If the trend holds, you will be able to run a model with Fable 5 capabilities on your MacBook around May 2027. Open source matched Opus 4.5 in nine months.
The model is rarely the bottleneck on a long project, the missing context is. I use Karpathy's "LLM wiki" system so the model maintains its own context (and mine). Here's how it works.
Most people do their AI work in chat, where it stays on the provider's server. Keep it in your own files instead and you can switch models anytime - and your work keeps getting more useful as they improve.