Post about everything latest in AI | Founder: AI Toast| DM for Collabs
DM for Collabs
- Tweets 76,960
- Following 687
- Followers 213,601
- Likes 173,087
before → after
same prompt, and both one-shotted.
the one on the right used frontend skills for the design.
6
1
13
3,220
Designers, this one is for you.
Ideogram just launched a new feature to convert images to editable SVG.
Prompt I used:
Design an intricate yet clean vector illustration of a royal Indian peacock with its tail fully expanded into a perfectly symmetrical mandala. Each feather contains geometric floral motifs. Emerald green, sapphire blue, turquoise, and warm gold flat colors. Traditional Indian craftsmanship meets modern luxury packaging. Crisp outlines, separate editable vector elements, cream background, no gradients or photographic textures.
Today, we’re introducing Ideogram's first vector model.
Generate clean, editable SVGs with smooth gradients and sharp typography from a text prompt, or turn any image into a detailed, scalable vector.
Live in Ideogram and the API: ideogram.ai/models/vector
3
1
24
4,053
NVIDIA CEO Jensen Huang asked 4 tiny robots to sit, stand, nap, and do somersaults.
And they actually did it.
NVIDIA is bringing AI agents from virtual world to real world with these robotic ducks.
57
123
873
98,082
Anthropic just launched the OpenAI Dots competition.
Claude Duet is here; it's an AI agent that doesn't just wait for your prompts.
Here’s what it does:
→ Identifies what needs your attention
→ Completes tasks while you're away
→ Works across your connected apps
→ Checks with you before taking important actions
It’s built around Claude Opus 5.5, with memory and skills carried over from your Claude account.
We’re opening a research preview of Claude Duet, a new app where Claude works as a proactive agent.
It comes to you with what needs your attention, completes tasks you ask of it, keeps working when you step away, and checks with you before it acts on your behalf.
9
5
32
5,338
This is crazy...
AI just helped build an AI model from scratch in under 4 months.
Recursive just released Flint, a 40B-parameter coding model built using its automated AI research system, the Eureka Machine.
Here's what makes this interesting:
→ The AI research system helped improve the model's architecture and training.
→ Flint is open-weight under Apache 2.0.
→ It runs locally, so your code can stay on your own hardware.
→ It delivers 125+ tokens/sec on a MacBook at short contexts.
AI is starting to help accelerate the research that builds the next generation of AI.
And this could be just the beginning.
Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine.
• Best-in-class fast local coding
• Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark
• Built from scratch in under 4 months
Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.
6
2
17
2,679
I asked Claude to rank the top 50 AI tools. It put itself at #1. 😂
But that's not all.
- Claude Code ranked #1 for coding.
- Nano Banana Pro took the top spot for image generation.
- Gemini Omni won video.
- ElevenLabs topped voice.
I asked Claude to pick the best tools across 10 categories, from AI assistants and coding to research, automation, and music.
Some rankings make sense. Others are going to start arguments. 👀
How many of these rankings do you disagree with?
Save this list for later.
1
2
14
2,976
🚨 Google just turned Gemini into an AI employee for small businesses.
Whether you’re a freelancer, a founder, or a creator selling your own stuff, this one is for you.
Here’s what’s new:
→ There’s a new business hub inside Gemini where you can handle branding, finances, and customer insights in one place.
→ You upload your logo, fonts, and brand colors once, and Gemini remembers them for every post or product image it makes.
→ You can drop in messy spreadsheets or receipts, and it turns them into simple charts so you can actually understand your numbers.
→ It connects with Google Business Profile, Google Ads, and Workspace, plus apps like Canva, Shopify, Zoho, and Xero.
Google says tens of millions of small business owners already use Gemini. One farmer in Maharashtra even used it to plan his whole ginger farm without any formal training.
It’s rolling out to all Gemini app users over the next few days.
13
1
15
4,224
General Compute just signed a multi-year partnership with @positron_ai to bring Positron’s AI chips into its cloud.
