Dense Slides: Turn 10,994 Notes Into Memory - Paul Iusztin, Decoding AI & Louis-François Bouchard, Towards AI

Source Video

Turn 10,994 Notes Into Memory - Paul Iusztin, Decoding AI & Louis-François Bouchard, Towards AI

Method

This deck is slide-only. The existing captured video frame set supplies candidate frames, then local OpenCV rejects sponsor/title/speaker-only frames, crops visible slide surfaces, deduplicates, and saves the cropped slide images.

Cropped Visible Slides

Recreated text/layout view: open HTML recreation
AI slide classifier: content_slide confidence 0.98
Text source: agent_vision.

Slide text:

My Second Brain

5,489 notes in Obsidian

5,505 files in Readwise

Plus, Notion, Google Drive.

Growing ~250 files per month

Recreated text/layout view: open HTML recreation
AI slide classifier: content_slide confidence 0.98
Text source: agent_vision.

Slide text:

My OS didn't start from zero

Years of notes — every video and course I've made, all moved into Obsidian

Meeting recaps I save after every call -> Obsidian

Highlights from posts, tweets, and articles -> Obsidian

Even my agent skills -> Obsidian

Recreated text/layout view: open HTML recreation
AI slide classifier: content_slide confidence 0.99
Text source: agent_vision.

Slide text:

Inside v1: fan out, rank, scrape

Topic

- golden links

Scrape links

seed context

Main agent

Query rounds

Gems grounded in Google

subagent Q1

subagent Q2

subagent Q3

subagent Q4

subagent Q5

subagent Q6

x3 rounds

Recreated text/layout view: open HTML recreation
AI slide classifier: content_slide confidence 0.99
Text source: agent_vision.

Slide text:

Forget the infrastructure you think you need

No vector database. No knowledge graph.

No semantic search. No text search.

Recreated text/layout view: open HTML recreation
AI slide classifier: content_slide confidence 0.95
Text source: advanced OCR rapidocr-live/bright-screen/contrast.
OCR decision: ready — Small embedded code/screenshot text on the right is dense and better suited for OCR.

Slide text:

No database, just the index

cta 0

1

tuIce s

1 1

VT11k1/IeOs/agt·mry/AROHITECTE.d

reads first index.yaml g1thub://re04).1abs/agn.e0s01y9ae78c79940 1k1/sovIces/agent·meory.o

catalog + summaries origin:gitnub

Agent _c0*2026-05-17 D.blicatio- acpliodtoaxeteoeoiy. cite 1MIy n004j.labs 1snA0a(0·2.0Pyton1ib1 meieaid flcachart tR suogi

Recreated text/layout view: open HTML recreation
AI slide classifier: content_slide confidence 0.96
Text source: advanced OCR rapidocr-live/bright-screen/contrast.
OCR decision: ready — The file-tree and graph labels are small and dense, so OCR is likely better than manual transcription.

Slide text:

How does the wiki look like?

research-coding-agent-architectures

raw corteat-compaction ARCHTECTURIARCHITECTURE

wiki

comparisons

concepts tool-rogistry

entities do01-000 cvervow

notes sposuado

repos aqer!-ptrrtittion-low-opencoce-vs- -hermtn-agtns

sources caude-code

open-questions subogerts-architecture-cpencode iratruction-fles

synthesis overview mcp sandboaing ARCHITECTURE

log index TNTIO EYST quntiors

Recreated text/layout view: open HTML recreation
AI slide classifier: content_slide confidence 0.98
Text source: agent_vision.

Slide text:

Now, how do we query this wiki?

Recreated text/layout view: open HTML recreation
AI slide classifier: content_slide confidence 0.98
Text source: agent_vision.

Slide text:

Every question leaves a trace

Recreated text/layout view: open HTML recreation
AI slide classifier: content_slide confidence 0.98
Text source: agent_vision.

Slide text:

One immutable PARA snapshot

Recreated text/layout view: open HTML recreation
AI slide classifier: title_card confidence 0.99
Text source: agent_vision.

Slide text:

Want to build systems like this?

Agent Engineering: Building Multi-Agent Systems

Towards AI Academy

- the engineering behind this repo

- agent workflows, tool use, memory

- multi-agent systems & orchestration

- evals & production-grade agents

Hidden Non-Slide Evidence

`slide-002.jpg` — title_card confidence 0.97; speaker intro card
`slide-003.jpg` — title_card confidence 0.97; speaker intro card

Classification audit: raw/sources/slide-ai-classification/dense/ZRM_TfEZcIo/audit.json