Government AI leaked on GitHub

Dear Readers,

A country is building its own super AI. And no one would have noticed—if it hadn't been for a GitHub leak.

What initially looks like a digital accident turns out to be a significant political signal: with the AI.gov project, the US government wants nothing less than to usher in a new era of administrative digitization.

Uniform access to AI providers such as OpenAI, Google, and Anthropic, a cross-platform monitoring dashboard called “CONSOLE,” and a government chatbot – all under the control of a federal authority. Launch date: July 4. It couldn't be more symbolic.

This development is a turning point in several respects. First, it shows that artificial intelligence is no longer seen as a purely economic tool, but as a political infrastructure project. Second, it brings central questions of transparency, control, and ethics to the fore: Who writes the answers for the government bot? And who controls whether AI-based administration makes decisions – or just simulates them?

In Today’s Issue:

• A surprise leak reveals America's ambitious plan to centralize its government AI

• Powerful Qwen3 models are now optimized and ready for your Apple devices

• Why choose one answer when Codex can now give you multiple options at once?

• Jan-nano: the small new AI agent that's acing research benchmarks.

• And more AI goodness…

All the best,

Government AI leaked on GitHub

A few days ago, a surprising leak shook the political world: planned AI access by the US government was accidentally revealed in a public GitHub repository—shortly before it was taken offline again.

At the center of it all is AI.gov, a project of the US federal agency GSA led by Thomas Shedd (formerly of Tesla/Musk Alliance). It is scheduled to launch on July 4 and comprises three core elements: a government chatbot, a centralized API to AI providers such as OpenAI, Google, Anthropic, AWS Bedrock, and Meta LLaMA, and a real-time monitoring dashboard called “CONSOLE” that monitors the use of AI tools within government agencies.

This is relevant for the AI community because it shows how large states want to radically digitize their administration and centrally control AI functions – a direct insight into real-world AI strategy. At the same time, it raises questions about transparency, data protection, and ethical guidelines: Who controls what a government chatbot recommends? And who monitors its use?

Looking on the bright side, how will this model develop further, and can such a platform also be used for democratic participation and global cooperation?

Why it matters: This initiative shows for the first time how AI is being systematically integrated into administration, with far-reaching effects on efficiency, control, and security.

At the same time, it opens up the discourse on responsible, transparent AI use in government institutions.

Sources:

🔗 https://de.yahoo.com/?p=us

🔗https://www.theregister.com/2025/06/10/trump_admin_leak_government_ai_plans/

Find out why 1M+ professionals read Superhuman AI daily.

In 2 years you will be working for AI

Or an AI will be working for you

Here's how you can future-proof yourself:

1. Join the Superhuman AI newsletter – read by 1M+ people at top companies

2. Master AI tools, tutorials, and news in just 3 minutes a day

3. Become 10X more productive using AI

Join 1,000,000+ pros at companies like Google, Meta, and Amazon that are using AI to get ahead.

Sign up and start learning AI

Qwen3 Models Arrive on Apple's MLX Framework

The powerful Qwen3 AI models have been officially released in the MLX format, specifically optimized to run efficiently on Apple's machine learning framework. To suit various hardware and performance needs, developers can access them in four distinct quantization levels: 4-bit, 6-bit, 8-bit, and BF16.

Accuracy takes time: LLMs with higher benchmark scores take longer to run.

Are machines smarter than venture capitalists?

Quantitative AI-based VC funds such as China's Baiont and the US company QuantumLight use AI algorithms to make investment decisions largely autonomously.

While Baiont uses AI for short-term price forecasts, QuantumLight invests in 17 start-ups based on data and dispenses with traditional VC roles such as lead investments or board memberships

Turning point: This signals a structural change in the venture capital sector – from human intuition to algorithm-based decision-making logic.

Macro relevance: If AI gains a foothold in the illiquid VC market, it could accelerate and diversify fund allocation and establish new valuation patterns – with implications for venture capital strategies and regulatory design.

The young fundies using AI to beat the index

Minotaur Capital, an Australian fund, has achieved a YTD return of 23.5% with six employees and the AI platform “Taurient,” compared to MSCI ACWI with 17.4%.

It analyzes 35,000 articles weekly and develops AI agents to estimate addressable markets.

Turning point: A new era of lightweight asset alphas without large teams – AI leads to efficiency gains in research and portfolio construction.

Economic relevance: Such models could shift industry standards – traditional asset managers must integrate AI or fight for survival, challenging margins, structure, and the cost landscape in asset management.

How artificial intelligence is transforming finance (IESE Insight)

A report by CEPR & IESE highlights how AI is revolutionizing efficiency, credit and risk analysis, and trading processes in financial institutions—but at the same time creating new systemic risks, for example through model homogeneity or reduced data transparency.

Turning point: AI is shifting the architectural foundations of the financial system – from static models to dynamically learning systems.

Consequences: Central banks and regulators must adapt macro- and microprudential approaches, such as real-time monitoring, model stress tests, anti-herding measures, and AI governance, in order to maintain stability and competitiveness.