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

Today:

• Meta Seeks Funding For Llama

• Google Gives Students Free AI Access

• Anthropic Invests In AI Transparency

• OpenAI Introduces Flex Pricing Option

• Gemini Flash Adds Thinking Budget

Meta Seeks Funding For Llama

Meta asked Microsoft, Amazon, and others to help pay for training its open-source AI model, Llama, but got a lukewarm response. The plan, called the “Llama Consortium,” faced pushback as companies didn’t want to fund a free product. Meta’s ad-driven business model complicates monetizing Llama. With rivals like DeepSeek rising and AI costs soaring, Meta faces pressure to fund growth without hurting shareholder returns.

Why this matters

1. Funding strain: Even top tech firms are feeling the financial pressure of AI development, highlighting how costly large-scale models are becoming.
   

2. Open-source tensions: Meta’s struggle to attract partners shows the challenge of monetizing open-source AI when rivals are building profitable closed systems.
   

3. Industry shake-up: DeepSeek’s rise and Meta’s funding search may shift power dynamics among AI model creators, cloud platforms, and business users.

Google Gives Students Free AI Access

Google is giving U.S. college students free access to Gemini Advanced, NotebookLM Plus, Veo 2, and 2TB of storage through spring 2026. Students get top AI tools for writing, research, creativity, and studying—plus AI built into Docs, Sheets, and Slides. This offer, available until June 30, aims to support smarter learning during exam season and beyond.

Why this matters

1. Wider adoption: It accelerates AI familiarity and usage among young users, shaping future expectations of productivity tools.
   

2. Competitive edge: Google boosts its AI ecosystem against rivals like OpenAI and Microsoft by embedding powerful tools in its everyday apps.
   

3. Education shift: It pushes AI deeper into academic workflows, potentially reshaping how students learn, research, and create.

Anthropic Invests In AI Transparency

Anthropic invested $1 million in Goodfire, a startup using AI to explain how other AI models think. Goodfire trains interpreter models that reveal hidden concepts inside large models, helping businesses like Rakuten spot risks like hallucinations. This work builds on Anthropic’s own research and aims to make AI more understandable and controllable. As AI grows more powerful, tools like Goodfire’s offer a path toward safer, more transparent systems.

Why this matters

1. AI transparency: Goodfire helps crack open black-box models, enabling developers to understand and guide AI behavior more clearly.
   

2. AI safety: By spotting risks like hallucinations and sensitive data leaks, Goodfire’s tech offers practical tools to reduce harm in deployed AI systems.
   

3. Strategic alliances: Anthropic’s support signals growing momentum behind “mechanistic interpretability” as a key strategy for aligning advanced AI with human goals.

🧠RESEARCH

ColorBench: Can VLMs See and Understand the Colorful World? A Comprehensive Benchmark for Color Perception, Reasoning, and Robustness

ColorBench tests if AI models truly understand color. Evaluating 32 models, researchers found that language components matter more than vision encoders. Step-by-step reasoning helps, but models still often miss or misuse color cues. Results show color understanding is weak in current systems, highlighting a key blind spot in AI.

BitNet b1.58 2B4T Technical Report

BitNet b1.58 2B4T is the first open-source 1-bit large language model with 2 billion parameters. Despite using ultra-efficient 1-bit precision, it matches the performance of full-precision models in tasks like coding and reasoning. It uses far less memory and energy, and is available on Hugging Face for easy use.

ReTool: Reinforcement Learning for Strategic Tool Use in LLMs

ReTool is a new method that teaches language models to use tools like code interpreters during reasoning. It combines real-time code execution with reinforcement learning to help models decide when and how to use tools. ReTool outperforms baselines on tough math tasks, showing improved accuracy, efficiency, and even self-correction behaviors.

🛠️TOP TOOLS

Prompt Silo - AI-powered tool designed to generate creative content ideas for MidJourney, an AI art generation platform.

Sketch2App - AI-powered tool that transforms hand-drawn sketches into functional application code.

FineCam - AI-powered virtual camera software that transforms various devices into high-definition webcams for video conferencing and content creation.

Scrip AI - AI-powered content creation tool designed to streamline the process of generating engaging digital content across multiple platforms.

Scribble AI - automatic text generator powered by OpenAI’s GPT-3 Language Model, designed to create a wide range of content quickly and efficiently.

📲SOCIAL MEDIA

🗞️MORE NEWS

• OpenAI launched “Flex” pricing for its o3 and o4-mini models, cutting API costs in half for slower, less urgent tasks. It targets budget use cases as rivals like Google release cheaper, efficient AI alternatives.

• Google’s Gemini 2.5 Flash introduces a “thinking budget” that lets developers control how much reasoning AI uses, cutting costs by up to 600%. It balances smart outputs and speed, giving businesses flexible, cost-efficient AI.

• OpenAI’s $500B Stargate project, backed by Oracle and SoftBank, may expand beyond the U.S. to the U.K., Germany, and France. Originally focused on U.S. AI infrastructure, it’s still raising its first $100B in funding.

• Scale AI is partnering with Qatar to build AI tools for education, healthcare, civil services, and more. The five-year deal includes AI agents for schools, hospitals, and transportation. Scale expects 2025 revenue to double to $2 billion, driven by global government contracts and a growing push into AI-powered infrastructure and services.

• The Trump administration is considering banning Chinese AI lab DeepSeek from buying Nvidia chips and restricting U.S. access to its services. This move aims to curb China’s AI rise amid IP theft allegations and rising U.S. developer use.