Hey,

Last week, we talked about "The Infrastructure Flip."

Google and Microsoft were quietly rewriting how browsers talk to AI.

But this week, the dam broke.

Every major player just shipped autonomous agents.
And they all made the exact same terrifying admission.

They have no idea how to secure them.

Here is the signal for the week of Mar 08 – Mar 15.

1. Microsoft Betrays OpenAI

Microsoft just built its new flagship AI on Anthropic.
They called it Copilot Cowork.

They literally licensed the tech from Claude to build it.
And they are charging $99 a user for the privilege.

The OpenAI exclusivity era is officially over.
Microsoft knows multi-model is the only way to survive.

Why this matters:
You can stop worrying about picking the "best" AI model. The biggest software company on Earth just admitted you need them all.

→ The Move:
Stop buying single-model subscriptions. Build your workflows to route tasks to the best model for the job.
(Source: https://fortune.com/2026/03/09/microsoft-copilot-cowork-ai-agents-anthropic-e7-m365-saas/)

2. Your Agent Is Already Hacked

OpenAI just admitted that prompt injection "may never be fully solved."
Let that sink in.

Perplexity's new browser just got hijacked to steal 1Password accounts.
It took zero clicks.

The industry is shipping tools they cannot secure.
They know it, and they are doing it anyway.

Why this matters:
If you give an AI agent read and write access to your private data today, you are playing Russian roulette.

→ The Move:
Isolate your agents. Put them in Docker sandboxes and never give them access to your core accounts.
(Source: https://techcrunch.com/2025/12/22/openai-says-ai-browsers-may-always-be-vulnerable-to-prompt-injection-attacks/)

3. The Death of Data Entry

Google just turned Sheets into a live data pipeline.
You don't type anymore.

You tell Gemini to "organize my move."
It reads your Gmail, finds the quotes, and builds the spreadsheet.

Meanwhile, GPT-5.4 just dropped with native computer use.
It doesn't just write code — it takes over your mouse and tests it.

Why this matters:
If your job involves moving data from one window to another, you are officially competing with software that does it instantly.

→ The Move:
Open Google Sheets and test "Fill with Gemini." Stop copying and pasting. Let the AI pull the live data for you.
(Source: https://blog.google/products-and-platform/products/workspace/gemini-workspace-updates-march-2026/)

The Lightning Round

Claude Code gets interactive: Anthropic quietly updated Claude Code to ask you questions mid-task. It no longer fails silently — it stops and asks for the missing URL. (Source: https://releasebot.io/updates/anthropic/claude-code)

Grok drops four agents at once: Musk just launched Grok 4.20 with a four-agent parallel reasoning setup. It boasts a 78% non-hallucination rate and undercuts competitors on price. (Source: https://www.eonmsk.com/2026/03/03/xai-released-grok-4-20-beta-2/)

Sandboxed agents go mainstream: NanoClaw just partnered with Docker to run AI agents in isolated MicroVMs. This is the only safe way to run autonomous code right now. (Source: https://venturebeat.com/infrastructure/nanoclaw-and-docker-partner-to-make-sandboxes-the-safest-way-for-enterprises)

What's Happening This Week

I see a lot of creators teaching legacy automation tools.
Things like Relay and Lindy.

Those are browser-only apps.
And they are already dead.

Inside Agent J Plus, we teach the future that is already here.
We build with Claude Code, Claude Skills, and OpenClaw.

These are tools that actually replace employees.
Claude Code can do everything those legacy tools do, plus 1000x more. And it is easier to set up.

A lot of people try OpenClaw and give up because it feels too complex.
That is because they skip the fundamentals.
It is like trying to design car engines without understanding how a car works.

We teach the full stack.

If you want to see how we do it, come to the Friday Webinar.
We are building an AI agent live. Only 100 spots.

Or join us in the free Agent J community for the weekly sessions:

• Tuesday: The Builder's Hour — live build of a new workflow.
• Wednesday: Vibe Code Session — from idea to live product.
• Thursday: AI Prompting Masterclass — learn to think like the AI.

The Sunday Action Plan

  1. Cancel your single-model subscriptions. Set up a routing workflow that uses Claude, OpenAI, and Gemini.

  2. Audit your agent permissions. Revoke any access that lets an AI read your private emails or write to your calendar.

  3. Build one automated spreadsheet this week using Google's new natural language integration.

The Bottom Line

The tech giants are shipping features faster than they can secure them.

The winners won't be the ones with the smartest AI.
They will be the ones who know how to build the safest workflows.

See you next Sunday,
John

P.S. Reply "VIRAL" and I'll send you The God Prompt for Viral Content.

Reply

Avatar

or to participate

Keep Reading