bt_bb_section_bottom_section_coverage_image

Weekly AI Puls – Big Breakthroughs, Small Bugs, and a Lot of Fun

Holistically pontificate installed base portals after maintainable products. Phosfluorescently engage worldwide methodologies with technology. Seamlessly visualize quality intellectual capital without superior collaboration and idea-sharing. Holistically pontificate installed base portals after maintainable products.

Weekly AI Puls – Big Breakthroughs, Small Bugs, and a Lot of Fun

1. ChatGPT Agent Mode: Your Autonomous Sidekick

OpenAI launched a powerful new Agent Mode (currently for Pro users at $200/month) that lets ChatGPT operate a virtual browser, filter results, and handle multi-step tasks—like planning date nights, ordering clothes, or analyzing YouTube performance. In testing, simpler tasks (like a YouTube report or presentation) were completed fairly well. More complex workflows? They often stalled or needed manual intervention. Impressive in concept, buggy in execution. Promising, but not quite there yet. #chatgpt #openai #AgentMode

2. Record Mode Now for Plus Users Too

OpenAI’s Record Mode, which records and summarizes any audio on your Mac, is now available to Plus users. Perfect for Zoom calls, YouTube videos, or podcasts when you don’t feel like taking notes. #RecordMode #OpeanAI #ChatGpt

3. Anthropic Launches Claude’s Tool Directory

Claude now has a growing library of connectors—from Notion and Canva to Stripe and Figma—designed to let you interact with various services through Claude. The interface is promising, but right now, connections are flaky and some integrations don’t yet work. Still, a solid glimpse at the future of AI operating hubs. #Anthropic #Claude

4. NVIDIA AI Twins: Digital Avatars of You

NVIDIA released “AI Twins”—a feature that creates a digital clone of you from a short video recording. Your avatar can then promote products or deliver messages using your voice and likeness. It’s fun, slick, and free to try at invido.io. #Nvidia #AITwist

5. Hume AI: Clone Not Just Your Voice, But Your Vibe

Hume AI’s EVI3 model doesn’t just replicate your voice—it mimics your speech style and personality. The AI version of Matt Wolf talks like him… a little too much. Not an exact voice match (yet), but definitely a step toward emotionally expressive AI personas. #EVI3 #Hume

6. Runway Act 2: You Move, It Animates

Runway’s new model turns your webcam gestures into animations. Upload a video, pick a character, and it mimics your motion. It’s great at facial and upper-body movement but struggles with props (like lightsabers). Still, an exciting toy for creators.

7. Deart Mirage LSD: Real-Time Video Reskin

Live Stream Diffusion (yes, “LSD” is probably intentional) reskins your webcam feed live—turning your background into Versailles, Minecraft, Goo World, or Yarnland. Lag is minimal and the creativity is off the charts. A psychedelic playground for livestreamers and digital artists.

8. Adobe Firefly: Voice to Sound Effects

Firefly now turns voice clips into realistic sound effects. Say “raaar!” and get back multiple lion roars timed to your original voice. Perfect for syncing effects in videos or animations. Simple, but clever. #Adobe #AdobeFirely

9. XAI Grok: Meet Annie and Rudy

Elon Musk’s Grok app introduced AI companions Annie and Rudy—complete with an NSFW mode (for Annie). On launch day, the servers were overwhelmed, and no one could get them to talk. That’s… probably for the best. #GROK

10. IDE Drama: Windsurf, DeepMind & Cognition

Windsurf—an AI-enhanced IDE forked from VS Code—saw its execs jump ship to Google DeepMind just as the company was acquired by Cognition (makers of AI coder Devon). All this while Microsoft still holds IP ties through VS Code. A full-on soap opera in the dev tools world. #

🔥 Quick Hits

  • Google’s AI Calling Feature now makes business calls on your behalf and checks availability & pricing.

  • Kimmy K2, a new Chinese open-source model, landed in the top 5 on LM Arena—rivaling DeepSeek R1.

  • Claude for Finance is now available—tailored for deep financial analysis tasks.

  • Mistral adds Deep Research + Voxrol, a speech-to-text model that’s ultra-cheap ($0.001/minute).

  • Amazon Kira IDE arrives with a planning-first approach to software development—ideate before you code. #

🔎 Final Thoughts

AI is clearly racing toward a fully hands-off future—handling bookings, shopping, analytics, and even cloning your personality. But we’re still in the early stages. Agents are promising but needy. Avatars talk too much. Tools are smart but buggy. Still, this week proves one thing: AI is evolving faster than ever—and getting weirder in the best possible ways. Just be careful what you let it order for you. #