Aug 18th - Weekly AI update
From GPT-5 updates, the personalization revolution, to AI-generated podcasts

Link to the Hebrew version

Let’s start with a story from last week in China: the first Humanoid Robot Olympics.
Hundreds of robots from 16 countries competed in sprinting, soccer, kickboxing, and even daily tasks like sorting medicine and cleaning.

The fascinating part wasn’t just the successes – but the failures. Robots that stumbled, collided, or moved painfully slow. A vivid reminder of how wide the gap still is between the dream of human-like robots and today’s reality.

For China, this was more than a spectacle. It was a strategic move to showcase leadership in robotics and to test, in real time, how humanoid robots might integrate into industry and daily life in the future.

Now, back to what’s a bit more real – and already reaching us.

1. GPT-5: A Week and a Half Later 📱

We’re past the initial hype. What has actually changed since the launch?

• Fast vs. Thinking mode – You can now choose between quick answers (Fast) or deeper reasoning (Thinking). Like deciding between an espresso shot or a slow-brew Turkish coffee.
• All models unlocked – Previous models, including GPT-4o, are available side by side.
• Reduced hallucinations – Benchmarks show significant improvement, and you can even see how the model verifies its own reasoning in Thinking mode.
• Code execution & UI generation – GPT-5 can generate not just code, but actual UIs (HTML, landing pages, even simple games). Early, but promising.
• Extended memory – Try a time check: the model now recalls more about you and past conversations.
  

Is it perfect? No. Some answers feel overly concise, and the warmth it tries to project sometimes feels… artificial.
But it’s definitely a step forward.

👉 Curious about how it stacks up against Google Gemini 2.5? Here’s a head-to-head test.

2. The Personalization Revolution 🎯

If memory and personalization are what keep me coming back to ChatGPT, this week showed the competition is catching up.

• Claude’s new memory & personalization
  
• Search across past conversations (a long-awaited feature).
• Practical advantage: when a thread gets too long and cuts off, you can now pick up where you left off.
• Rolling out now to Teams and Max, soon for all users.
  
• Gemini’s memory with privacy controls
  
• Opt-in memory (you choose when to turn it on).
• Temporary chats that leave no trace.
• Ability to view, edit, and delete stored data. Read more here.
  

For the first time, all major AI assistants are moving toward real personalization.
The big question is: Who will win our trust – the one who knows us best, or the one who respects our privacy most?

3. AI-Generated Podcasts – GenSpark AI Pod 🎙️

Turning research into an audio conversation isn’t new – but GenSpark took it up a notch.

Why I find it useful: when I need to dive into a new topic, I can gather articles, feed them in, and listen to the material as a podcast in the background. It helps me absorb and process much faster.

GenSpark goes further:

• Builds podcasts from any material.
  
• Positions itself as a “super-agent” that can also call on your behalf, create presentations, produce videos, and manage documents.
  
• Recently raised $36M in just 45 days with a team of only 20 people.
  

They even introduced a coding agent this week – translating natural language instructions directly into working code.

📝 Wrapping Up

This week felt like the shift from “wow, a cool tool” to “this is becoming my actual work partner.”

Personalization and memory aren’t just features – they’re what transform AI from a machine into something closer to a personal assistant.

The real question isn’t just which model is best, but rather:
👉 How are you building a relationship with your AI?

✨ What’s new on my side

• My AI course kicks off in early September (now with a module on AI-powered coding). Reach out if you’d like to join.
  
• For companies and teams, I offer workshops and tailored consulting on AI adoption and automation.
  
  Let's talk

Here’s to a week full of learning, experimentation, and better partnerships with our AI.
— Dalit ❤️