LWiAI Podcast #242 - ChatGPT Images 2.0, Qwen 3.6 Max, Kimi-K2.6
Summary
Hosts Andrei Kerenkov and Jeremy Harris discuss OpenAI’s text-capable image model, Alibaba’s Qwen 3.6, and Google’s Deep Research agents. They analyze Mozilla using Claude for bug fixes, SpaceX’s potential $60B Cursor acquisition, and Cerebras’ IPO. The episode also covers Anthropic’s $5B Amazon deal and new funding for alternative AI paradigms.
Topics discussed
Intro, episode format, and workout app discussion
Listener feedback on nuclear treaties and AI safety
OpenAI's new image generation and SVG capabilities
Qwen 3.6 Max and the open-source model race
Google Gemini 3.1 Pro and Deep Search features
AI in cybersecurity and software vulnerability detection
Starbucks ChatGPT integration for ordering
Elon Musk's $60B offer to acquire Cursor
Cerebras Systems files for IPO with $23B valuation
New AI startups: Superintelligence and alternative training
Amazon's $5B investment in Anthropic
Talent shifts: OpenAI, Meta, and XAI departures
Mandatory AI training via mouse and keystroke tracking
Chinese chip fabs and mature node competition
Google TPUs, Xanadu quantum computing, and inference chips
Chinese models: Minimax, Kimi, and MLA attention
Hugging Face ML Intern and adversarial training attacks
Mythos security breach and NSA access controversy
Parquet stable looped transformer architecture
OccuBench for evaluating AI in economic sectors
AI music saturation, deepfakes, and episode wrap-up
Listen ad-free on Castria