🚨 The AI game just changed AGAIN—here’s the 90-second recap you can’t scroll past 👇 You know that moment when ChatGPT is mid-essay and you *finally* remember the detail that would make it perfect? No more Ctrl-C ➜ new prompt ➜ “ignore everything above…” OpenIA just dropped an UPDATE button that lets you **inject extra context while the model is still running**. Deep-research mode just became freestyle jazz. The main character this week is Kimi K2 THINKING. 256k context, 200–300 tool calls *in a row*, no human hand-holding. Benchmarks? - 44.9 % tool-use - 60.2 % browser tasks - 71.3 % SWE-Bench verified Same price tag as the old K2: $0.15 cache hit → $2.50 output. Turbo flavor exists if you hate money.
- 0 replies
- 0 recasts
- 0 reactions