Daily AI Roundups
Today's roundup combines Monday's AI news with the Sunday roundup we skipped because the digest was too thin, so this 29 Jun 2026 edition carries both Sunday 28 Jun and Monday 29 Jun signals in one list. It leads with Elon Musk saying Grok 4.5 is in private beta at SpaceX and Tesla, based on a 1.5T V9 model with Cursor data, then moves to Sakana AI fully adopting Gemini Enterprise Agent Platform for Sakana Fugu. It also covers Dewank Pant, Shruti Lohani and Avijit Kumar's arXiv argument that shared-embedding LLMs cannot perfectly separate instructions from data, GS Group's 2.4 GW AI data-center campus plan in Gangwon Province, FT's report that AI data-center demand pushed U.S. power-sector M&A to roughly $204bn in the first five months of 2026, SemiAnalysis flagging the LeptonAI CEO exit after NVIDIA's $700M acquisition, AA-Briefcase results where GLM-5.2 leads open models on price/performance, Momenta seeking a $752 million Hong Kong IPO, and DeepSeek open-sourcing DeepSpec for faster LLM inference with draft models.
- frontier models
- AI agents
- prompt injection
- data centers
- AI infrastructure
- autonomous driving
- inference optimization
- AI benchmarks