@Oblivious9021 Use an LLM agnostic router that routes by task against LLM/model capability benchmarks with mid-task escalation and de-escalation routing with context injection and host locals on prem for routine tasks.
@dklineii I have a very simple methodology. Progression or regression towards desired state? Knowing or doing issue when desired state is not achieved? That’s all. It’s either a training issue or an ability issue. Accountability always.
@mcuban I am building the framework for an airgapped sovereign local, specifically for regression scenarios where a generation may have to learn old skills, so yes there will be. But for nontechnical people to use AI there need to be governance, that’s why TaG exists. Democratization.
@cryptoPaulRudd@AmericaReal3@grok Wrong. This was an online only commercial in the early days of the internet. It been around for over 20 years and is one of the first examples of viral”social media” marketing years before social networks existed.
@E_MindsetHQ 45 minutes late every appointment? Someone is not coding out every time. If it has happened more than 3 times your assessment is correct, but because his team does not understand time Mangement or the shop he works for is greedy and overbooks.
@chamath I have a shared memory across claudecode, codex, and openclaw that syncs every thirty minutes and creates an engram of where where all the projects stand and next steps with references to the full memory of each project and a link to the stored text logs of every session.