A lot happened in AI this week.

OpenAI made a major strategic hire that signals where agents are heading.

Google officially launched Gemini 3.1 and upgraded its reasoning engine.

Several product updates quietly pushed AI further into everyday workflows.

In this issue, I’ll break down the key updates that matter for business operators, and then go deeper on two stories:

Let’s get into it.

OK, let's get into it!

As of this past weekend, Peter Steinberger, founder of OpenClaw (formerly Clawdbot), is joining OpenAI to lead their push into personal agents.

Steinberger built the original prototype in about an hour. He connected WhatsApp to an AI coding tool, sent it a message, and got a useful reply back. That was November. By January, OpenClaw had become one of the fastest-growing repositories in GitHub history, pulling in over 200,000 stars and 2 million visitors in a single week.

Before you think this was just a lucky side project, Peter previously built PSPDFKit, a software used by companies like Apple and Dropbox. Nearly a billion people use apps powered by his technology. He bootstrapped that business for 13 years and exited successfully. This is a seasoned operator.

The backstory is one of the more fascinating sequences in the AI space this year. The project originally launched as “Clawdbot.” Anthropic’s legal team asked him to change the name due to its similarity to Claude. Fair enough. He rebranded, and ultimately landed on OpenClaw. Days before the announcement, he shared in a Lex Fridman interview that he was choosing between Meta and OpenAI. Mark Zuckerberg was personally testing the product and texting him feedback. Satya Nadella called him directly. Every major VC firm was reaching out.

Ultimately he chose to go with OpenAI. In his blog post, he explained his decision by saying “It’s always been important to me that OpenClaw stays open source and given the freedom to flourish. Ultimately, I felt OpenAI was the best place to continue pushing on my vision and expand its reach.”

OpenClaw will now move to an independent foundation and remain open-source, with OpenAI sponsoring the project.

So what does this actually mean for you?

This hire signals where AI is heading next. We are moving from AI you type questions into, to AI that executes tasks on your behalf. Sam Altman has said the future is multi-agent.

Think about what that looks like in your day-to-day. Instead of asking ChatGPT to draft a follow-up email, imagine an agent that drafts it, checks your CRM, schedules the send based on your calendar, and automatically follows up if there is no reply in three days. That is the shift.

Steinberger has suggested that agent-style systems could eventually replace a large percentage of the apps on your phone. That might sound aggressive today. But when you consider how many apps exist just to manage tasks and handoffs, it becomes more plausible.

Here is the practical move that puts you ahead of most business owners.

Start identifying the tasks in your business that are repetitive, time-consuming, and require minimal judgment. For example:

These are prime candidates for agent-style automation.

The businesses mapping these workflows now will be ready when agent tools become more seamless and embedded. Based on this week’s moves, that timeline is accelerating.

What to watch next.

If OpenAI integrates OpenClaw-style capabilities directly into ChatGPT, your AI assistant could manage your calendar, handle customer communication, and coordinate across tools from a single conversation. We have already seen AI move from answering questions to generating content. The next step is operational execution.

I will keep you posted as this develops.

Claude has been making some serious moves lately, but the biggest hesitation I keep hearing is: “I’ve spent months building all that context in ChatGPT. I don’t want to start over.”

A few days ago I posted in my Circle community about why ChatGPT is still my daily driver, even though I also love Claude. A lot of you felt the same way. It’s not about which model wins a benchmark. It’s about the fact that ChatGPT knows you. Your business. Your style. Your preferences. That context compounds and makes every conversation better.

The good news is you do not have to rebuild that from scratch.

If you want to move your ChatGPT context into Claude like I recently did, you can do it in about 10 minutes:

1. Export your data from ChatGPT.

Go to Settings → Data Controls → Export Data. ChatGPT will email you a download link.

2. Locate the file called chat.html.

Extract the ZIP file and find chat.html. That file contains your full conversation history.

3. Create a new Project in Claude and upload it.

Name it something like “ChatGPT Hub” and upload the file. If the file exceeds Claude’s size limit for your plan, open it in your browser, select all, copy, and paste the contents directly into the Project instead.

4. Ask Claude what it knows about you.

Inside that Project, start a conversation and ask:

“What do you know about me based on my ChatGPT history?”

You should see your business context, writing style, and patterns reflected back.

5. Save the important patterns to Claude’s memory.

This is the step that makes it stick. Ask Claude to extract the most important insights and save them to its memory. Once stored, those patterns can apply beyond just that single Project.

That’s it.

No starting from scratch. No re-explaining your business for the hundredth time. Just your existing context, now portable across tools.

From an operational standpoint, this gives you flexibility. You can test different models without losing the compounding value of your history.

Try it out and hit reply to let me know what Claude surfaces. I’m curious what patterns it picks up for you.

If you want to go deeper on any of this, that's exactly what we do inside the Circle community.

See you next week,

Julien

PS: If you found this useful, forward it to someone on your team or a colleague who could use it. www.ampra.ai/join-our-newsletter.