cuaca, harga saham

Tahu tak? ChatGPT telah dilatih dengan data daripada hampir keseluruhan internet awam sehingga trilion perkataan — lebih dari yang mampu dibaca manusia dalam seribu tahun!

---

Kesimpulan

AI adalah alat yang sangat berkuasa, tetapi ia masih memerlukan manusia untuk memimpinnya. Ia bukan sihir, bukan musuh — ia adalah teknologi yang, jika difahami dan digunakan dengan betul, boleh membantu kita bekerja lebih cerdas dan hidup lebih mudah.

Dalam post seterusnya, kita akan bincang tentang Model Bahasa Besar (LLM) seperti ChatGPT dan Claude, dan cara menggunakannya dengan berkesan dalam kehidupan seharian.

---

Kongsi post ini kepada rakan-rakan yang ingin tahu tentang AI! 🇲🇾

AI in 2026: What's Actually Going On?

The hype is real. But so is the confusion. Here's a plain-English breakdown of what's trending in AI right now — and why it matters to you. 

You don't have to be a tech person to notice it — AI is everywhere right now. Your phone, your apps, your workplace, your news feed. But what's actually happening under the hood? Let's break it down, no jargon, no fluff.

The AI Model War is Heating Up

Imagine four blockbuster movies dropping in the same month. That's basically what happened in March 2026 with AI. GPT-5.4, Gemini 3.1, Grok 4.20, and Mistral Small 4 all launched within 23 days of each other. It's intense out here.

Just a couple of years ago, a major AI model would drop once a year and the whole internet would lose its mind. Now? It's every 2–3 weeks. The competition between the big players — OpenAI, Google, Anthropic, xAI — is pushing them to ship faster than ever.

AI That Actually Does Things For You

Up until recently, AI was mostly a "question and answer" machine. You ask, it answers. But in 2026, we're entering the era of Agentic AI — AI that doesn't just talk, it does.

Think of it like hiring an assistant who can read your emails, book your meetings, fill out forms, and send follow-ups — all by themselves. No hand-holding needed. Gartner (a big research firm) predicts that 40% of business software will use these kinds of AI agents by the end of this year.

This is a big deal. We're moving from "AI as a search engine" to "AI as a doer." And that changes everything about how we work.

The Money Is Absolutely Insane

Let's talk numbers for a second, because they're wild. OpenAI has crossed $25 billion in annual revenue. Anthropic (the company behind Claude) is closing in on $19 billion. The global AI market is expected to hit $2.52 trillion in 2026.

For context — that's bigger than the entire GDP of many countries. AI has gone from a nerdy science project to one of the biggest economic forces on the planet in just a few years.

Governments Are Starting to Push Back

With great power comes great responsibility — and a lot of legal headaches. In the UK, a parliamentary committee called generative AI a "clear and present danger" because many AI companies trained their models on copyrighted books, articles, and creative works without permission or payment to the creators.

In the US, the state of Washington passed two big AI bills around disclosure and safety. More countries are expected to follow. The era of "move fast and break things" in AI is slowly coming to an end.

AI Is Helping Cure Diseases

Okay, this one is genuinely exciting. AI isn't just writing emails and making memes — it's being used to discover new medicines. In 2026, several drug candidates that were identified using AI are now entering mid-to-late stage clinical trials, especially for cancer and rare diseases.

The traditional drug discovery process takes 10–15 years. AI is compressing that timeline dramatically. We might be living in the era where some of the biggest medical breakthroughs come with "discovered by AI" in the footnotes.

The Big Picture — What Does It All Mean?

Here's the honest truth: AI is no longer a "future thing." It's a now thing. Whether you're a student, a small business owner, a teacher, or just someone trying to get through the day — AI is either already in your life, or it's about to be.

The question isn't whether to pay attention. The question is: how do you want to show up in this new world? As someone who just uses these tools, or someone who understands them well enough to use them to your advantage?

Because here's the thing — AI doesn't replace people who think critically, who are creative, who understand humans. It replaces tasks. And that's a very different thing.

From WhatsApp Ordering to Full F&B Automation — The Orderla Journey Since 2020

 In 2020, when many businesses were struggling to survive, one thing became clear:

F&B merchants needed a simpler way to accept orders.

That was the beginning of Orderla.

Not as a big tech company.

Not backed by investors.

But built from real conversations with real merchants.

It Started With WhatsApp

At the beginning, many small businesses relied heavily on WhatsApp to take orders.

It was simple.

But it had problems:

• Messages buried during peak hours

• Manual order writing

• Wrong items sent to customers

• No proper sales tracking

• No structured customer database

We saw merchants losing orders — not because customers didn’t want to buy, but because the system couldn’t keep up.

So Orderla started as a structured WhatsApp ordering platform.

It helped merchants:

• Organize incoming orders

• Reduce human mistakes

• Respond faster

• Create better customer flow

That was our first step.

---

From Ordering to E-Commerce

As merchants became more comfortable with digital tools, their needs evolved.

They didn’t just want order collection.

