🤖

Series · Part 8 of 21

Foundation
AI Demystified
Abhishek Saha
Abhishek Saha · 🤖 AI / ML

How AI Reasons (And Why It Sometimes Breaks)

o1, o3, DeepSeek-R1 — reasoning models behave differently. What chain-of-thought actually is, what 'thinking longer' means, and where it still fails.

#ai #reasoning #chain-of-thought #llm #o1 #test-time-compute

Standard LLMs generate an answer immediately and can’t revise. Reasoning models write out intermediate steps first — and that small change makes a surprising difference on hard problems.

Same problem · three approaches
Model generates answer immediately. Fast, but commits early — errors in the first tokens compound.
user →
A bat and a ball cost $1.10 total. The bat costs $1.00 more than the ball. How much does the ball cost?
model →
The bat-and-ball problem — a classic CRT (Cognitive Reflection Test) item that trips up both humans and direct-answer LLMs.

Writing steps doesn’t mean AI is “thinking.” It means the intermediate text creates a richer context that makes correct final tokens more probable. The effect is real; the introspection isn’t.

Next up: You know how AI works. Part 9 is about how to talk to it — five prompt engineering techniques that cover 95% of real-world use cases.

AI Demystified · 16 of 21 published

  1. 0 Grounding 5 Mental Models You Need Before Diving Into AI
  2. 1 Foundation What Happens When You Ask AI Something?
  3. 2 Foundation Transformers — The Architecture That Changed Everything
  4. 3 Foundation How AI Learns, Thinks, and Decides
  5. 4 Foundation How AI Reads Your Words
  6. 5 Foundation Why AI Forgets
  7. 6 Foundation Why AI Lies (And Doesn't Know It)
  8. 7 Foundation What AI Cannot Do
  9. 8 Foundation How AI Reasons (And Why It Sometimes Breaks)
  10. 9 Practice Prompt Engineering — How to Talk to AI
  11. 10 Practice Embeddings & Vector Databases — The Memory Layer of AI
  12. 11 Practice RAG Explained — How AI Knows What You Didn't Train It On
  13. 12 Practice Fine-tuning vs. Prompting — When to Use Which
  14. 13 Practice Do You Really Need GPT-4?
  15. 14 Practice Latency, Tokens, and Cost — The Physics of AI Products
  16. 15 Practice How Do You Know AI Is Actually Working?
  17. 16 Hands-On Coding Setup — Your AI Development Environment soon
  18. 17 Hands-On MCP Tool Calling — How AI Uses Tools soon
  19. 18 Hands-On AI Agents — Beyond Chatbots soon
  20. 19 Hands-On Build Your First Real AI App soon
  21. 20 Hands-On Token Optimization — Spend Less, Get More soon
← Part 7 What AI Cannot Do Part 9 → Prompt Engineering — How to Talk to AI

Related posts

Enterprise Master Data Cleaning: The Business Case for SAP Supplier Data
AI / ML ·

Part 1 — Enterprise Master Data Cleaning

Enterprise Master Data Cleaning: The Business Case for SAP Supplier Data

Why dirty vendor master data quietly costs enterprises millions, how AI helps, and the operating model that makes it safe. Part 1 of the series, for business and leadership.

Read
master-data sap data-quality enterprise
Embeddings and Vector Search, Step by Step: Finding Duplicate Suppliers
AI / ML ·

Part 2 — Enterprise Master Data Cleaning

Embeddings and Vector Search, Step by Step: Finding Duplicate Suppliers

Learn embeddings, distance, vector databases, semantic search and a mini RAG one step at a time, by finding duplicate suppliers in a small dataset.

Read
master-data sap data-quality embeddings
newsletter

Get new posts in your inbox

No spam. No digest. Just a note when I publish something new.

Discussion