THE BUILDER’S READING LIST
A workflow, an agent,
or just a good script?
Start with the decision. Understand the trade-offs before adding another tool.
Read the field guideUNDERSTAND→COMPARE→PUT TO WORK
THE PRACTICAL AI INDEX
Less noise.
More useful AI.
Understand what matters. Find the right tools.
Turn a good question into your next useful step.
Original sourcesClear limitationsNo mystery scores
A workflow, an agent,
A good place to begin.
Clear explanations and practical decisions, with the sources attached.
Guide
Read an AI launch without inheriting its hype
Turn a model announcement into a small decision record: what shipped, under which conditions, and what you still need to test.
Lab roadmapsPeople & claims
Source-based4 min ↗
Explainer
Open weights, open code, open enough for what?
Inspect the actual artifacts and permissions behind an openness claim instead of treating open as a single product feature.
Open models & licensesData & training
Source-based4 min ↗
Guide
Your local AI assistant may still use the internet
Draw the model, retrieval, tool and logging boundaries before assuming that a local application is fully offline.
Local & edge AIData flow
Source-based4 min ↗
Explainer
Memory, context and training are different levers
Locate where information is stored and how it reaches the model before choosing a memory feature, retrieval system or training job.
Context & memoryTraining & adaptation
Source-based4 min ↗
Research note
Test the plan, not the confidence of the answer
Turn an impressive reasoning demonstration into a task with explicit states, constraints, perturbations and a checkable result.
Reasoning & planningBenchmarks & methods
Source-based4 min ↗
Guide
A real citation can still support the wrong claim
Audit a short AI-generated answer at the claim, passage and source-family levels. A working link is only the beginning.
Sources & evidenceResearch workflow
Source-based4 min ↗
Follow a useful thread.
A little structure for the things you want to understand.
01
Read AI claims clearly
Move from a launch headline to a scoped, source-backed decision without treating every claim as a benchmark result.
02Design an agent you can stop
Understand information flow, choose a simple control structure and specify human review before connecting real actions.
03Evaluate before adopting
Define observable success, read productivity evidence carefully and bring those habits to code and voice workflows.
Know what a claim rests on.
Source-based notes are not hands-on tests. We keep that distinction visible.
Inside the notebook.
Changes to this library, not an automated industry news feed.
Established the eight-track knowledge base
Organized ongoing research into eight tracks and 41 subtopics, with source identity, evidence limits and open questions kept together.
Connected technical depth with the wider AI landscape
Added industry, ecosystem, multimodal and adoption context around the technical core. Media production and business validation retain their separate responsibilities.
Separated source review from performance claims
Revisited a selected source set and recorded limited-use, hold and routing decisions. Proposed experiments remained proposals, not reported results.