They wanted:

• Online storefronts

• Payment integration

• Product management

• Delivery management

• Sales reporting

So Orderla expanded into a full e-commerce solution under Orderla Commerce.

We focused on helping SMEs digitalize without needing technical knowledge.

No complicated setup.

No heavy systems.

Just practical tools that work.

---

Listening to Merchants (The Real R&D)

Since 2020, one thing has remained constant:

We listen.

Through WhatsApp support, calls, demos, and feedback, we continuously improve based on what merchants actually need.

Not trends.

Not hype.

But real operational problems.

This close relationship shaped what came next.

---

The Birth of Orderla FOS (Food Ordering System)

We noticed a major gap in the F&B industry:

Staff were overwhelmed.

Orders were miscommunicated.

Peak hours were chaotic.

And labor cost kept increasing.

Customers, on the other hand, wanted:

• Faster pickup

• Self-order convenience

• Loyalty rewards

• Cashless payments

So we built Orderla FOS — a web-based ordering system designed specifically for F&B.

It allows customers to:

• Order at their own time

• Customize items

• Pay online

• Skip long queues

While merchants:

• Receive structured orders

• Reduce manual order taking

• Increase order accuracy

• Improve operational speed

This isn’t just a system.

It’s a shift from manual chaos to structured automation.

---

Built for the Future

The F&B industry is changing.

Digital records.

Automation.

Loyalty ecosystems.

Data-driven decisions.

E-invoicing trends.

Businesses that adapt early will grow faster.

Orderla’s mission is simple:

To empower SMEs with practical technology that increases revenue, reduces mistakes, and prepares them for the future.

---

What’s Next?

From WhatsApp ordering…

To e-commerce…

To F&B automation…

Orderla continues to evolve.

Because our journey isn’t about building software.

It’s about building better businesses.

Orderla 6th Anniversary — January 2026

January 2026 marks an important milestone for Orderla.my — six years of building, learning, and growing together with entrepreneurs across Malaysia and beyond.

This anniversary is not just about celebrating time passed, but reflecting on how far we’ve come, the community that shaped us, and where we’re heading next.

⸻

From a Simple Idea to a Growing Platform

Orderla.my began in 2020 with a clear mission: to help SMEs — especially those in the F&B space — operate digitally without the burden of building complex technology from scratch.

What started as a simple online ordering tool quickly evolved as we listened closely to merchants, adapted to real-world needs, and embraced new opportunities. Over time, Orderla.my grew from a startup idea into a reliable partner supporting everyday business operations.

⸻

Growing Beyond the MCO Era

After the MCO period, Orderla.my matured into a more flexible order management platform designed for speed, simplicity, and real-world usage.

As merchant needs became more specialised, the Orderla ecosystem expanded into three focused solutions:

• Orderla.my — a general-purpose, WhatsApp-friendly ordering platform that is quick to set up and easy to use across many business types.

• Orderla Commerce — built for businesses selling physical and digital products, with support for multiple variants and detailed product management.

• Orderla FOS (Food Ordering System) — designed specifically for F&B brands, complete with a customer portal and features suitable for HQs managing multiple outlets.

Each platform serves a different business need, while sharing the same goal: making digital ordering practical and accessible.

⸻

The Heart of Orderla: Our Community

What truly makes this journey meaningful is our community — merchants, partners, and early adopters.

From roadside stalls to busy cafés, from solo founders to growing restaurant groups, your feedback has shaped every feature we build. Many of the improvements in Orderla exist because users asked for them. Supporting businesses from their first online order to hundreds of daily transactions has been both humbling and motivating.

⸻

What We’ve Built So Far

Over the past year, Orderla.my has continued to strengthen its foundation:

• More Flexible Ordering Tools

Improved menus, customizable checkout experiences, and better order flow control to simplify daily operations.

• Customer-Driven Enhancements

Features built directly from merchant feedback, including better reporting, order reminders, and faster support.

• A Reliable Backbone for Growth

A stable system designed to perform during peak hours, festive seasons, and periods of rapid business growth.

⸻

Looking Ahead: The Next Chapter

Anniversaries are also about vision. The next phase of Orderla.my is already taking shape:

1. Orderla FOS — Scaling F&B Digitally

An advanced web-based food ordering system inspired by large brands like ZUS, but built to be accessible, cost-effective, and merchant-friendly — without middleman fees.

2. Tech + Consultancy Partnerships

Opening our technology through partnership frameworks, allowing consultants and businesses to deploy Orderla while we continue strengthening the core platform.

3. Platform Resilience & Independence

Reducing reliance on external communication channels by strengthening native platform capabilities, ensuring long-term stability for merchants.

4. Knowledge & Growth Initiatives

Beyond software, we aim to support merchants through learning resources, workshops, and shared insights to help them grow sustainably.

⸻

Thank You

Six years of growth wouldn’t be possible without the trust and support of our users, partners, and team.

Thank you for believing in Orderla.my, for your feedback, your patience, and for building this journey with us.

The next chapter is just beginning 🚀