I'm on the 20 USD tier, and it works quite well for me. Basically I send one very carefully crafted task to the LLM per 4h limit, do a couple of minor questions, and the rest of the time I'm thinking/exploring/coding. I am producing around a tenth of the code of my colleagues, but around the same number of features.