AI Learning
Success is the sum of small efforts, repeated day in and day out.
September 2026
8 days studied
AI Learning, Day 8: First Agent, Deployed
Built and deployed my first real agent with smolagents — CodeAgent, custom tools, live on Hugging Face Spaces.
AI Learning, Day 7: Building the Loop by Hand
A dummy agent in Colab with no framework — and the one parameter that keeps the model from hallucinating its own tool results.
AI Learning, Day 6: Observations Close the Loop
The last piece of the Thought-Action-Observation cycle — how results get fed back as the agent's only form of 'memory.'
AI Learning, Day 5: The Thought-Action-Observation Loop
How Thought, Action, and Observation fit together into the loop that actually drives an agent.
AI Learning, Day 4: Tools Are Just Functions With a Description
How agents actually take action in the world — a tool is a function, a name, and a description the LLM can read.
AI Learning, Day 3: Agents Don't Actually Remember
Messages and special tokens — the illusion of memory is just the full conversation getting re-sent every time.
AI Learning, Day 2: Tokens, Revisited
Reading about how LLMs work — the token concept ties directly back to an NLP class from my master's.
AI Learning, Day 1: What Even Is an Agent?
Starting a formal, systematic study of AI — day 1 covers the basics of agents from Hugging Face's Agents Course.