@karpathy because the AI gets immediate feedback from the test output - so it becomes this nice loop of write code/tests -> run tests -> fix code/tests -> repeat. It's like having a built-in QA process that doesn't require you to context-switch constantly to check if things actually work.
@karpathy Also, something that's been a game-changer for me: asking the AI after it completes a feature to run and fix static analysis tools and tests. The beautiful thing about this is you don't have to manually test everything yourself each iteration
We just dropped something HUGE! TrackMage's new MCP Server lets you track shipments through AI chat.
https://t.co/wXNlwYx1GT
Just ask your AI: "Show me late shipments" or "Track package LZ123456" and get answers across 1,290+ carriers.
#AI#Logistics#ShipmentTracking#MCP
@vvsevolodovich на нем столько библиотек для нейронок и математики, что проще запустить демон на питоне и дергать его api, чем в php пытаться сделать что то подобное.