Support colleague: "Customer isn't getting certain emails, can you check?"

Pasted her message into Codex CLI, right in the email service source directory. 5 minutes later it explained why
the system behaves exactly as designed — no bug.

Then asked it to draft a support reply a non-technical person can understand.

AI coding agents aren't just for writing code. They're incredibly fast at understanding existing codebases and
translating technical behavior into human language.

@tompson Can you tell us about the actual root cause in this case, except „works as designed“? How did Codex solve it?
@saxx the customer thought he and his colleagues should get a mail in that kind of event. But it was designed as only the administrator should get an email in that kind of event. Codex figured that out by the subject of the mails and looking at the logic in the code for sending out those mails.
@tompson Cool 😊 I know you‘re quite hyped about these possibilities for a while now: Out of curiosity, how much would you be willing to pay for the current level of functionality/usage allowance in case OpenAI will increase prices to cover costs?
@saxx I'm not sure if it will get that more expensive. Currently we are using multiple tools in parallel and I hope we can consolidate later but we are already spending a lot per month
@tompson Since even OpenAI are admitting that they loose money which each subscription, I think it’s safe to assume it will need to get much more expensive to cover even their OpEx. Not sure how they’ll ever get ROI on their CapEx.
@saxx I think I don't really get the point here. The logic would then be to use them now when they are cheap and decide later, when they are more expensive, if you still want to use them.
@tompson Let me rephrase 😇How much is the (very cool, no question about it) functionality of Codex or Claude Code or ChatGPT worth to you? 100€/M? 1000€/M?
@tompson I‘m struggling a lot with this question: How much should I rebuild my company to depend on these tools? How much am I risking if I downsize (eg. support personnel in your example), become dependent on them, and then they suddenly substantially increase prices to repay investors? Moving away from Azure or AWS to another to another hoster is painful, but possible. But no other provider can replace OpenAI and Anthropic in the foreseeable future?