AI Infographic
One A2 sheet, one idea. Each poster is a print-first map of how an AI system actually behaves — hooks, context, loops, gates — so you can pin it on a wall and still read it from a chair.
Not a slide deck and not a blog essay. The job is glanceable structure: what fires when, what is stable vs what changes each turn, and where a human still decides.
On this site you get a WebP preview and a PDF to download. The authoring art stays off-site; what ships here is the finished sheet.
We need an AI IDE - buy Cursor?
Same job, different room…
We need RAG - buy a vector DB?
Same shelf, different job…
What happens next…
That prompt, through hooks and context, to the message you see.
Pipeline
Topics to do next
Queued A2 sheets — one idea each. Propose something missing on the right.
Embedding and Chunking
How text becomes vectors — and why the cut points decide what retrieval can find.
Claude vs ChatGPT
Same job, different product shapes: context, tools, and what each system refuses to pretend.
What the hell is LoRA?
Low-rank adapters in one sheet — what changes, what stays frozen, and when it beats full fine-tunes.
Context windows
Token budgets as a hard wall: what fits, what gets dropped, and why “longer” is not free.
Prompt cache
What stays warm across turns — stable prefix vs one-shot noise — and who pays for a miss.
Agents vs workflows
A loop that chooses tools is not a flowchart. When autonomy helps, and when it just hides the branch.
Fine-tuning vs RAG
Weights that remember vs documents you retrieve — which lever moves facts, style, or both.
Quantization
Fewer bits, smaller boxes: what 8-bit and 4-bit actually trade for speed and VRAM.
Tool calling
Who proposes the call, who may run it, and where human gatekeeping still sits in the loop.
Evals that matter
What “good” means when demos lie — a poster-sized map of checks that catch real failure modes.