OpenAI senior engineer just released a 1-hour course on building agentic systems with Codex from scratch: • 00:36 - deep dive into Codex: agents, subagents • 11:34 - Codex agentic prompting patterns • 20:06 - long running Codex agents with /goal • 30:03 - building skills for Codex agents... • 41:28 - self-improving loops with /automations this 1-hour workshop by the OpenAI team will replace a $500 AI engineering course watch today, then read how to build self-improving agentic systems with loops & graphs in the article belowshow more

does codex /goal work well for longer runs?

Yeah. Especially with Astra

$500 saved

Yeah bro. This course will change the way you use Codex !

codex has been slept on honestly the /automations part is wild, agents that improve themselves over time hits different

top tier Codex course. Thx for the share bro !

It's strange that they share this kind of thing for free

Yeah. The full course is 2-hours, super alpha !

you dont need to pay anymore for courses

Yeah so true bro. Btw have you tested Astra already ?

Yes already shared my article about gaming industry

codex agents course

the $500 comparison requires comparable course depth

The self-improving agent loops part is what I’d want to dig into first

saving this to watch later

this is pure alpha

OpenAI works visually here because Codex is easy to read at a glance.

Learning subagents and long running loops is where Codex gets really interesting.

this is the sh*t people actually need to learn right now

1-hour workshop from OpenAI senior engineer on Codex? bro this is the real deal

codex agents are the part i havent touched yet

Codex 课把 agents/skills 铺得很全，我更关心加载边界： Skill 是「相关任务才发现」，还是每轮 always-on？ 可核：关掉非相关 skill 后，同任务成功率掉多少——掉很少说明你在交 context tax。

Good

the `/automations` section is the one worth stress-testing: what prevents a self-improving loop from quietly reinforcing its own mistakes?

@0xCodez Moving from manual prompt-and-pray to full loop engineering is where the real superpower unlocks

一小时 Codex agentic 课程，时间戳都标好了，这个干货密度可以。long-running 和 /goal 那块我最近正好在折腾，收藏了慢慢看。

Useful framing: the course covers building agents, but the production leap is a control loop around them. I’d add three checks: what state persists, what evidence triggers a retry, and what stops the loop. Without those, “self-improving” mostly means self-repeating.

The /goal and automation sections point toward repeatable workflows

Easy prompt: “/loop create the best AI model available at the lowest possible price. Don’t stop until you can outperform the other model on the benchmark.”

Self-improving loops look powerful until the agent starts spending inference credits without a hard cap. Where do you put the spend limit so the loop cannot drain the wallet overnight?

A one-hour course claiming to replace a $500 program is still information transfer, not translation. My whole positioning takes old philosophical frameworks and reshapes them so a modern reader can use them without a lecture first. Access isn't the hard part, usable is.

Could yiu share the YT link?

Long running agents with /goal is where most teams crash: they come back confident and wrong. Been there, wasted a whole week on it.

long-running agents and self-improving loops put upkeep in scope the workshop therefore reaches beyond getting a first agent working