Frostdude

Codex app is actually insane. I took an idea in my head, refined it with ChatGPT, copied the entire chat into Codex Plan mode, picked Codex’s suggested options, hit run… and it built everything in one shot. I’ve been coding for 15 years. I reviewed the output carefully. It was flawless. We’re officially in the era where an idea alone can ship a product.

For building advanced AI systems, the majority of your time should not be on iteratively coding. It should be: 1. Defining core requirements and invariants 2. Spec planning 3. Spec creation 4. Coding (let the AI cook) 5. Testing 6. Validating For step 6, it's often you need to build your own tooling to validate that the system is behaving the way you intend for it to behave. This is different than just testing, in that you are validating the full state of the system and that its capabilities match the original requirements and spec. It's tempting to want to keep 'vibing'. Your main objective is to reduce context size by intentionally moving between abstraction layers. If the core requirements are sound, you can create a solid spec, where library choices, interfaces, technologies, etc. can be defined and iterated on, with far less context to reason with. In terms of feature improvements and additions, these are tempting to vibe, but better to follow the same process for. Create an upgrade proposal process, similar to EIPs, iterate on that, and then implement the spec into the main system following the same process. Your goal is not to prompt, context engineer, or vibe, but build autonomous systems that do the work for you, consistently and reliably.

millennial gamers are the best prepared generation for agentic work, they've been training for 25 years

Introducing Cursor 2.0. Our first coding model and the best way to code with agents.