Uploaded July 2026 | Updated September 2026, 2 weeks ago
Ever wonder how AI coding engines actually work behind the scenes? In this webinar clip, we break down the foundational mechanics of Large Language Models (LLMs) and how they function as the base for today's coding engines.
Miki Tebeka explores why these generative AI models operate with inherent randomness—often compared to pulling a slot machine—and why they are exceptionally good at generating code due to the repetitive nature of programming. Despite their power, LLMs lack real thinking or reasoning, which can occasionally lead to hilarious logic fails or subtle, hard-to-catch coding mistakes. Crucially, because LLMs do not possess memory, they rely entirely on the active conversation's "context" to function.
He touches on why supplying an AI with a massive context window (such as a 100,000-line codebase) actually degrades its efficiency, causing the model to focus primarily on the edges of the prompt. Ultimately, you'll learn why context engineering (the practice of limiting the AI's focus to only the most meaningful parts of your code) is an essential skill for getting the best results.
Topics covered:
● How LLMs predict text and the inherent randomness of Generative AI
● Why LLMs excel at repetitive tasks like coding but sometimes fail at basic logic
● The memory limitations of LLMs and how context windows work
● Why large context windows degrade AI efficiency
● The importance of context engineering and limiting code scans to meaningful snippets
---
Explore more from Ardan Labs
Online Courses: ardanlabs.com/education
Live Training Events: ardanlabs.com/live-training-events
Technical Blog: ardanlabs.com/blog
Github: github.com/ardanlabs
Connect with Ardan Labs
Website: ardanlabs.com
X: https://x.com/ardanlabs
LinkedIn: linkedin.com/company/ardanlabs
#llm #generativeai #contextengineering #aicoding #softwaredevelopment
Ever wonder how AI coding engines actually work behind the scenes? In this webinar clip, we break down the foundational mechanics of Large Language Models (LLMs) and how they function as the base for today's coding engines.
Miki Tebeka explores why these generative AI models operate with inherent randomness—often compared to pulling a slot machine—and why they are exceptionally good at generating code due to the repetitive nature of programming. Despite their power, LLMs lack real thinking or reasoning, which can occasionally lead to hilarious logic fails or subtle, hard-to-catch coding mistakes. Crucially, because LLMs do not possess memory, they rely entirely on the active conversation's "context" to function.
He touches on why supplying an AI with a massive context window (such as a 100,000-line codebase) actually degrades its efficiency, causing the model to focus primarily on the edges of the prompt. Ultimately, you'll learn why context engineering (the practice of limiting the AI's focus to only the most meaningful parts of your code) is an essential skill for getting the best results.
Topics covered:
● How LLMs predict text and the inherent randomness of Generative AI
● Why LLMs excel at repetitive tasks like coding but sometimes fail at basic logic
● The memory limitations of LLMs and how context windows work
● Why large context windows degrade AI efficiency
● The importance of context engineering and limiting code scans to meaningful snippets
---
Explore more from Ardan Labs
Online Courses: ardanlabs.com/education
Live Training Events: ardanlabs.com/live-training-events
Technical Blog: ardanlabs.com/blog
Github: github.com/ardanlabs
Connect with Ardan Labs
Website: ardanlabs.com
X: https://x.com/ardanlabs
LinkedIn: linkedin.com/company/ardanlabs
#llm #generativeai #contextengineering #aicoding #softwaredevelopment
