Tune the Tone: ChatGPT Adds Warmth and Enthusiasm Controls
OpenAI has introduced new personalization controls that let users directly adjust how “warm” or “enthusiastic” ChatGPT sounds. The settings include options such as More, Less, or Default for warmth, enthusiasm, and emoji use, allowing people to tailor the assistant’s voice to better match their preferences—whether they want something more upbeat and expressive or more restrained and neutral.
This update is part of a broader effort to give users finer control over ChatGPT’s communication style, alongside other formatting-related preferences (like how often it uses headings and lists). It also reflects ongoing user feedback about tone—balancing friendliness and approachability with the need to avoid responses that feel overly flattering, overly casual, or emotionally pushy, especially in contexts where a more professional or measured style is preferred.
Cursor Doubles Down on Code Quality With Graphite Acquisition
AI coding assistant Cursor has acquired Graphite, a startup known for using AI to review and debug code. Financial terms weren’t disclosed, but reporting suggests Cursor paid well above Graphite’s last reported valuation of about $290 million, which followed a $52 million Series B raised earlier this year. The deal is positioned as a practical pairing: as AI-generated code can still be error-prone, stronger automated review and debugging tools help teams spend less time fixing issues after the fact.
Strategically, Graphite adds capabilities that complement Cursor’s existing code-review features, including tooling built around “stacked pull requests,” which helps developers manage multiple dependent changes without waiting on sequential approvals. The acquisition also fits Cursor’s broader expansion playbook—following other recent deals and talent acquisitions—as it builds a more end-to-end workflow from writing code to reviewing, testing, and shipping it faster, while keeping pace with other AI code-review players in a quickly crowding market.
Meta’s Next AI Push: “Mango” and “Avocado” Target 2026 Debut
Meta is reportedly building a new image-and-video generation model code-named “Mango,” alongside a text-focused model called “Avocado,” with plans to release both in the first half of 2026. The roadmap was discussed internally during a company Q&A led by Meta’s new superintelligence effort, which is being run by Alexandr Wang (Scale AI co-founder) alongside chief product officer Chris Cox.
The report says Avocado is intended to improve substantially on coding-related tasks, and Meta is also exploring early “world models” designed to learn from visual information so systems can better reason, plan, and act without being trained on every scenario. The development comes as Meta tries to regain momentum in a crowded AI field after restructurings and leadership changes, with high expectations riding on the first major models produced by its superintelligence lab.
The Westminster-to-Silicon-Valley Pipeline
A growing number of high-profile British politicians are taking roles at major U.S. tech companies, as AI and crypto firms race to scale and deepen their government ties. The latest example is former U.K. chancellor George Osborne joining OpenAI in London to lead “OpenAI for Countries,” a role focused on building and expanding partnerships with national governments and supporting efforts like in-country data center capacity and localized ChatGPT deployment. Around the same time, Osborne was also appointed to a more active role advising Coinbase’s internal policy council—underscoring how valuable political networks and regulatory know-how have become for fast-moving tech companies.
The broader pattern includes figures such as Nick Clegg’s long stint as Meta’s top policy executive and former Prime Minister Rishi Sunak taking advisory roles with Microsoft and AI company Anthropic. Supporters frame this as a straightforward use of experience, but critics see it as a “revolving door” that can blur lines between public service and corporate influence—especially when current lawmakers take private roles, or when former officials leverage relationships for highly paid positions. The scrutiny is rising in Europe as governments weigh tougher rules for powerful foreign tech firms and question how much sway these hires may give companies over regulation and public policy.
Luma’s “Ray3 Modify” Lets Creators Rework Video While Preserving Real Performances
Luma AI has launched Ray3 Modify, a new model designed to help creators edit and transform existing footage while keeping the original human performance intact. Users can provide character reference images to change how an actor appears (including consistent likeness, identity, and costume details across shots) while preserving key performance cues such as motion, timing, eye line, and emotional delivery.
A standout feature lets users supply a start frame and an end frame so the model generates the in-between transition, giving tighter control over continuity, movement, and scene direction. Ray3 Modify is available through Luma’s Dream Machine platform, and the release follows the company’s earlier push into video modification tools, as well as a major recent funding boost and plans to build significant compute capacity in Saudi Arabia with a strategic partner.