5 Things You Should Know in AI This Week — January 9, 2026

Meta just bought Manus, a Singapore-based AI agent startup, for over $2 billion.

Haven't heard of Manus? You're not the only one.

Here's an article I wrote when Manus launched back in March.

It's been one of the most overlooked agent tools out there, even though it's probably the most approachable one if you're non-technical.

I regularly use Manus for online research.

Unlike standard chatbots that can give you old info, Manus pulls fresh data.

I use it for things like researching competitor sites, prepping for client meetings, pulling social content, and generating niche strategies.

It's also great at creating support documents based on its findings.

The Meta acquisition is intriguing.

They'll keep Manus running as its own product, but also start integrating agents into WhatsApp, Instagram, and Facebook.

The signal?

This is Meta's first move as they try to re-enter the AI mix.

They've been quiet since their underwhelming Llama 4 launch and hiring spree last year.

This move makes them a legitimate agent provider overnight.

Expect to see Manus-style features showing up across Meta's apps.

If you haven't tried Manus yet, it's worth trying out.

(I may do a deeper dive on it in an upcoming Sunday newsletter.)

They offer a free tier, so you can test it out for research without handing over sensitive company data.

xAI launched Grok Business and Grok Enterprise plans this week.

This means higher usage limits, tighter security controls, and a promise not to use your data to train models.

While Grok started consumer first with its edgy personality, they've been going after enterprise customers the past few months.

Their main challenge? Gaining corporate trust.

This launch comes as they face more public backlash around deepfakes and non-consensual image use.

Constant hiccups make enterprise adoption harder, despite their efforts.

Grok also secured $20B in a Series E funding this week (above their $15B target).

Investors include Nvidia, Cisco, Fidelity, and Qatar Investment Authority.

The funding shows xAI's has staying power in the AI space, especially for a company that was a smaller competitor until now.

The signal?

This is all about revenue.

AI orgs need money from:

• Enterprises (big contracts)
• Consumer subscriptions (only 3-5% of users pay)
• Someday ads...

Enterprise contracts are where the money is.

But most enterprises won't make Grok their primary platform.

It's more likely that specific teams test Grok for niche uses.

Grok's differentiator is real-time social data from X.

Social, marketing, and research teams could benefit from this.

For you, it's worth knowing Grok is in the mix now despite their issues.

What does that mean?

AI competitors can now adopt the Skills format and roll it out on their platforms.

If you've followed along, I've been hyping Skills for months:

How to Set Up Claude Skills in <15 Minutes (for Beginners)
How to Set Up Claude Skills (for Teams)

Quick recap: Skills are reusable instructions that tell AI how to do specific tasks consistently.

They're genuinely useful.

And if you use Claude, I urge you to try them if you haven't already.

Sidenote, Anthropic also raised funds this week ($10B).

The signal?

Expect to see Skills (or something similar) on ChatGPT and Gemini soon.

And if other platforms adopt the standard, you'd be able to use your Skill files across multiple AI tools.

This isn't Anthropic's first time doing this.

In 2024, they created and open-sourced Model Context Protocol (MCP = a standard way for AI to connect with apps and data).

Anthropic's strategy: build the best tools, then share the blueprint so it becomes the industry standard.

A dedicated health tab with separate chat history and memory.

You can connect health data from Apple Health, MyFitnessPal, Peloton, Weight Watchers, and even healthcare providers (through b.well for US users).

Plus, the data won't be used to train OpenAI's models.

As expected, there are disclaimers protecting OpenAI.

They've framed it as a tool to prepare for your doctor visit, not a replacement for your doctor.

I recall OpenAI mentioning health use cases several times during their GPT-5 model release in the summer.

At the time, I thought it was risky given health liabilities.

And strange how quiet they'd been about it since that launch.

With 230 million users asking health-related questions weekly, they're leaning into emerging behaviors instead of avoiding risk.

The signal?

Health aside, this is a blueprint for AI tools in other domains:

• Responses tuned for a specific industry
• Easily connect your data
• Separate memory and privacy controls

Expect similar vertical experiences for things like legal, finance, and other regulated fields.

This also aligns with one of my 2026 predictions:

This will be the year specialized tools emerge.

• More than 20% of YouTube's feed is now AI slop.
  • According to Kapwing, 1/5 of videos recommended to new users are low-quality, auto-generated content made to farm views.
  • Source: Kapwing
• Nvidia's next-gen AI chip is 5x faster and arriving six months early.
  • Their Vera Rubin platform will reach customers the second half of 2026, cut inference costs by 10x, and need far fewer GPUs to train large models.
  • Source: Nvidia