@Scobleizer
@OpenAI @NotebookLM And same in bullet points, thanks to @blevlabs's AI agent he and I built together. Here's what OpenAI announced today (March 5, 2026): • GPT-5.4 launched — new frontier model with 1 million token context window, "extreme" reasoning mode, and the ability to interrupt the model mid-response to redirect it https://t.co/5aGF9aIdcp • GPT-5.4 Pro and Thinking versions — Pro for maximum intelligence, Thinking for deep reasoning (rolling out in ChatGPT) https://t.co/lmdbFUoUuX • First AI to beat humans at operating a computer — 75% on OSWorld vs 72.4% human baseline https://t.co/kVWETi97c6 • First model to pass 50% on APEX-Agents — autonomous agent benchmark, step function improvement https://t.co/pkHew7MFop • /fast mode — 1.5x faster responses with same intelligence and priority processing https://t.co/CzNIQu0ee8 • Codex overhaul — now on Windows, /fast mode, multi-agent support. Sam says "fix 3 things" to make it perfect https://t.co/dX5gQH135n • Computer-use with Playwright — production-ready, not a preview. Navigate web, fill forms, click buttons https://t.co/muyfVMCrVt • ChatGPT directly in Excel and Google Sheets — new financial services tools suite https://t.co/b394EVtUPU • Aggressive pricing — $2.50 input / $15 output per million tokens (standard). Roughly half the cost of Claude Opus https://t.co/evXiWXwGhr • 33% fewer false claims and 18% fewer errors per response vs GPT-5.2 https://t.co/waZENFVhMp • Medium reasoning = GPT-5.2 extra-high at 83% less cost — same quality, dramatically cheaper and faster https://t.co/l2HpISwMgi • Key benchmarks: SWE-Bench Pro 57.7%, BrowseComp 82.7%, ARC-AGI-2 Pro 83.3%, Vibe Code Bench 67.4% (#1 SOTA), FrontierMath Tier 4 38% (up from 2% a year ago), GPQA Diamond 94.4% • Same-day third-party integrations: Cursor, Windsurf, Perplexity, Notion, Bolt, Roam Research, https://t.co/KEaCPJtVL0