They’re starting with Positron’s current system, called Atlas. Next comes Titan, a bigger system with far more memory. Customers can use it from Q1 2027.
Here’s why this matters.
When an AI model writes an answer, it creates one word at a time. For every word, the chip has to pull data from memory. Most of that time is spent waiting on memory, not doing math.
The special memory GPUs rely on is in short supply right now. Positron’s chips are built around memory and don’t depend on it.
So General Compute is splitting the work. GPUs read your prompt. Positron’s chips write the answer. You get one endpoint and one contract.
Our heterogeneous cloud just got better. We're announcing a cluster scale deployment of @positron_ai as our next partner.
We believe demand for fast inference will outstrip supply for the next decade. Our bet on Positron is another step towards getting fast inference for opensource models into builders' hands.
Hardware goes live Q1 2027.
Thanks to @mitesh711, @trsohmers and the whole Positron team! We're excited for this partnership.
LFG!
3
4
15
3,010
AI made posting easy. But made every brand look the same.
That’s the worst part.
Vyb is AI too, but it starts from your actual brand. You put in your website and it picks up your logo, fonts and colors. Your product photo stays your product photo, it doesn’t redraw it.
Everything sits on its own layer, so you can change just the headline and leave the rest alone. Then export to Figma and keep going.
I launched the first image and video generation models at Google, and kept asking why all AI designs look the same.
AI has brought us the WORST era of design.
It made content creation effortless, but it also made 90% of it look painfully generic.
My co-founder and I spent 2 years building in stealth to change that.
Meet Vyb, the first AI design model that understands your taste like a designer would.
RT + comment your website URL and we’ll make your first design.
4
2
11
3,027
Finally, a robot I actually want in my home.
Tuesday Labs just launched Voxel, a droid butler designed to make living with robots feel natural.
He can:
→ Reset your home by putting things back where they belong
→ Fetch things for you
→ Find misplaced items like your sunglasses
→ Take care of your plants
Voxel is expected to ship in 2028, and the first production run is limited to 999 numbered units.
You can reserve one for $200.
4
2
9
2,483
the Viggle meme app went viral off one piece of what their tech can do: making a character move.
JST-2 is much more of it, 3D characters, motion and scenes from one physics-native world model.
if one piece got that big, I want to see what they build on the rest
Today we are releasing JST-2, the world's first full-stack physics-native world foundation model.
Many world models learn in pixel/video space. We went back to first principles and built the entire model stack in 3D.
Generate a 3D scene in seconds and expand it in real time. Create characters with motion built in, all through one general architecture.
9
2
18
3,609
This is insane.
You can now turn a rough 2D sketch into a full CAD build sequence, and it’s 100x faster than GPT Astra.
You can now design with an AI that fixes the “ugly” side of LLMs.
LLMs are good at general learning, common sense and zero-shot prompts.
They’re bad at hallucinations, simulating the future and memory.
The ugly part:
→ Inference that eats tokens
→ Planning that breaks on long tasks
Skyfall AI built SKY-001, a world model for engineering, to go after exactly that:
→ 100x faster and 150x cheaper than GPT Astra at CAD and BIM
→ Handles designs with thousands of connected parts
→ Sketch or prompt in, full CAD build sequence out
→ Opens in tools like Autodesk Fusion
Introducing SKY-001, the first Engineering World Model.
On CAD and BIM tasks, it beats GPT-6 Astra and every frontier model we tested:
-> 51-125x faster
- > 150-210x cheaper
-> Higher IOU than all frontier models
SKY-001 lets you dream a physical object before you build it.
18 months of work. Built by the team out of Maluuba's deep learning lab (acq. by @MSRResearch).
The first ‘AI General Engineer’ will be the backbone for Autonomous Manufacturing.
6
4
18
3,551
asked Google’s Argon 4, Claude Opus 5.5, GPT 6.1 Sol, and Grok 4.6 to create interactive 3D cats.
the result is insane 🤯
AI has improved so much; I couldn’t pick a winner.
Which one do you think did the best?
7
5
54
9,606
JUST IN: OpenAI finds secret spy device planted by it’s own engineering lead inside it’s office.
