if you continue despite 5 failed businesses over 7 years, you deserve all your successes simply because you have more resilience than the vast majority of people
on Codex? what workload are you doing, because I've got multiple agents running a few days a week at least, and I've never hit weekly limits
might hit it for the first time this week though because I've set everything to xhigh. but tbf I've got a few resets left so I'm not worried
very rare moment right now where both Codex and ChatGPT web ui are down
I've been using ChatGPT almost daily since 2023 and this is the first time I've seen both go down concurrently. the influx of new users must be absolutely massive, so definitely a good problem to have from their perspective
if you can spare about 10-15 hours a week, you can definitely pick up the fundamentals of SWE over the course of say 6-12 months. may turn out to be a great investment over the long term
it is a bit difficult at first, but if you are able to find the time, I'd recommend FreeCodeCamp and then building beginner projects (AI is great at suggesting these) and using AI + YouTube to learn different tools / frameworks
one thing that Anthropic doesn't get enough credit for is their UI / UX - it just feels far nicer than the interface of any of their competitors
they also avoid cramming numerous references in the chat response which makes the reading experience more pleasant, which helps especially over long chat sessions
Gemini on the other hand (using 3.6 thinking here) seems to restrict the number of tokens that can be used for any given response, unlike ChatGPT, which means any response which requires depth will by its nature be lossy
I think even if someone were to give their playbook away, that knowledge can only take one so far
information eventually spreads in any industry, especially nowadays with the accelerating capabilities of AI, and the moat eventually becomes (or will become_ the ability to execute, and do so for a sustained period of time, while being known as a more reliable company than one's competitors
I can see more and more labs trading extra usage / additional credits for rights to train on user data
honestly surprised that more labs haven't done it as of now
@Tech_girl as long as its pretty cheap, better than Opus 4.8, and available via a subscription I think most people will be fine with it
most labs now seem to be going with a frontier model + cheaper, smaller model that they distill from their frontier model
although OpenAI doesn't open weight their models, there's no denying that they have tremendously increased the supply of high quality intelligence by releasing even their top tier models for a relatively affordable subscription with much higher rate limits than any of their competitors
@ItsKieranDrew I plotted your follower count and income growth and it validates the idea that follower count is a strong leading indicator for income growth