A LITTLE MORE PERSPECTIVE
Find your next read.
Search the collection by title, subject or format.
59 stories
What is a large language model?
How the system behind AI assistants learns and writes its answers.
What are AI agents? How they plan and use tools
How agents choose steps, use tools and where their limits lie.
How to start using AI at work
Start with one task, then measure the work left afterwards.
How AI image generators work
From a written brief to a picture, and where control is lost.
AI and jobs: what the evidence says
What exposure, productivity and hiring data really say about work.
Open vs. closed AI models: what the difference means
What “open” really means, and how to choose between the two.
What AI benchmarks actually measure
What benchmark scores capture, and how to test models on your work.
Claude review: research and writing workflows
Claude for research and writing: plans, citations, limits and data.
Claude vs. ChatGPT vs. Gemini: choosing an assistant
Three assistants compared on price, context, tools and data terms.
ChatGPT review: everyday work and limitations
ChatGPT's plans, features and documented shortcomings for everyday work.
How reasoning models work, and what changes
Why some models think before answering, and what it costs you.
Context windows explained: what fits and what gets lost
What fills a model's working space, and what gets overlooked.
What is RAG? Retrieval-augmented generation explained
How AI looks things up before answering, and where it breaks.
AI hallucinations explained: why plausible answers fail
Why models make confident errors, and the checks that catch them.
Claude vs. ChatGPT for long-document work
Which assistant handles long reports and contracts, and how.
ChatGPT vs. Gemini for spreadsheet analysis
Which assistant shows its working on spreadsheet analysis.
How to judge a new AI model release
A checklist for reading launch posts, scores and system cards.
Why AI agents fail, and how to diagnose the failure
Trace a bad answer back through the run before you fix it.
Build your first AI agent without code
Design one small, supervised agent, step by step.
AI agent vs. AI workflow: how to choose
The real difference is who decides the next step.
Build an ecommerce product-question agent
A supervised agent that answers shoppers from your own catalogue.
Build a research agent with source verification
An agent that drafts briefs in which every claim is sourced.
Build a SaaS onboarding investigation agent
A read-only agent that finds why new accounts stall.
Build a growth reporting agent
A weekly reporting agent whose every number traces to a query.
Build an invoice-exception investigation agent
An agent that investigates failed invoice matches, with people approving.
What AI agent demos leave out
The questions to ask before you buy or build from a demo.
LangChain's role in AI agent development
From open-source library to platform company: what LangChain sells.
Vapi's approach to voice agents
How Vapi wires AI phone agents together, and charges for them.
What AI agent platforms disclose about permissions
Ten agent platforms' permission documentation, read against ten criteria.
How to verify AI-generated research
A practical way to check AI citations before you rely on them.
Perplexity vs. ChatGPT for research
Two cited-answer tools compared on sources, control, price and data.
Cursor vs. GitHub Copilot
A standalone editor against a plug-in assistant, compared feature by feature.
Best AI coding assistants for defined coding tasks
Seven coding tools sorted by job, permissions and cost.
AI for marketing: useful tasks and required checks
Where AI helps a marketing team, and the check each task needs.
AI for ecommerce: practical starting workflows
Five low-risk AI workflows a small online shop can start now.
Perplexity review: research quality and limitations
Perplexity's plans, citations and what independent checks found.
Cursor review: coding workflow and review effort
How Cursor's features fit together, and the review work left.
GitHub Copilot review: coding tasks and limitations
Copilot's plans, agents and limits after the move to credits.
Claude Code review: from task to reviewed change
One coding task followed from brief to merge in Claude Code.
Claude Code vs. Cursor: choosing a coding workflow
Terminal agent or agent-first editor: where review happens and what it costs.
Best AI research tools for evidence-led work
Seven research tools sorted by job, sources and cost.
How to measure AI productivity without hiding review work
Count the checking and fixing, not just the draft time saved.
Glean and the workplace AI search problem
How Glean searches company knowledge without breaking sharing rules.
What workplace AI studies actually show
What the best workplace studies measured, and how far it travels.
Best AI video generators for creator workflows
Six video tools sorted by job, credits and commercial terms.
How to keep characters consistent in AI video
A tool-neutral workflow for keeping characters the same across shots.
How AI video generators work
How video generators handle time, and why hands and physics fail.
Build an AI voiceover workflow with review
A reviewed process for synthetic narration you can stand behind.
Midjourney vs. Stable Diffusion for creator control
Hosted polish against open weights: control, licence and risk.
Best AI image generators for creator workflows
Seven image tools compared on cost, ownership and legal cover.
Runway review: usable video outputs and editing
Runway's models, credits, clip limits and terms, examined.
ElevenLabs review: voice production and editing
ElevenLabs for narration and dubbing: models, costs and voice rules.
ElevenLabs and the AI voice business
What the voice AI company sells, charges and safeguards.
The EU AI Act explained
Risk tiers, roles and deadlines in Europe's AI law.
How to spot AI-generated content: signals and limits
Why visual tells fail, and four origin checks that work better.
Content credentials explained: what provenance can show
What a signed provenance record can prove, and what it cannot.
AI training and copyright: the important distinctions
Four separate copyright questions, and where each stands.
AI detectors vs. content credentials
Detectors guess; credentials record. Why the difference matters.
Who is accountable when an AI agent acts?
How responsibility splits when software takes an action.