The FBI arrested Abdulrahman Alfozan, a 35-year-old Saudi engineer at OpenAI, who now faces upto 5 years of in prison if convicted.
Reportedly, device was hidden behind the TV, inside toilet paper.
We discovered a sophisticated recording device that was hidden inside one of our offices. We used our models to understand the device and where it had come from, identifying that a specific employee had planted it a few days earlier. We referred the matter to the FBI, and we’re grateful to the agents and prosecutors who moved quickly to make an arrest. The FBI’s investigation is ongoing and we will continue to support their work.
At this time, we have no indications of this individual inappropriately accessing or exfiltrating model weights, source code, or customer data. The employee was terminated.
We take protecting our technology, data, and people seriously, and as a matter of national importance. Our security program is designed to protect against a wide range of threat actors, from insider risks to cybercriminals to nation states, and this case is an example of how AI can help address these sophisticated security threats.
3
2
16
5,907
🚨Google is officially rolling out Argon 4
I tried building a dancing doll with it. It absolutely nailed it.
→ Change the outfit color
→ Pick skin and hair
→ Switch dance moves
→ Adjust tempo, bounce and jiggle
8
1
30
6,093
Claude Opus 5.5 vs GPT-6.1 Sol
This time, I prefer GPT-6.1 Sol - Super Mario style game, but the player is the Claude logo and the coins are tokens.
Wait till the end. :)
25
8
97
16,971
ChatGPT Shortcuts 🔥
9 camera angles. 1 photo. 0 photographers. (save these)
→ /groundlevel — eye-level, straight-on shot
→ /droneview — top-down, wide aerial look
→ /lowangle — shot from below, looks powerful
→ /topview — directly overhead, flat-lay style
→ /closeup — tight portrait, face-focused
→ /birdseye — high angle, slight tilt down
→ /wormsview — extreme low angle, looking up
→ /pullout — wider shot, more surroundings visible
→ /sideview — profile shot, side-on angle
one photo, nine completely different shots. no reshoot, no gear, no photographer.
4
4
38
6,442
Step 5 Preview is now #1 on OpenRouter trending, now available in Cline, Kilo, OpenCode and OpenRouter:
openrouter.ai/rankings?view=…
4
487
→ Step 5 Preview made a clean, dark slide deck. It covers growth by business unit, margins, regional performance, risks and what to watch in FY2026.
→ Qwen made denser, report-style pages. It has more charts and a sources and methodology page at the end.
1
4
532
After that, I gave both models a chance to fix anything that was wrong.
I changed/fix a number in the file and checked whether the related calculations and charts updated correctly.
Then I compared the final results side by side.
1
3
111
Then I checked the first delivery.
I opened the actual files and checked:
• Were the numbers and sources correct?
• Did the charts match the data?
This is where the differences started showing up.
1
3
171
The task was more than just summarizing a report.
They had to:
→ explain what changed in the business
→ find key problems
→ support the findings with sources
→ make an 8–12 page presentation
And most importantly, the numbers had to match across everything.
1
4
392
I gave @StepFun_ai Preview and Qwen 3.8 Max the exact same task.
Analyze Apple's FY2025 and build the business analysis deck.
Here’s how the two models handled it 👇
22
3
39
6,350
Codex for creative work is here.
That’s how Moda describes Voyager, an open harness for video, graphics and games.
Most AI tools are built for coding. Voyager is designed to get better creative results from models like Opus.
What it does:
→ Comes with free graphics, music and video editing tools
→ Can also drive Blender, DaVinci Resolve, After Effects, Ableton and 100+ other tools
→ Moda says everything in its launch thread was made with Voyager
Introducing Voyager: the open harness for S-tier video, graphics and games.
The Codex for creative work.
Most AI tools are built for coding. Voyager is designed to get the best creative results from models like Opus, Astra and DeepSeek.
It comes with free graphics, music and video editing tools - or, it can drive Blender, Resolve, After Effects, Ableton and 100+ others.
