This Week’s AI Drops: Slide-Making Bots, 3D Pizza, and a Robot That Does Laundry
Feeling like AI is evolving faster than you can Google “What’s Claude 3 again?” You’re not alone. But here’s the upside: every week brings shiny new tools that can save you time, boost your creative output—or just make you laugh at a robot folding your t-shirt (ok, almost folding).
Grab a drink, crack your knuckles, and let’s run through this week’s wildest updates you can actually use.
Run Local, Run Fast: ChatGLM-4 Enters the Arena
Hosting your own LLM just got a whole lot more real.
ChatGLM-4—often misquoted as “GLM 4.5”—is the newest open-weight model from Zhipu AI. And devs are buzzing for good reason:
- Full-weight download: host it locally or in your own cloud.
- Competitive scores across reasoning and code tasks—right up there with GPT-4o and Claude Opus.
- Speedy inference with free public access (for now) on Zhipu’s site.
But what really makes it pop? Two party tricks:
- Slide-deck generator: Prompt it with “Why pigeons are government drones,” and bam—6 slick slides with visuals and citations.
- One-shot coding: Request a vampire-themed mini-game and get playable JavaScript/Canvas code on the first try.
If you’re eyeing self-hosted LLM solutions for work (or side hustles), this one’s a no-brainer addition to your shortlist.
Video Editing: Write It, Watch It
The bar for video creation just sank faster than your productivity after lunch. Three major tools let you “edit with words”—and the results are uncanny.
Runway’s ALF Edit
Paying Runway users now get ALF (Additive Layer Fill). Want to replace the sky with outer space? Turn your living room into a sea of hoagies? Just type it.
It’s especially sharp at:
- Background or sky swaps
- Single-object replacements
- Keeping everything else untouched
Luma Dream Machine
Similar vibes, with a smart twist: a Strength slider. Dial it down for subtle changes. Crank it for “what if my friends were muppets.” Bonus: audio stays synced, so social clips feel seamless.
Google Veo’s Frame Magic
Veo hides a fun trick—write things like “Dune buggy rolls in, chopper flies above” directly on the first frame… and it plays out exactly like that.
Pro tip: Trim those first few frames post-export to remove the onscreen instructions.
MidJourney Now Does Animation—and Loops
MidJourney just unlocked a new trick: pair a start and end image, and it auto-generates a slick 5-second animation.
Use it to:
- Morph selfies into cyborgs (or seasons, or moods)
- Loop animations perfectly by matching start → end images
- Spice up visual stories with scene-shifting transitions
Heads-up: It still pulls a few surreal stunts—planes doing 180s mid-flight? Not unheard of. But for mood pieces, it slaps.
Memes, Marketing, and You: Face Swap Goes Prime Time
Ideogram’s new Character Mode skips the “20-photo-training-set” hassle. Upload a single selfie, pick a template—say, astronaut or old oil painting—and boom: you’re in.
Even better:
- Paste your face into existing photos
- No editing hoops, just real results
Yes, your Oscars selfie with Jennifer Lawrence can finally be a reality. Sort of.
3D Assets Without the Blender Headaches
If you’ve got a prompt, Meshy 5 has a mesh.
- Text → 3D: “Deep-dish pizza with pepperoni and jalapeños”? Meshy turns it into a textured, game-ready asset—OBJ, GLB, you name it.
- Image → 3D: Upload that cute spaceship doodle from your sketchbook and spin it in full 3D minutes later.
Also worth watching: Tencent’s Hunyuan 3D World generates explorable scenes from text. Still in public beta, but signs point toward something big.
Productivity Upgrades Worth Your Clicks
Some AI features snuck in this week that could seriously shrink your to-do list:
- ChatGPT “Study” Mode: Instead of blurting answers, it walks you through problems step-by-step. Gold for candidates, quizzes, and deep learning.
- Edge Copilot Upgrade: Microsoft’s browser bot can now click, fill forms, and summarize full sessions like a real sidekick.
- Cursor’s Bug Bot: Catch logic lapses and edge-case errors before your team grabs the pitchforks.
Meet the Bots Testing Our Chores
Robotic progress check: we’re not quite in Jetsons territory—but we’re close enough to laugh (or worry).
- Figure 01 from Brett Adcock: This humanoid now moves laundry from hamper to washer. Folding? Coming soon.
- Unitree G1 (not “R1”): A nimble little bot doing cartwheels, handstands, and kung-fu kicks. $9–16k depending on config. Utility: low. Entertainment: high.
Blink and You’ll Miss These
A few quick hits that didn’t get front-page status but pack real value:
- Higgsfield Halo – Flexible video transformer, currently free to try. Guardrails optional.
- Amazon x Fable – 30-minute AI-generated cartoons from one prompt? Netflix’s got competition.
- Google Search Generative rolls into the UK – Now with image understanding baked in.
- Notebook LM – Auto-slide-making and synced audio summaries hit beta.
Try These This Weekend 🔧
No pressure—but if you’re the hands-on type, here’s a punchy test lab for your weekend:
- Sign up at Zhipu.ai and run the slide command in ChatGLM-4.
- Take a short phone video and pit Runway ALF vs. Luma’s Modify slider.
- Pop yourself into a movie poster with Ideogram, then take it up a notch in Photoshop Beta’s upscale.
- Search “gold-plated steampunk pistol” in Meshy 5 and import into a 3D viewer.
- Let Cursor’s Bug Bot roast your code before your teammates do it for real.
Why This Matters (And What Comes Next)
LLMs you can self-host. Slides generated in seconds. 3D models from imagination to render faster than your coffee brews. This isn’t about tech for tech’s sake—it’s about lowering the time from idea to shipped.
You’re living in the era where solo devs build tools that used to take entire teams. And if that sounds overwhelming… don’t worry. We’ve got your back, weekly.
Want an easy place to sharpen your AI skills from scratch?
Check out Tixu—a beginner-friendly AI platform that teaches you how to prompt, code, and build projects without getting lost in the buzzwords.
Ready when you are.
Leave a Reply