OpenAI published Codex 0.169.0-alpha.3 on November 7, 2026, marked pre-release with no release notes. The releases page lists 0.167.0 as the latest stable version.
How we checked
Read the release page and releases list on GitHub; confirmed tag, date, pre-release label and missing notes.
Why it matters
With no notes, this alpha gives builders nothing specific to act on, and anyone wanting a stable Codex should not pick it up by accident.
Try this
If you pin Codex in scripts or CI, keep the pin on stable 0.167.0 rather than this alpha.
Grok Bot routines can run on schedules or on events: a Slack message, GitHub, Linear, Sentry or PagerDuty events, an email to the Bot's own address, or a webhook call.
How we checked · Why it matters · Try this
How we checked
Video's schedule and email trigger examples checked against xAI's routines docs. No hands-on test.
Why it matters
Agent work can start from engineering alerts or messages, not only timers, which widens what can run unattended.
Try this
Start with a Slack trigger on explicit mentions rather than any message, so a busy channel cannot set off constant runs.
A Ray Fu video says each Grok Bot gets its own virtual computer. xAI's docs say otherwise: all Bots on an account share one cloud computer, with separate screens that are not security boundaries.
How we checked · Why it matters · Try this
How we checked
Read xAI docs on computer and bots; they contradict the video's per-bot computer claim. No hands-on test.
Why it matters
Files and sign-ins are not isolated per Bot, so anything one Bot is signed into or saves is within reach of the others.
Try this
Before signing a Grok Bot into any service, assume every Bot on your account can use that session and read its files.
Grok Bot's Teach a task feature, where available, lets you perform a workflow once in a one-to-one chat. It records visible computer interaction for up to ten minutes and produces a draft skill to review.
How we checked · Why it matters · Try this
How we checked
Video notes matched against xAI's skills and routines docs. No hands-on test.
Why it matters
You can capture a repeatable browser workflow without writing it out, though the result is a draft that still needs checking.
Try this
Teach one short, read-only task, then read the draft skill's steps and approval requirements before putting it on a schedule.
Several Grok Bots can share a group chat and pass work among themselves, but xAI's docs say every Bot reply counts toward usage and no switch stops Bots messaging each other.
How we checked · Why it matters · Try this
How we checked
Video's group chat and handoff claim checked against xAI's chat and collaboration docs. No hands-on test.
Why it matters
A multi-agent team can spend usage on its own chatter, and approval requests do not appear in the group, so they are easy to miss.
Try this
Give each stage one owner in the kickoff message and use as few Bots as the job needs.