Everything in this thread was made with Voyager. A few of our favorites:
1
15
3,485
The perfect prompt doesn't exist.
The right direction does.
Most AI tools expect you to know exactly what you want.
But creative work often starts with a brief, a reference, or just a hunch.
That’s why we built Burst mode.
Turn the start of an idea into up to six image directions at once, up to 10× faster, for just one AI credit.
More directions to explore. More chances to find the one worth pursuing. elements.envato.com/lp/burst…
2
1
17
4,487
A Chinese developer built an open-source tool that makes YouTube Shorts from a single keyword.
It's called MoneyPrinterTurbo, and it has passed 125K stars on GitHub.
You type a topic, and it puts the whole video together: script, voiceover, subtitles, footage and the final edit.
The same video works for TikTok and Reels.
1
12
84
9,246
Norway becomes the first country to propose a temporary ban on AI glasses in public places
2
2
27
5,873
Hollywood is in trouble.
You can now fix a performance after the scene is shot, without reshooting anything.
That's what DRAMA does.
→ Fix the acting
→ Keep the shot
→ Keep the face
→ Keep the scene
Dakota Johnson might finally have a use case for AI 👀
Faking just got better.
Acting is solved! Just shoot the video, add performance later.
Officially launching DRAMA 1.0: Our cinematic expression-editing model.
Change the emotion. Dial up the intensity and make performances more convincing. No shoot goes to waste now!!
DRAMA edits acting performance in existing footage while preserving identity, facial features, lighting, and camera angles. 10 emotions accessible at one click.
Rather than regenerating the scene, DRAMA targets only the expression layer.
It is designed to minimize visual drift and temporal artifacts while retaining the original shot's cinematic quality. Our proprietary model disentangles performance from identity and pose, thereby achieving extremely high fidelity compared to state of the art text to video models.
From a subtle reaction to a scene-stealing performance. You control how far it goes.
Today DRAMA comes out of Beta. Creators have been using Drama for all sorts of cases and are excited to share the public version today.
Try DRAMA → trydrama.io
@trydrama_io @mandar2404 @deepakpathak @onzonebanerjee @Shaileshg21 @RMantri @mohitrakhade20 @agrahari18 @ExfinityVP @navamcapital
8
17
73
12,865
Holy shit… Google just made a 740M-parameter model that understands text, images, video AND audio.
EmbeddingGemma 2 is an open multimodal embedding model built for on-device use.
It can:
→ Put text, code, images, video and audio into one shared search space
→ Run on phones and laptops, with vision and audio parts you load only when needed
→ Understand 100+ languages
→ Shrink vector storage with minimal quality loss
→ Power offline, privacy-first RAG when paired with Gemma 4
Introducing EmbeddingGemma 2, a new open multimodal model that sets the standard for on-device efficiency.
- our first open, natively multimodal embedding model
- handles text, code, image, video, and audio tasks within a lightweight, modular 740M parameter form factor
- ideal for offline, privacy-first RAG when paired with Gemma 4
- outperforms some specialist models more than twice its size
Weights available now on Hugging Face.
3
2
35
6,908
3. Build the interactive experience
Give Astra the brief and the creatures, then ask for:
> clickable animal cards
> a slow 360° rotation
> an info panel that swaps per animal (name, tags, key traits, habitat)
> one caption line under each creature
> smooth transitions between cards
Then screen record at 16:9 and time the cursor clicks to the panel changes.
1
503
2. Generate each creature as a hero object
Ask Astra for one animal at a time, in a clean front or three-quarter view.
Keep the same lighting, background and detail level for every creature so the set feels like one collection.
1
273
1. Design the interface first
Describe the app before any animals: cream background, sage-green accents, serif headings, a card row on top, an info panel on the right, and a big rotating creature in the center.
The clearer the layout, the fewer fixes later.
1
518
Built an interactive 3D Animal Atlas with GPT-6 Astra
No 3D software or extra plugins.
Here's the full process 👇
12
7
76
8,464
Vibe coding is a QA nightmare, and the tool for it is finally here.
