Tibo's Day 12 | Reuse subagent context for more effective usage

The Day 12 announcement describes three efficiency improvements in Codex and Work: reuse context across subagents, retain caches longer, and improve routing to produce more cache hits.

The benefit comes from less repeated processing

Subagents working on the same project can repeatedly receive shared instructions and background. Reusing that material reduces repeated processing and leaves more usage for new task steps.

Multi-agent work with substantial common background is therefore especially relevant. Tasks with little repetition or frequently changing context may benefit differently. The announcement gives no uniform percentage increase for every account or new weekly allowance figure.

Why some users may still report fast consumption

A relevant user reply says weekly usage can still run out quickly. This is one account’s experience. Without complete model, task and window conditions, the reply cannot estimate average improvement.

For a personal comparison, keep similar tasks, the model, subagent use and usage-window records. Comparing only two remaining percentages after tasks finish loses useful context.

Which later update adds detail?

Day 16’s caching article describes a 12-hour API option and the claim about restoring old conversations in signed-in products. It adds reuse conditions, without establishing one usage multiplier for everyone.

In the same thread, Tibo’s additional reply mentions GPT-6 Pro speed, tools and parallel-agent changes. That has a separate product scope from this article’s Codex context reuse.