TIDESa civilization in progress

Articles & research

What we're learning,
written down.

Long reads, design papers and short explainers on the technology we teach and build with.

  • In the library24 pieces
  • Long reads10
  • Design papers7
  • TipPoint at a card to preview it

The library

Read something worth your time.

Papers describe how we design and build things. They report designs, not experiments.

Showing 1–6 of 24

PaperLong read10 minMulti-agent systems, explained with a newsroom: roles, handoffs, and when one agent is enoughA multi-agent system splits one job between several AI agents, each with a narrow role. This paper uses a newspaper newsroom to explain how the roles, handoffs and controls work, what each extra agent costs, and how to tell when a single agent with good tools is the better design.Multi-agentRead →ArticleLong read10 minFrom Matric to building AI agents: what to learn, in what orderA suggested self-study order for a student in Pakistan who wants to go from basic computer skills to building working AI agents. Each stage has one small project, and the order matters more than the speed.LearningRead →PaperLong read9 minWhy AI agents fail in production: nine failure modes and how to design them outA design note on the nine ways an AI agent most often goes wrong once real people start using it. For each one we describe what it looks like, why it happens, and the design decision that removes it.AI agentsRead →ArticleLong read9 minWhere the money goes in a software project (and how to avoid paying twice)A plain account of what a client actually pays for in a software project, from the first conversation to the years after launch, and the handful of decisions that stop you paying for the same work twice.SoftwareRead →PaperLong read9 minRetrieval done properly: chunking, search and the mistakes that make agents lieA technical note on retrieval-augmented generation in plain words. It follows a question from the document store to the final answer and shows, at each step, the mistake that leads an agent to say something false with confidence.RetrievalRead →ArticleLong read9 minAI and jobs: which tasks change first, and what stays humanAI does not replace jobs in one piece. It changes particular tasks inside them. This article reasons through which tasks change first, which change later, and what a student or worker in Pakistan can sensibly do this year.WorkRead →

Want to learn this properly?

The AI Automation & Agentic Mastery bootcamp teaches these ideas hands-on, over four months.