AI is Rewriting the Rules of Governance, Science & Work

• Albania Appoints First AI Minister – Diella, the world’s first AI-powered virtual minister, takes charge of public procurement to ensure transparency and fight corruption.

• Gauss: AI for Mathematics – Math, Inc. debuts Gauss, an autoformalization agent that formalized the strong Prime Number Theorem in record time, redefining mathematical verification.

• Claude AI for Teams – Anthropic rolls out smart memory, allowing Claude to remember projects and context, streamlining collaboration and productivity for teams.

From politics to mathematics to the workplace, AI isn’t just assisting, it’s transforming how the world operates. The future is arriving faster than expected.

Albania Names Diella the First AI Minister to Fight Corruption

Albania has appointed the world’s first AI-powered virtual minister named Diella, meaning "sun" in Albanian, to oversee public procurement and combat corruption in government tenders. Announced by Prime Minister Edi Rama, Diella is a digital assistant who has been supporting citizens on the e-Albania platform since January 2025 by helping with bureaucratic tasks and issuing digital documents. Now as a cabinet member, though not physically present, Diella will manage and make all decisions regarding public tenders, aiming to ensure the process is fully transparent and free from corruption. This innovative move seeks to eliminate human bias and manipulation in awarding public contracts, strengthening Albania’s efforts to join the European Union by 2030. The appointment is seen as a historic step in governance by using AI to enhance transparency and integrity in public administration.

This AI Agent Transforming Mathematical Formalization

Math, Inc. has introduced Gauss, a groundbreaking autoformalization agent designed to assist expert mathematicians in formal verification. Gauss successfully completed a challenge set by Fields Medallist Terence Tao and Alex Kontorovich to formalize the strong Prime Number Theorem using the Lean proof assistant, a task that had faced major delays due to complex analysis difficulties. Working autonomously for hours, Gauss produced about 25,000 lines of Lean code encompassing over 1,000 theorems and definitions, achieving what typically takes years of human effort within just three weeks. The project was supported by the advanced Trinity environments infrastructure from Morph Labs, enabling thousands of concurrent Lean runtimes. While currently requiring expert guidance and natural language scaffolding, Gauss represents a significant leap towards scalable formalization, promising to drastically reduce the time needed for massive mathematical verification projects. Math, Inc. plans to deploy Gauss for beta testing with mathematicians and aims to increase formal code output by 2-3 orders of magnitude within a year, pushing towards verified superintelligence and machine polymaths. The initiative is supported by DARPA’s expMath program.

Anthropic Claude AI Launches Smart Memory for Teams

Anthropic has introduced a new memory feature in its Claude AI app designed for teams at work, enabling the AI to remember projects, preferences, and context to avoid repeated explanations and enhance productivity. This memory is fully optional and comes with granular controls allowing users to manage what Claude retains, including project-specific memories that keep details compartmentalized for complex, concurrent initiatives and sensitive information. Additionally, Incognito chat mode lets users interact with Claude without saving the conversation to memory, ideal for confidential or fresh discussions. The memory feature is rolling out to Team and Enterprise plan users, with enterprise admins able to control memory settings. This development aims to support seamless, context-aware workflows and make work more efficient by building on previous conversations and eliminating the need for constant context resetting.

Hand Picked Video

In this video, we explore an AI-powered job-hunting agent that automates job searches based on title, experience, and location. It fetches job listings from multiple websites, provides skill recommendations, and analyzes industry trends to help you prepare for applications.

Top AI Products from this week

• Mayson – Meet "Mayson", the low-code, AI-enabled BaaS that builds production-ready backends instantly. Auto-generate scalable APIs, deploy to your own cloud, and export clean Python code. No vendor lock-in, full control. Ready for your MVP and beyond.

• Stash MCP Server – Stash MCP Server lets AI agents like Cursor, Claude, and Copilot access your team’s real context (code, docs, issues) so they can resolve tickets without endless prompting. Just say "solve my assigned issue with the ID of …" - that’s it.

• Mozart AI v0.5 – The AI music platform to create, improvise, and finish music — built by artists for artists. Start with a loop or a prompt, improvise every layer, and finish your music in minutes. Every generation is context aware. Free during beta.

• LockedIn - LockedIn is a tool where you can paste any job URL and watch our AI research the role like a human would, then generate personalized cover letters and resume tips for you to review and use.

• Epoch – Craft intelligent AI agents with Epoch, your AI co-pilot. Powered by SmartBots.

• Genspark AI Browser – Excited to announce Genspark AI Browser — now available on Windows and Mac with world's first AI features that will revolutionize how you work online. From on-device AI models to autonomous web navigation – your browser now works for you.

This week in AI

• AI Podcast Boom - A new startup plans to produce 3,000 AI-focused podcast episodes weekly, exploring AI’s impact across industries and entertainment.

  Related

• Qwen3-Next AI - Qwen3-Next by Alibaba is an 80B parameter model activating only 3B at inference, cutting training cost by 90%, 10x faster inference, excels at long-context and complex reasoning tasks.

• AI Boosts Cancer Care - AI empowers patients and doctors by analyzing complex data to tailor cancer treatment, enabling faster, more precise, and personalized care that standard protocols often miss

• Agent 3 AI - Agent 3 is 10× more autonomous than predecessors, continuing software development tasks with minimal supervision—marking the "Full Self-Driving" moment for coding.

• Stable Audio 2.5 - Stable Audio 2.5 delivers fast, high-quality, customizable audio for enterprises, generating 3-minute tracks in under 2 seconds with advanced controls including audio inpainting.

Paper of The Day

TORSO: Template-Oriented Reasoning Towards General Tasks" introduces a novel method to enhance large language models' inherent reasoning abilities without relying on few-shot prompts or additional training. TORSO guides model decoding by injecting specific tokens to unlock internal reasoning and synthesize final answers, achieving superior performance across diverse benchmarks. This approach offers a more consistent and efficient reasoning process, improving both answer accuracy and rationale quality across a broad range of tasks.

To read the whole paper 👉️ here.