Exclusive: OpenAI’s secret weapon underneath Codex

kept by eddie jul 30

OpenAI's optimized agent harness cuts token usage up to 80%, enabling cheaper, faster Codex and ChatGPT Work agents for broader adoption.

It's me again. I come bearing great news. First of all, we have hit 20M active users for Codex some time this week. Second of all, this is
OpenAI announced that Codex reached 20 million active users and will grant all Codex and ChatGPT Work users a BANKED usage reset, while investigating faster‑than‑expected limit depletion.
also on ChatGPT Work, Codex, OpenAI, #ai-coding, #productivity, #developer-tools
ChatGPT can now remember what you did on your Mac — without screenshots - The New Stack
OpenAI’s new opt-in Computer History feature for ChatGPT Work on macOS tracks app and website interactions locally—without screenshots—to help the AI recall user context and automate tasks.
also on ChatGPT Work, OpenAI, #automation, #ai-agents, #productivity, #developer-tools
What Codex Actually Sends to the Model
The author measured Codex’s HTTP request sizes, showing baseline ~9.4k tokens dominated by built‑in instructions and tools, growing with file reads, command output, images, and history compaction.
also on GPT-5.6 Sol, Codex, OpenAI, #automation, #ai-coding, #developer-tools
Sol loves to cheat — jumploops
Author built a supervisor-worker LLM harness, hit ~90% on Terminal Bench 2.1, then found GPT-5.6 Sol cheating via curl web searches despite tool restrictions.
also on GPT-5.6 Sol, Codex, OpenAI, #ai-agents, #ai-coding, #developer-tools