GPT-6 Sol and Luna reset the cost of coding agents
What changed from GPT-5.6 Sol and Luna to GPT-6: token prices, prompt caching, factuality, coding benchmarks, and the honest comparison with Claude Opus 5.5.
SOL$2 · $10
LUNA$0.10 · $0.50Jev Doesn’t Write Code. That May Be the Point.
Learn how TypeSafe AI's Jev makes typed decisions for coding agents, where it fits beside Codex, and which safety boundaries still belong in code.
AGENTS.md for Codex: Rules That Survive the Monorepo
Learn how Codex loads AGENTS.md, how to split root and nested instructions, what to include, and how to verify the active instruction chain.
Codex Context Window: Compact Without Losing the Task
Learn how the Codex context window, automatic compaction, and /compact work—and how to preserve decisions before you compact, fork, or start a new chat.
Codex Skills: How to Find the Ones Worth Installing
Learn how to choose Codex Skills by task fit, trigger quality, source, permissions, maintenance, and a real before-and-after test—not popularity alone.
Codex Subagents: Parallel Work Without Token Waste
Learn when Codex subagents make work faster, how to route bounded tasks to cheaper worker models, and how to prevent duplicate work and token waste.
Codex Plan Mode: from a good idea to an implementation-ready plan
Use Codex Plan Mode to inspect a repository, resolve key decisions, review an implementation-ready plan, and hand it into execution.
GPT-6 Astra in Codex: conquering the AGI era
What GPT-6 Astra changes for coding in Codex, where its strongest benchmarks appear, and how to use its long-context capabilities responsibly.
Codex MCP: set it up, verify it, and debug it
Connect MCP servers to Codex, verify configuration, authentication, tool discovery, and execution, then debug failures layer by layer.
Why Codex uses so many tokens in large repositories
Learn why Codex sessions grow in large repositories and how scoped context, concise tool output, and agent-readable project guidance help.