The "Thinking" Machine

Transcript

1. The “Thinking” Machine Naufal Hakim
2. NAUFAL HAKIM 👋 Software Engineer PT Indonesia Satu Tujuh instagram.com/@falll.hkm linkedin.com/in/naufalhkm tiktok.com/@kimnopal Pemecut AI
3. HOT TOPIC
4. AI AI AI Antek asIng
5. Why is AI so HOT ? AI AI AI AI AI AI LLM LLM LLM Large Language Model AI Provider systems that can process large amounts of text can understand and reproduce human language
6. Simi-simi VS LLM
7. Planning/Reasoning Tools Memory Evolution of LLM AI Agent LLM yang lebih sakti
8. Normal LLM vs AI Agent Prompt Answer Planning the steps Prompt Recall user preference from memory Using available tools Reasoning the goals Answer
9. Normal LLM vs AI Agent Cariin tiket murah ke Bali buat minggu depan dong Kamu bisa cek Traveloka, Tiket.com, atau Google Flights. Pilih tanggal fleksibel agar harga lebih murah.
10. Normal LLM vs AI Agent 1.Cari tiket murah 2.Tujuannya ke Bali 3.Waktunya minggu depan Reasoning 1.Membuka website pencarian tiket. 2.Mengecek beberapa tanggal. 3.Membandingkan harga. 4.Memilih opsi terbaik. 5.Memberikan rekomendasi. 6.Bahkan bisa lanjut membuat itinerary atau reminder. Planning + Recall Memory Tools Menggunakan browser tools Menggunakan calender tools Cariin tiket murah ke Bali buat minggu depan dong Prompt Answer .......
11. Reasoning 101 How Agent Think? Goals Apa tujuan akhirnya? Data apa yang dibutuhkan? Apakah perlu memakai tools? Langkah apa yang harus dilakukan? Context Apa tujuan akhirnya? Data apa yang dibutuhkan? Apakah perlu memakai tools? Langkah apa yang harus dilakukan? Evaluate Apakah data ini relevan? Apakah hasilnya cukup? Apakah perlu mencari lagi? Apakah ada error?
12. Tools 101 LLM Tools MCP Skills Model Context Protocol Analogy
13. Tools 101 MCP Model Context Protocol
14. Tools 101 MCP Architecture
15. Tools 101 Skills Reusable agent capabilities (.md file) system prompt keeps growing Behavior is inconsistent Prompts live in code Knowledge is not reusable The Problems
16. Tools 101 Skills https://github.com/obra/s uperpowers/blob/main/ski lls/brainstorming/SKILL.md
17. Multi Agent System “a group of agents working together” Single agents have limitations: Cognitive overload on complex tasks Jack-of-all-trades, master of none Difficult to maintain context across diverse domains Hard to parallelise work The Problems
18. RUDI (Research Utility for Data and Information) Research Agent Tugas: Mencari referensi SARI Summarization Agent for Reading Information Tugas: Meringkas jurnal Summarizer Agent CICI Citation Intelligence for Cataloging Information Tugas: Mengatur daftar pustaka Citation Agent WATI Writing Agent for Text and Ideas Tugas: Membuat draft laporan Writer Agent BUDI Balanced Understanding and Document Inspector Tugas: Mengecek grammar Reviewer Agent Multi Agent System SKS Agents System Skripsi Kebut Semalam
19. Agent RUDI Agent SARI Agent WATI Agent CICI Agent BUDI Multi Agent System SKS Agents System Prompt Coba dong buatin laporan soal IoT buat topik smart farming. Result
20. Multi Agent System Architecture
21. YES and NO So, Will Engineer Really Be Replaced? Will replace those who can't use AI those who can use AI
23. THANK YOU