Shine
- 4 Followers
- 67 Following
- 190 Posts
In general, I like the concept #OpenAI #Codex introduced: task -> variant -> feedback loop -> pull request. Feels almost like talking to a human through PR review, but there are quite a few things that I don't like:
- Lack of Docker containers inside the Codex environment.
- #Codex UI always shows the whole thing, without an option to see the diff of the latest changes in the branch.
- Integration with GitHub pull requests is apparently not as great as it could be...
- Lack of Docker containers inside the Codex environment.
- #Codex UI always shows the whole thing, without an option to see the diff of the latest changes in the branch.
- Integration with GitHub pull requests is apparently not as great as it could be...
Trying to work with #AI as if it's a human engineer in a micro project. Quite an interesting challenge: on one hand, the amount of knowledge poured into LLM is stunning, but there's so many issues with âlocalâ contextâŚ
Boring reminder: If an alert isnât actionable, it shouldnât exist. Alerts are for action, not noise.
The beauty of microservices and containerization: Language barriers have faded. Develop in Python, deploy in Go, connect via gRPC or REST - mix and match as needed. No more rewriting everything in Java just because the stack started that way!
Captain Obvious strikes again: when setting up blackbox_exporter to monitor your endpoints, host it on a separate server/cloud. Not on the same cluster you're going to monitor. A downed blackbox_exporter doesnât export much đ
#Terraform lets you keep modules in the same repo as your code, which is great for getting started. But as your deployment logic grows, consider moving modules to a separate repo.
This brings major benefits: better version control and easier QA/testing for module changes.
When planning a new service that handles customer data, and if maintenance procedures are required, always ask: how can this operation be safeguarded? How will we observe progress? How will we handle failures?
Design maintenance operations to be safe by default.