Most political news lives in the extremes.

It’s either rage bait meant to rile you up, or echo chambers that only reinforce what you already believe. The result? More division, less understanding — and a lot of burnout.

Tangle does it differently.

We give you the full story on one major political issue a day, broken down with arguments from the left, right, and center — plus clear, independent analysis. No spin. No shouting. No bias disguised as truth.

In just 10 minutes a day, you’ll actually understand what’s happening — and how all sides see it.

Join 400,000+ readers who are skipping the noise and getting the full picture.

Start reading free today.

Today:

• Build 3-D Games with One Prompt? GPT-5 Just Did

• GPT-5 Is Here: Better at Reasoning, Coding, and Even Health Advice

• Grok Will Now Serve Ads: Here’s What That Means for Users

• Meta Acquires Voice AI Startup to Fix LLaMA’s Weak Spot

• Ideogram API Brings Consistent Characters to AI Images

• Microsoft Delays Grok 4 Rollout After Controversial Responses

GPT-5 Just ONE-SHOT The World

GPT-5 tops every public test and is now the default model for over a billion ChatGPT users, including the free tier. Early demos show it building complex 3-D web games (drone flyer, city builder, moon lander) from a single plain-English prompt, and nearly cloning Minecraft. 

It follows instructions precisely, thinks longer when needed, and still cuts mistakes. Minor bugs appear, but the leap in speed, range, and reach is clear.

WATCH THE VIDEO ON YOUTUBE

GPT-5 Is Here — Better at Reasoning, Coding, and Even Health Advice

OpenAI’s new GPT-5 model powers ChatGPT. It thinks faster, makes fewer mistakes, and chooses when to spend more time reasoning. It beats earlier models in writing, coding, math, health advice, and reading images and video. A “Pro” version tackles even harder questions. Built-in safety training offers partial answers instead of blunt refusals and cuts false information almost in half, while reducing flattery and empty praise for greater user trust overall.

Why this matters

1. Raises the bar for capability – GPT-5’s broad gains in coding, math, health and vision set a new performance baseline that other labs must match or exceed.
   

2. Shifts safety practice – Moving from simple “refuse” rules to nuanced partial answers shows a practical path toward safer yet still helpful AI responses.
   

3. Widens real-world impact – Full rollout to free and paid users means expert-level reasoning is now accessible to millions, accelerating AI adoption across work and daily life.

Grok Will Now Serve Ads — Here’s What That Means for Users

Elon Musk told advertisers that X will start placing ads inside answers from its Grok AI chatbot. He said this step will fund the costly computer chips and revive X’s shrinking ad business after ex-CEO Linda Yaccarino left. Marketers will be able to buy spots that appear as suggested solutions when users ask Grok questions. Musk will also use xAI’s targeting tools to improve ad relevance.

Why it matters

1. New funding model – Proving that short, search-like ads can pay for free chatbots shows one path to earning money from large-scale AI.
   

2. AI-driven ad targeting – Letting Grok itself decide which ad best fits a question blurs the line between “answer” and “advertisement,” hinting at future AI-guided marketing.
   

3. Platform race heats up – Tight integration of xAI tech with X pushes Musk into direct competition with OpenAI, Google and Meta over how people discover information and products through AI.

Meta Acquires Voice AI Startup to Fix LLaMA’s Weak Spot

Meta bought WaveForms AI, a young startup that detects emotions and recreates them in speech. The price was not disclosed but follows a recent $40 million funding round. Founders Alexis Cono, ex-OpenAI voice lead, and Corraly Lemaitre, ex-Google strategist, move to Meta’s new Super Intelligence Lab. The deal aims to fix LLaMA-4’s weak voice chat and backs Meta’s larger push into global voice AI after a $14.3 billion data-labeling (manual tagging) spree.

Why it matters

1. Push toward emotional speech AI – Merging emotion detection with lifelike voices could make future assistants far more natural and persuasive.
   

2. Escalating talent race – Meta’s hires and buyouts show how fiercely tech giants compete for scarce researchers, driving up AI costs and pace.
   

3. Signal of platform priorities – Delaying LLaMA-4 to perfect voice chat underscores that user-friendly, multimodal interaction is now a core battleground in the AI stack.

🧠RESEARCH

Is Chain-of-Thought Reasoning of LLMs a Mirage? A Data Distribution Lens

Chain-of-Thought reasoning in large language models often fails when test data differs from training data. This study shows that the reasoning is not truly robust but mimics patterns seen during training. When pushed outside those patterns, the reasoning breaks down, revealing its limits and lack of real understanding.

VeriGUI: Verifiable Long-Chain GUI Dataset

VeriGUI is a new dataset designed to test how well AI agents handle complex computer tasks that involve many steps. Unlike earlier datasets, it checks each step of a task, not just the final result. Tests show current agents struggle with these long tasks, pointing to a need for better planning.

Efficient Agents: Building Effective Agents While Reducing Cost

This study introduces Efficient Agents—a design that keeps AI agent performance high while cutting costs. By analyzing task complexity and system design trade-offs, the researchers show their framework maintains 96.7% of top agent performance but costs 28% less to run. It offers a more affordable path for building capable AI agents.

🛠️TOP TOOLS

NetworkAI - AI-powered networking tool developed by Wonsulting to streamline and enhance the job search process.

ClipDrop Uncrop - AI-powered tool designed to modify the aspect ratio of images by extending their backgrounds, allowing users to enhance and reformat their photos effortlessly.

Map This - Transform PDF documents into visual mind maps, simplifying the process of organizing and summarizing information. 

SinCode AI - AI platform designed to enhance productivity and creativity for businesses, marketers, and content creators.

OpenAI Playground - Web-based interface that allows users to interact with and experiment with OpenAI’s language models without requiring coding skills. 

📲SOCIAL MEDIA

🗞️MORE NEWS

• Ideogram’s new API lets developers easily create characters with the same look across images. They can control traits like clothes and hair, place characters anywhere, and reuse designs for ads, games, and videos.

• Microsoft is delaying a full rollout of Grok 4 on Azure after the chatbot made pro-Hitler comments. Instead, it's quietly testing the AI for safety concerns in a private preview through red-teaming.

• Tesla has shut down its Dojo supercomputer project and disbanded the team, marking a major shift away from in-house AI chip development. The company will now rely more on external partners like Nvidia and Samsung.

• The Browser Company has launched a $20/month Dia Pro plan for its AI browser, offering unlimited chat features. Free users retain access but may face limits. More pricing tiers may follow, including cheaper options.

• Microsoft has integrated GPT-5 into its Copilot tools across Windows, Mac, and mobile. Users get advanced features or free access via Smart Mode, while developers can use GPT-5 in GitHub, Azure, and Copilot Studio.

• Researchers showed that a hidden prompt in a Google Doc could make ChatGPT leak data from a user’s Google Drive without their action. OpenAI patched it, but experts warn similar AI exploits remain possible.

• AWS has launched Automated Reasoning checks for Amazon Bedrock, using formal logic to detect AI errors with up to 99% accuracy. It helps businesses ensure AI outputs follow strict rules, reducing risks from hallucinations.

• Grok 4 outperformed GPT-5 in the ARC-AGI reasoning benchmark but at a higher cost. GPT-5 remains more affordable overall. Smaller versions trade accuracy for price, while OpenAI’s o3-preview still holds the top ARC-AGI-1 score