Elon Just Dropped an AI 4x Cheaper Than Claude at the Same Level (+14 AI Updates)

12 Major AI Updates You Need to Know

This video covers the biggest AI developments of the week, including new AI agents, local AI models, coding models, AI-powered browsers, image generation, AI infrastructure, and automated AI workers.

1. Grok Bots: AI Employees That Can Do Real Work

xAI introduced Grok Bots as AI teammates that can perform real tasks. Users can teach a bot how to complete a workflow, allowing it to learn the steps and repeat the task later.

Each Grok Bot gets its own computer in the cloud, allowing it to open websites, send emails, access connected services, and complete tasks while the user continues working. Bots can also communicate with one another and hand work from one bot to another.

2. Meta's Local AI Model

Meta released Muse Glimmer, a free 30-billion-parameter model designed to run locally in under 20 GB of memory. The video demonstrates how a local AI model can build and operate a dashboard while keeping data on the user's computer.

3. Z.ai Launches GLM 5.3

Z.ai launched GLM 5.3 with a focus on coding, AI agents, and cybersecurity. The model is presented as an improvement over GLM 5.2, particularly for coding tasks and agentic workflows.

4. Four Major OpenAI Updates

1. GPT 5.6 Cyber for advanced cybersecurity tasks.
2. An ultrafast mode designed to make GPT 5.6 work significantly faster.
3. Integration with Google Docs, Sheets, Slides, and Google Drive.
4. A computer history feature that can summarize previous activity across applications.

5. Grok 4.6

Grok 4.6 is presented as having improved understanding of large codebases, longer task execution, integrations with tools such as GitHub, Slack, Google Calendar, and Figma, and stronger built-in coding capabilities.

The video also highlights improvements in design generation and compares its token pricing with competing AI models.

6. Google's Offline AI Translator

Google demonstrated a DIY AI translator running on a Raspberry Pi 5. The system includes a touchscreen, speaker, keyboard, battery, and a small Gemma model. It can translate spoken language without requiring an internet connection.

7. Claude Becomes an AI Co-Worker Inside Chrome

Claude's Chrome side panel can work with information across open browser tabs. The video demonstrates using Claude to read invoices, prepare a month-end report, and update a connected Google Sheet.

8. Nvidia's AI Infrastructure Investment Plan

The video discusses Nvidia's plans involving major financial partners to support investment in AI infrastructure and data-center hardware.

9. Grok Imagine Image 2.0

Grok's image model is presented as moving toward a broader AI design workflow. Features highlighted include selective image editing, smart resizing, improved instruction following, and templates for e-commerce photography, UGC images, and game assets.

10. DeepSeek V4 Pro

DeepSeek V4 Pro is designed for more demanding agentic workloads, including coding, tool use, and long-running workflows. The video highlights its performance on agent benchmarks and changes to off-peak API pricing.

11. Hermes Agent Bot Mode

Hermes Agent introduced a bot mode where users can create individual agent profiles with their own roles, descriptions, identities, and memory. Multiple bots can communicate and work together, such as a developer bot and a testing bot.

12. Claude AI-Generated Content Watermarking

Claude is presented as introducing invisible watermarks for AI-generated text, images, and files. The video explains that verification tools can potentially provide information about the origin and editing history of generated content.

Grok Bots Tutorial

The second part of the video provides a practical demonstration of Grok Bots.

Creating AI Teammates

Users can create multiple bots for different jobs, such as AI research, design, YouTube analytics, partnerships, and content workflows.

AI Research Bot

A research bot can monitor AI developments, compare new information with previous reports, identify important updates, provide sources, and request approval before posting information to a team Slack channel.

Design Bot

A design bot can be assigned creative tasks such as creating animations and visual assets.

YouTube Growth Bot

A YouTube analytics bot can connect to channel data, analyze views, subscribers, watch time, CTR, and video performance, then create recurring reports and recommendations.

Connecting Business Tools

Grok Bots can connect with tools including Gmail, Google Calendar, Google Drive, Slack, and other applications. This allows bots to perform tasks using the same services people use in their daily workflows.

Building a Partnership Email Agent

The video demonstrates creating a partnership agent that checks incoming emails, identifies relevant messages, drafts replies, and waits for user approval before sending them.

Teaching a Bot by Demonstration

With the teacher task feature, users can record themselves completing a workflow. Grok observes the actions and converts the demonstrated process into a reusable skill that can later be repeated automatically.

AI Bots Working Together

A chief-of-staff bot can coordinate other specialized bots. For example, it can ask a research bot for the newest AI update and return the result to the user. This creates a workflow where multiple AI workers operate as a team.

Key Takeaways

1. AI agents are moving from answering questions toward completing real-world tasks.
2. AI Bots can learn repetitive workflows through demonstrations.
3. Cloud-based AI workers can operate their own virtual computers.
4. Local AI models can perform useful tasks without sending data to the cloud.
5. AI tools are increasingly connecting directly with workplace applications.
6. Multi-agent systems allow specialized AI workers to collaborate.
7. Recurring AI routines can automate research, reporting, email, and analytics.