AI

Explore how AI industry growth can translate into profits for companies and returns for shareholders. Compare demand, price competition, margins, payback, financing, and dilution in a hypothetical valuation scenario, then use a worksheet to test your assumptions at the next earnings release.

Read Microsoft's annual financial statements to distinguish revenue growth, margins, capital spending, cash flow, and customer concentration.

Trace a document request through retrieval and inference to see how compute, memory, networking, power, cooling, throughput, and utilization affect costs.

Compare 12 months of costs for an internal AI document service, including setup, subscriptions, maintenance, usage, and migration.

Separate model quality from customer outcomes, then evaluate data, workflow, permissions, and switching costs in a hypothetical document service.

Compare flat-rate, usage-based, and hybrid pricing for a hypothetical document service. Set included usage, handle failures and trials, and check the margin on extra tasks.

Calculate the cost of a successful AI task from model calls, retries, and incremental expenses, and see how caching and quality change the result.

Separate VAT, development subscriptions, infrastructure, and usage costs to calculate contribution margin and break-even for a hypothetical AI service.

Use a search feature to connect task requests, contracts, context, model choice, implementation, verification, handoff, and PR review with Claude and Codex.

Split coding tasks between Claude and Codex and isolate changes with Git worktrees. Define file ownership and API contracts, check shared ports and databases, and integrate results sequentially. Learn to resolve conflicts, verify the integrated feature, and compare time and cost.

Learn how to resume long-running work with Claude and Codex. Record the goal, code state, verification evidence, and next action, then check them against the repository. Understand how conversation compaction differs from resuming a session and how to check uncommitted changes and stale records.

Learn how prompt caching can reduce the cost of repeated input with Claude and Codex. Arrange stable instructions and task data, interpret OpenAI and Anthropic usage, and diagnose misses. Compare the controls available in APIs and coding tools, then verify the impact on total cost and quality.

Measure Claude and Codex token usage alongside actual development efficiency. Track completion, human rework, time through verification, and cost per completed task. Include failures and stops, compare under the same conditions, and use the results to choose improvements.

Verify code changes made by Claude and Codex, from reproducing the failure through type checks, tests, and browser behavior. Use a search filter example to write a regression test, report failures, decide when to stop retrying, and document what was checked and what remains unverified.

Choose models and reasoning effort based on the difficulty and risk of an AI coding task. Define starting routes and escalation rules for quick edits, routine implementation, and complex design or verification, then compare Claude and Codex on cost, latency, and quality.

See when skill descriptions, instructions, MCP tool definitions, and tool responses enter an agent's context. Compare Claude Code and Codex loading patterns and reduce token usage by loading only what the task needs.

Give AI coding agents a clear goal, scope, constraints, completion criteria, verification steps, and boundaries for decisions. Use a practical request template to improve implementation and reduce review rounds.

Organize repositories so coding agents can find the information each task needs. Separate AGENTS.md, task documentation, and scoped directory guidance to reduce rework and connect shared and local rules in monorepos.

Design AGENTS.md and CLAUDE.md without duplicating rules. Separate shared and tool-specific guidance and use templates that fit how Codex and Claude Code load instructions to reduce context costs and rework.

Reduce token usage and rework when developing with Claude and Codex. Define task scope, organize repository instructions, narrow tool output, and build a verification loop with practical templates, a checklist, and comparison metrics.

Use ETA as a fairness signal for meetinthemiddle recommendations. Build ETA distribution, meeting purpose, category, and time features. Compare feature sets with NDCG and consider correlation, feature importance, and deployment checks.

Build a meetinthemiddle recommendation dataset around a Station, Place, and Edge schema. Follow the exercise code through seed collection, cleaning rules, duplicate, missing-value and outlier handling, quality reporting, validation, and scaling to larger crawls.

Turn meetup-place recommendations into a machine learning ranking problem, with labels based on the user funnel and evaluation metrics.

Handle messaging API character limits with error routing, preserved data, and file-based fallback delivery.

Build a two-agent n8n pipeline that selects AI news, uses RSS sources and Pinecone, and generates a concise briefing.

Build a RAG pipeline with Gemini embeddings and Pinecone to ground LLM responses in stored AI policy guidelines.

Schedule a recurring job with Schedule Trigger, collect data from RSS feeds, and practice scraping sites without feeds using HTTP Request and HTML extraction nodes.

Learn to extract and clean complex JSON data. This hands-on guide uses array operations and regular expressions to format dates and merge fragmented payloads into a single message.

Learn n8n event-driven architecture and deploy a local automation pipeline with Docker and PostgreSQL.

Compare open-source alternatives to commercial no-code and low-code automation platforms, including their costs and vendor lock-in limitations, and learn why we chose n8n for its architecture.