News · AI summarised to understand what matters
goppo — News
A selection of AI news worth paying attention to, summarised to understand the essentials: what happened, why it matters, what impact it may have and where to read the original source.
150 results
Gemma 4 12B is an open-weight Google model designed to run locally on machines with 16GB of memory or VRAM. Its mix of multimodality, Apache 2.0 licensing and local execution makes it relevant for teams trying to reduce cost and data exposure.
Meta introduced Business Agent to help companies respond to customers, handle requests and create personalized experiences inside its messaging channels. The impact may be strongest for small businesses already selling through WhatsApp and Instagram.
OpenAI introduced new Codex capabilities for building internal sites, dashboards and work materials from natural language instructions. The signal is clear: software agents are moving out of the code editor and into operational work.
A Nature paper presents Co-Scientist, a Gemini-based multi-agent system for generating and refining scientific hypotheses. The first validations focus on biomedicine, with in vitro results that still require preclinical and clinical assessment.
Cursor launched Composer 2.5 on the open-weight Kimi K2.5 base, with 85% of compute in Cursor's own RL. Third on the Coding Agent Index at $0.07 per task — 10 to 60× cheaper than rivals. Proof that specialised fine-tuning reaches the coding frontier.
At Google I/O 2026, Google shipped Gemini 3.5 Flash — which outperforms the previous Pro model on agentic evals at $1.50/M — plus multimodal Omni and an API that hands you remote Linux per call. Agent infra is no longer a differentiator.
V4-Pro is now permanently priced at $0.87 per million output tokens. DeepSeek turned a temporary discount into a strategic commitment that puts direct pressure on Anthropic and OpenAI to justify their price gap.