llmcamp

#100DaysOfAgentMemory

100 days of agent memory. Annotated, in one book.

I posted about agent memory for 100 days on LinkedIn while the industry decided it was the problem of the decade. This is the series with the part LinkedIn couldn't hold: what held up, what I got wrong, and what I know now. Get Days 1–10 free.

Free 40-page sample · joins the llmcamp newsletter · unsubscribe anytime

  • 100 days
  • 106 posts
  • 5,000+ reactions
  • 1 DeepLearning.AI course
Book cover for 100 Days of Agent Memory by Richmond Alake
40 pagesDays 1–10Annotated

The receipts

It happened in public. All of it.

Eight moments from the series. Open any post to see the original conversation.

LinkedIn post, Day 1: why agent memory matters, 214 reactions
Day 01 Why agent memory matters
LinkedIn post, Day 6: RAG is the beginning, not the end of agent memory, 116 reactions
Day 06 RAG is the beginning, not the end
LinkedIn post, Day 50: prompt to context to memory, 114 reactions
Day 50 Prompt → context → memory
LinkedIn post, Day 59: the memory core in practice, 77 reactions
Day 59 The memory core in practice
LinkedIn post, Day 62: sessions, context, and memory, 97 reactions
Day 62 Sessions, context, and memory
LinkedIn post, Day 87: Memory is the battleground of 2026, 99 reactions
Day 87 Memory is the battleground of 2026
LinkedIn post, Day 98: teaching memory-aware agents, 189 reactions
Day 98 Teaching memory-aware agents
LinkedIn post, Day 100: agent memory predictions, 124 reactions and 36 comments
Day 100 One subject. One hundred days.

The first ten days, revisited

What's in the free sample

Days 1–10, annotated.

Every original post, plus 500-word commentary written with hindsight: the receipts, the corrections, and what each idea grew into.

The frameworks behind the series.

The memory ladder—prompt → RAG → context → harness → memory engineering—the three lenses of agent memory, and why RAG is the beginning, not the end.

The origin story.

How MemoRizz began, and the bet that memory is the last job to be done on the path to reliable, believable, and capable agents.

Portrait of Richmond Alake
Richmond Alake

About the author

Memory engineering, from the person naming the work.

Richmond Alake is the creator of Memory Engineering and Director of AI Developer Experience at Oracle. He teaches agent memory through a DeepLearning.AI course and the quarterly O'Reilly AI Agent Memory Management Bootcamp, and built MemoRizz, the open-source memory framework used throughout the book.

O'ReillyDeepLearning.AI

A look at the format

Original post. Fresh annotation. One clear takeaway.

The sample is designed to be marked up, argued with, and put to work.

Days 1–10

Agent
Memory

Annotated field notes

Richmond Alake
DAY 06ORIGINAL + ANNOTATION

RAG is the beginning, not the end.

THE POST

Retrieval is the familiar entry point. Memory adds a write path, time, and a lifecycle.

100 Days of Agent Memory · 18
THE TAKEAWAY01 / 10
Memory starts where the current session stops.
Prompt→RAG→Context→Harness→Memory

The horizon keeps getting longer. The engineering discipline has to follow it.

100 Days of Agent Memory · 31

Your copy is one confirmation away

Read the first ten days tonight.

Start with the posts that opened the series. Stay for the annotations LinkedIn couldn't hold—and get the launch price when the complete edition ships.

joins the llmcamp newsletter · unsubscribe anytime

Before you subscribe

A few honest answers.

Is the sample really free?

Yes. You get the 40-page PDF by email, and you join the llmcamp newsletter. Unsubscribe anytime and keep the sample.

What happens after I subscribe?

Confirm your email with one click, the sample lands in your inbox, and you'll get a couple of follow-up emails with the best of the series. When the full edition launches, subscribers get the launch price.

Who is this for?

AI engineers, tech leads, and founders building agentic systems who want the memory layer treated as engineering, not vibes.