It's called Ship, from ContextQA. It acts like an AI QA engineer for your app.
Here's what it does:
→ Tests your PR on the deployed app
→ Picks up bugs reported in Slack
→ Captures the evidence of what broke
→ Passes the context to Claude or Codex to fix it
Everyone got a coding agent. Nobody got a QA engineer.
Until today.
Meet Ship, your Autonomous Quality Engineer.
It tests deployed PRs, reproduces bugs from Slack and Linear, and hands Claude or Codex the context to fix them.
Try it free: ship.contextqa.com
7
21
5,754
Your brain gets SHARPER when you exercise it daily!
AI makes practicing easier.
STEAL These 7 Claude Prompts to Learn Anything 10x Faster. Use them in one chat, in order:
8
3
30
6,363
There are 7 skills worth installing in Claude right now.
Skills are basically shortcuts. You type one short command, and Claude runs a whole workflow for you. 👇
→ caveman: makes Claude talk in short, blunt lines so you spend fewer tokens
→ humanizer: cleans up the stuff that makes writing sound like AI
→ social-media-skills: helps you with hooks, posts and carousels
→ claude-ads: checks your Meta, Google and TikTok ads for mistakes
→ claude-seo: SEO that's built for AI search, not just Google
→ ai-second-brain: turns your old Claude and ChatGPT chats into a personal wiki
→ frontend-design: stops Claude from giving you the same boring UI every time
Save this one for later.
8
8
48
7,519
🚨 You need to see this.
Google just launched Advent of Agents Season 3.
Now you can master shipping production-ready agents in 31 days for free.
Here is everything you need to know:
→ It runs daily through October, one real production question per day
→ Every day comes with a short practitioner video and code you can run
→ Each day stands on its own, so you can start on any day
What the 31 days cover:
→ Days 1–4: Foundations. The agent lifecycle, and why securing agents isn't like securing chatbots or APIs
→ Days 5–9: Identity and access. A cryptographic identity for every agent, acting on behalf of users, least-privilege tool access
→ Days 10–16: Runtime guardrails. Blocking prompt injection, stopping PII leaks, sandboxing agent-generated code, verifying third-party MCP servers
→ Days 17–21: Fleet management. Terraform, CI/CD, and one agent registry across projects and clouds
→ Days 22–27: Observability and containment. Tracing agent behavior, threat detection, and an enterprise kill switch
→ Days 28–31: Day 2 operations. Cost circuit breakers, model routing, and continuous evaluation
If you've been meaning to actually understand how agents work, not just use them, this is a straightforward way in.
16
4
33
5,184
Someone just put 389 viral Opus 5.5 videos in one place.
And you can copy every single prompt.
What's inside:
→ Motion graphics
→ Explainer videos
→ 3D scenes
→ Even games
You can watch the original next to the remake, then copy the prompt behind it.
Save this for later.. 🔗👇
9
10
109
11,262
Influencers are in trouble..
Vivix A1 just dropped, You can now interrupt an AI character mid-sentence.
They can walk around, dance, pick things up, and keep listening the whole time.
So you can talk over them or change your mind halfway and they adjust instead of restarting.
Create your own character. Call the shots. Live.
Introducing A1 Playground + Agent: build a live AI character that talks, moves, and plays with you.
Create yours: platform.vivix.ai/vivix-a1-m…
Their look. Their personality. Their voice. Your choice. Describe who you have in mind, upload a picture, and let our Agent bring them to life.
And they do a lot more than talk.
💃 Full-body performance. From air guitar to victory dances.
🚶 Move around the scene. Walk and talk. Pace while making a point.
🛍️ Interact with objects. Pick up products, show them off, and pitch them.
🎮 Play web games together. Deal cards, keep score, and play along.
Solve a mystery with Sherlock Holmes, then interrupt him for a dance break.
Take a walk with Socrates. Hand him a protein shake and debate the meaning of gains.
Give Snape a bottle of shampoo. Ask for the most enthusiastic sales pitch of his life.
Who will you bring to life?
8
2
16
5,822