ai coding agents
Blog
Deep Dives for Engineers
Long-form AI practices, real-world case studies, and prompt strategies you can put into practice.
-
-
ai software developmentAI Software Development: What AI Does at Each SDLC Stage
AI software development means agents draft code and tests while engineers decide and check. Each SDLC stage, the evidence, and data from our own PRs.
- ai coding agents
How to Resume a Codex Session and Find Its History
Where Codex saves sessions on disk, how to resume one with codex resume or /resume, how to fork or archive it, and why a session goes missing.
- ai coding agents
Claude Code Session History and How to Resume It
Where Claude Code saves session history, how to resume a session, why one goes missing, and how to keep it past the 30-day cleanup.
- cursor skills
Cursor Skills and How to Add, Write and Share Them
Cursor skills are SKILL.md folders the agent loads by itself or you run with a slash. Where they live, how to write and share them, and what gets used.
-
engineering metricsEngineering KPIs: 12 to Track in 2026, With Formulas
12 engineering KPIs with formulas: DORA, PR cycle time and size, developer experience and AI cost, which ones to skip, and what 866 agent-written PRs show.
-
ai coding agentsHow Much Code Is Written by AI, and How to Measure It
Google says 75% of its new code is AI-generated; developers estimate 47%. What each figure counts, how to measure your own team's share, and our data.
-
AI coding toolsClaude Code Alternatives: 6 Coding Agents Compared (2026)
Claude Code alternatives for teams: Codex, Cursor, GitHub Copilot, Antigravity, Devin Desktop and OpenCode, compared by price, models and why to switch.
-
ai coding agentsClaude Code Best Practices for Teams, From Our Own PRs
Claude Code best practices for a team, from our own merged pull requests: smaller PRs, a check to run, short CLAUDE.md, hooks, and what to measure.
-
ai coding agentsClaude Code Monitoring: OpenTelemetry Setup and Queries
Set up Claude Code monitoring with OpenTelemetry: a tested collector config, managed settings rollout, metric names, PromQL queries and two traps to avoid.