160B Tokens so far - 5 threads in Sol Ultra Fast Mode to use up before 6PM PST. Creating Autocorrect from scratch.

preview.redd.it/jtbxlhj0lkmh1.png preview.redd.it/xue6hif5lkmh1.png This is just 1 of 2 Codex Accounts. Around 160B tokens so far on just Codex. And another 300B on Fable. These resets really helped. 14% fast mode. I have 5 threads coordinating with each other, with Fable babysitting my context language model ranker (it helps de jumble tangled words by looking into n-gram tokens (words prior)). Building a iOS keyboard created from scratch like grammarly, except I have toolbar buttons that have built in prompts like translation, re-write (for work, dating, apologies), interpreting ambiguous text to help you respond. Not only can you customize your own toolbar buttons (which sits on the suggestion bar), you can also customize each key . For example, holding spacebar would take the text in the textfield and ask a question to LLM; holding T, would translate the text to the preferred language; eventually I will open up to the community to design their own keys, like G sends a google meets, it could be anything fully customizable. The only third-party keyboard license available is keyboardkit is not good for autocorrect; so once done I'm probably going to wrap it into an SDK and also offer the autocorrect technology via license deals to Whisppr and other big companies using the keyboard as a productivity tool. All other third party keyboard SDKs were bought up by google. The current version on the App Store uses keyboardkit, which Im not a fan of so rebuilding from scratch and hoping to put the final product up soon. My flag ship feature is de-centralized end-to-end encryption through the keyboard. This will be a surprise.