Trending AI topics

Meet Gemini 2.5: Google’s Smartest AI Yet

The story in simple terms:

Google launched the most powerful AI model yet. Not in a hype-y way, but in a “this thing actually thinks before it talks” kind of way. The model’s called Gemini 2.5, and it might be Google’s smartest move yet in the AI race.

• Gemini 2.5 Pro scored #1 on LMArena, a benchmark where humans judge AI responses beating models from OpenAI, Anthropic, and DeepSeek.

• On Humanity’s Last Exam, a test with thousands of tough questions across math, science, and humanities, it scored 18.8%, the highest among most flagship models.

• It’s not just book smart on SWE-bench Verified, a coding test, it scored 63.8%, outperforming OpenAI’s o3-mini and DeepSeek’s R1.

• In another dev task, Aider Polyglot, it hit 68.6%, leading the pack in code editing ability.

• It can handle a 1 million token context window that’s like reading the entire Lord of the Rings in one go. Google says a 2 million token upgrade is coming soon.

• Gemini 2.5 isn’t some limited lab thing either. It’s already live in Google AI Studio and the Gemini app for advanced users.

Here’s what we think:

Every big AI company is building fast runners. Google just trained one to think before it runs. And that shift from speed to depth is where things are getting really interesting.

Read the complete article here.

Test the model here today, its live.

Sponsored by GrowthSchool

Join The World’s First 2-Day LIVE AI Bootcamp for $499 FREE🤖

AI is no longer the future. It’s here, and it's revolutionizing every industry & the way we work.

Introducing a 2-Day Free AI Bootcamp by Outskill which comes with 16 hours of intensive training on AI frameworks & use-cases that can make you an AI expert.

Inside the AI Bootcamp, you will learn:

👉 AI tools to automate repetitive tasks and free up time for high-value work.

👉 Smarter decision-making with Generative AI, Neural Networks, and LLMs.

👉 AI-generated images and videos to speed up content creation and marketing.

👉 AI-powered automations to eliminate manual, repetitive tasks.

👉 CustomGPTs and AI Agents to make AI work for you even while you’re asleep.

Join in now, (we have limited free seats! 🚨)

4o Model Turns ChatGPT Into a Design Tool That Understands You

The story in simple terms:

OpenAI’s new image model is actually useful now. For a while, image generation felt like a party trick. Cool, but limited. That just changed. OpenAI gave ChatGPT a serious image upgrade and this time, it’s not just pretty, it’s practical.

• The upgrade comes from GPT-4o, OpenAI’s newest model that now natively generates and edits images no more bolt-on tools like DALL·E.

• It can create photorealistic, text-accurate, and prompt-specific visuals, including logos, diagrams, comics, and even UI mockups.

• GPT-4o can edit existing images, including photos of people changing objects in the background, adjusting the layout, or "inpainting" missing details.

• On the technical side, it handles 10–20 distinct objects in one image, doubling what older models like DALL·E could manage.

• OpenAI trained GPT-4o on a blend of publicly available and licensed data (like from Shutterstock), avoiding legal traps by allowing creators to opt out of its datasets.

Here’s what I think:

OpenAI just made visual creation feel more like conversation. You describe it, it sketches it accurately. That’s a big deal for anyone designing, teaching, or just trying to explain ideas better. It’s no longer about making cool images; it’s about making the right ones.

Test it out, its live.

Poe just dropped a $5 AI pass

The story in simple terms:

Quora’s chatbot app, Poe, just got way more accessible. If you’ve been AI-curious but wallet-wary, this might be your green light.

• Poe now offers a $5/month subscription, its cheapest tier ever. It gives you 10,000 points per day to spend across different AI bots.

• On the flip side, there’s a $250/month plan for power users those who burn through messages with high-cost models like GPT-4.5, OpenAI o1-pro, or Veo 2 from DeepMind.

• Poe works like a digital arcade. You buy points and spend them across a lineup of bots: GPT-4o, Claude 3.7 Sonnet, DeepSeek-R1, ElevenLabs, and more. Each has a different point cost per message.

• For context: the previous cheapest plan was $20/month for 1 million points. Now, $5 gives you a daily drip of credits plenty for casual users who just want to explore.

• There’s also still a free plan, but with stricter limits on daily questions.

• Poe is cross-platform and available on iOS, Android, Mac, and Windows so you can chat with AI wherever.

Here’s what I think:

The $5 plan is smart. It lowers the barrier without making things too open-ended. You get enough power to try different bots, test ideas, or just mess around and that’s how new habits start. It’s not about replacing your tools yet. It’s about having all of them in one pocket.

Test it out here.

AI agents just got a common language and it’s a big deal

The story in simple terms:

There’s a new protocol in town, and it might quietly become the glue that holds the AI ecosystem together. It’s called MCP (Model Context Protocol), and the latest update just got serious backing.

• MCP helps AI agents talk to tools, apps, and data in a unified way—kind of like giving them one shared “API dialect.”

• OpenAI now supports MCP, with integration in its Agents SDK and incoming support for ChatGPT’s desktop app and the Responses API.

• Microsoft joined in too, releasing Playwright-MCP, which lets agents like Claude browse, click, type, and interact with websites—using the Chrome accessibility tree as their guide.

• The new MCP spec adds:
  • OAuth 2.1 for secure agent-to-server access
  • Streamable HTTP for real-time, two-way data
  • JSON-RPC batching for faster, multi-request flows
  • Tool annotations that give agents richer context

• Behind the scenes, it’s all modular—meaning developers can pick and use only what they need without breaking compatibility.

Here’s what I think:

We’re moving from cool demos to real-world agent utility. This isn’t just about talking to AI it’s about letting AI do things, across platforms. With OpenAI and Microsoft now on board, MCP could be the missing puzzle piece that lets agents go from isolated chatbots to full-blown digital coworkers.

Tools and resources

⚒️ AI tools and websites:

Here’re some of the best tools and websites you can visit now to save time:

• Presentations: Gamma

• AI CRM with Smart Suggestions: Folk

• AI Smart Portfolio Builder for Creatives: Carbonmade

• AI Automatic Video Chapters & Indexing: Wisecut

• AI Ad Script Generator: Script AI

AI social media:

OpenAI MCP

Ghibli film made with AI

𝕏, internet is flooded with Ghibli art created by AI

Anthropic research

Whenever you’re ready, there are 2 ways we can help you:

1. Help you promote your product and service to 150k+ engineers, AI enthusiasts, entrepreneurs, creators, and founders. Sponsor us

2. Help you build a irresistible brand on 𝕏/Twitter and LinkedIn in less than 6 months. We’ve helped 100+ creators from YouTube, entrepreneurs, founders, executives, and people like yourself. Contact us here: [email protected]

Thanks for reading…

That’s a wrap.

What's on your mind?

Share your best ideas with us at [email protected].

We'll bring your ideas to life. Send them our way, and we'll get to work on making them a reality.