OpenAI just made Codex way more dangerous.
Their new Chrome extension lets the AI agent hijack your actual browser the one where you're already logged into LinkedIn, Gmail, Salesforce, everything.
No more API limits or sandbox walls. Just pure automation using your real credentials.
But wait, there's more:
Why you should ditch Markdown and make AI spit out HTML instead
Meta just killed Instagram DM encryption and it's worse than you think
I'm Alex. Welcome to L8R by Innov8.
Let's dive deep 🐰
In today's post:
OpenAI's Codex now controls your actual browser'
Forget Markdown HTML is how you should prompt AI
Meta killed Instagram's encrypted DMs (and can read everything now)
Codex Just Took Over Your Browser
OpenAI just dropped a massive update that lets their Codex AI take over your active browser tabs.
The new Chrome extension allows the AI agent to click, type, and read right inside your signed-in accounts like Gmail and LinkedIn.
You no longer have to worry about broken APIs or getting locked out of secure tools.
The details:
The extension launched on May 7, 2026, and connects the Codex desktop app directly to Google Chrome.
Codex can now automate tasks on websites where you are already logged in, bypassing the need for complex API setups.
OpenAI added strict security rules, meaning the AI must ask your permission before it touches a new website.
The desktop app now boasts over 4 million weekly active users, growing eight times larger since early 2026.
It only works on regular Google Chrome right now, leaving users of Brave, Edge, and Arc waiting.
Why it matters:
This changes how we use AI at work.
Until now, AI agents struggled to use enterprise tools because they got blocked by login screens.
By piggybacking on your active browser session, Codex acts like a real human clicking through your daily tasks.
It is a huge leap toward AI that actually runs your computer.
💡 L8R's Take:
I love this update, but it is a security nightmare waiting to happen.
Giving an AI full access to your signed-in email or work tabs is a massive risk, even with OpenAI's permission pop-ups.
You should use this tool, but keep it on a very short leash..
Stop Asking AI for Markdown. Ask for HTML Instead.
Markdown is holding back your AI outputs.
An engineering lead at Anthropic just shared a massive tip for getting better results from Claude.
If you want interactive widgets and rich diagrams, you need to start asking your AI to write in HTML.
The details:
Anthropic's Thariq Shihipar says HTML is far better than Markdown for complex AI answers.
HTML lets AI build interactive playgrounds, SVG diagrams, and advanced navigation right inside your results.
Older models struggled with HTML because of the high token cost, but modern LLMs handle the extra tags easily.
You should still use Markdown if you plan to edit the text yourself, but use HTML for final reports or spec sheets.
Developer Simon Willison highlighted this shift on May 8, 2026, noting it completely changes his default way of talking to AI.
Why it matters:
We are used to getting flat text from AI.
Moving to HTML changes AI from a simple writer into a software builder.
You can now prompt an AI to build a working tool and use it instantly.
This unlocks a whole new level of quick prototyping for developers.
💡 L8R's Take:
I am switching my default prompts to HTML right now.
Markdown is great for quick notes, but it limits what AI can actually do.
If your AI can code a full interactive widget, do not hold it back with boring text formats..
Meta Just Killed Instagram DM Privacy
Meta just flipped the script and killed end-to-end encryption for Instagram DMs.
They rolled it out late last year as a huge safety feature, but now they want full access to scan your private messages.
This is a massive blow if you use Instagram to close deals or share sensitive business details.
The details:
Meta officially removed the opt-in encryption feature for all Instagram direct messages on May 8.
Wired broke the news that Meta will now scan every single DM you send to feed their ad machine and train AI models like Llama.
Creators and solopreneurs lose a major safe space for sending client contracts, pricing details, and NDA chats.
European Union regulators might launch GDPR probes soon, which could lead to massive privacy fines for Meta.
Why it matters:
Indian creators rely heavily on Instagram to build huge audiences without spending cash.
But Meta only cares about feeding its hungry AI models and selling targeted ads.
Every single business strategy or client chat you send is now free data for them.
Big tech companies will always throw you under the bus to please their shareholders.
💡 L8R's Take:
Meta is completely full of it.
They only used privacy as a PR stunt until they needed your chats to train their AI.
Stop feeding their machine and move your client chats to Signal or Telegram right now..
🚀 Quick L8R Summary
OpenAI Codex Extension: OpenAI dropped a Chrome extension that lets Codex control your actual browser logged-in sessions and all.
HTML Over Markdown: An Anthropic engineer says stop asking AI for Markdown and start demanding HTML for better formatting and interactivity.
Instagram Encryption Gone: Meta killed end-to-end encryption on Instagram DMs after barely rolling it out they can now read everything you send.
📩 Innathe L8R engane undarunnu 👇?
We read every reply - just reply to this email and let us know how we can improve !
Appo adutha L8R il kanaam bie…👋