2x £20 Plans Usage Findings

I’ve been pretty put off in the last few weeks, honestly both for Codex and Claude Code. But now I decided to try for my self. I’m using both to work on a single project switching between the two seeing the advantages of both models. I’ve been using only GPT-6.1 Sol Medium/High and Opus 5.5 High and Sonnet 5.5 High very rarely for medium to basic changes and honestly, with a well defined structure and well planned architecture docs (security, app foundation, design, etc) along with a custom handoff script I made which overlays no more than 50 lines of clearly defined work done and immediate next steps, run after concluding each task/phase, I’ve been loving it. A well orchestrated plan along with a dedicated orchestration chat, testing your app foundation at the start making sure everything works and going about it in phases, phase-1/app foundation, etc and using /clear and fresh chats goes a long way to save usage, get your work done without annoying hallucinations or AI amnesia and the new models, especially GPT 6.1 has been so very usage/token efficient, and Opus 5.5 is excellent at designing the plan and the architecture files and even implementing with Sonnet 5.5 Medium/Hugh is great but as long as the scope is well defined even a GPT 6 Luna High is a workhorse at implementing. If you expect to write one prompt and then wait until your usage depletes, it’s kind of on you because then you’re wasting more time telling it what to fix/change because no proper planning and testing was done, all saved to their respective MD files which each new /clear or fresh chat can read from is a game changer, they’re not version histories just the plan and the wording gets updated once something has been implemented and then the handoff skill I created puts what’s been done and the next session can orient itself with the changes made by either LLM and just continue from there. No hallucinations, no AI amnesia. I have £20 plans for both and I can see my usage lasting for the whole week at this rate + I got 4 banked resets saved on Codex with 93% remaining. Most of the usage complaints on here are from one-shot prompts with 0 planning, and then they spend their whole usage fixing stuff which should have been defined in AGENTS.md, architecture docs, config files (for max sub agents and which sub agents to use and a clearly defined scope of when to use them). You can say this is overkill but it’s just good practice especially when trying to stretch your usage. I was so anxious to touch any of these new models like Opus 5.5 or 6.1 Sol because of all the fear-mongering even from the 20/10x users with no 5h limits. From now on I’m gonna just test for myself before getting put off by usage posts in both communities. People getting by just fine who use well defined structures before they even type the first prompt just get by fine and don’t see a reason to post, so I thought I’d report my usage findings on the lowest tier plan. I got a reset yesterday along with 4 banked resets saved on Codex too 🙏