Google plans to charge developers $2.85/app and $3.65/game if a US user follows a link that takes them outside of the Play Store to install it within 24 hours (Sean Hollister/The Verge)
OpenAI rolls out an update to let users adjust ChatGPT's characteristics, including how warm or enthusiastic it is, and its use of headers, lists, and emojis (Emma Roth/The Verge)
Sources: AI chipmaker Cerebras plans to file for a US IPO as soon as next week, targeting a Q2 2026 listing, after withdrawing its prior IPO filing in October (Echo Wang/Reuters) https://www.reuters.com/business/ai-chip-firm-cerebras-set-file-us-ipo-after-delay-sources-say-2025-12-19/
New York Gov. Kathy Hochul signs the RAISE Act into law; the AI safety bill's text was modified earlier to more closely resemble California's SB 53 (Ashley Gold/Axios) https://www.axios.com/2025/12/19/new-york-ai-safety-bill-hochul
Source: Tencent has secured access to Nvidia's Blackwell chips through a cloud service operated by Tokyo-based Datasection (Jacob Adelman/Barron's Online) https://www.barrons.com/articles/china-tencent-nvidia-blackwell-chips-cloud-9d5e5998?st=8Mssvc&reflink=desktopwebshare_permalink
Meta board member Dina Powell McCormick, an ex-top adviser to Trump, steps down after joining in April; source: she is considering retaining an advisory role (Kurt Wagner/Bloomberg)
Sequoia's Shaun Maguire speculated in an X post that a Palestinian student was behind the Brown University shooting; the post was later deleted (Chris Stokel-Walker/Fast Company) https://www.fastcompany.com/91463942/sequoia-shaun-maguire-brown-university-shooter-palestine
The Delaware Supreme Court rules that Elon Musk's 2018 Tesla CEO pay package, worth around $56B when it vested, must be restored, ending a years-long fight (Lora Kolodny/CNBC)
2025 LLM Year in Review: shift toward RLVR, Claude Code emerged as the first convincing example of an LLM agent, Nano Banana was paradigm shifting, and more (Andrej Karpathy/karpathy)
AI robotics startup Physical Intelligence claims vision-language-action models learn to align human videos and robot data as pre-training is scaled up (Physical Intelligence)