Post about everything latest in AI | Founder: AI Toast| DM for Collabs
DM for Collabs
Joined September 2021
- Tweets76.9K
- Following686
- Followers213K
- Likes173K
Pinned Tweet
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.
Poonam Soni retweeted
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.
🤝 Paid partnership
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
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.
🤝 Paid partnership
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
SenseTime just open-sourced SenseNova-RoboRSI.
The embodied AI release includes:
→ A unified evaluation harness
→ Dockerized benchmarks
→ A common HTTP API
→ A 27B embodied model that runs at roughly 1/50 the task cost of GPT-6-Astra
Technical report:
arxiv.org/abs/2610.22471
🤝 Paid partnership
We release the most comprehensive 𝗲𝗺𝗽𝗶𝗿𝗶𝗰𝗮𝗹 𝘀𝘁𝘂𝗱𝘆 𝗼𝗳 𝗚𝗣𝗧-𝟲-𝗔𝘀𝘁𝗿𝗮’𝘀 𝗲𝗺𝗯𝗼𝗱𝗶𝗲𝗱 𝗶𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲 across 12 simulation benchmarks and 8 real-robot platforms, together with 𝗦𝗲𝗻𝘀𝗲𝗡𝗼𝘃𝗮-𝗥𝗼𝗯𝗼𝗥𝗦𝗜: an open-source harness framework and a 27B model that runs at ~1/50 the cost of GPT-6-Astra.
Here's what we found:
🔹 𝗧𝗵𝗲 𝗰𝗮𝗽𝗮𝗯𝗶𝗹𝗶𝘁𝘆 𝗰𝗲𝗶𝗹𝗶𝗻𝗴 𝗶𝘀 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝗺𝘂𝗰𝗵 𝗵𝗶𝗴𝗵𝗲𝗿. Under a unified protocol (ICL + up to 10× the interaction budget), GPT-6-Astra achieves clear SOTA across all 12 benchmarks: RoboDojo: 27.6% at 1× → 81.9% at 10×, DexJoCo: 87.3% vs. 40.3% prior SOTA, while the LIBERO benchmarks are pushed to saturation.
🔹 𝗧𝗼𝗼 𝗺𝘂𝗰𝗵 𝗯𝘂𝗰𝗸 𝗳𝗼𝗿 𝘁𝗵𝗲 𝗯𝗮𝗻𝗴. Even a simple task like placing pens in a holder costs ~$55 in inference.
🔹 𝗧𝗵𝗲 𝘄𝗼𝗿𝗹𝗱 𝗱𝗼𝗲𝘀𝗻’𝘁 𝘄𝗮𝗶𝘁. Time matters as real robots overheat, and an egg burned while the model was still thinking.
Our takeaway: GPT-6-Astra has largely crossed the threshold of 𝗲𝗺𝗯𝗼𝗱𝗶𝗲𝗱 𝗰𝗼𝗺𝗽𝗲𝘁𝗲𝗻𝗰𝗲, but 𝗽𝗿𝗮𝗰𝘁𝗶𝗰𝗮𝗹 𝗲𝗺𝗯𝗼𝗱𝗶𝗲𝗱 𝗶𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲 is not here yet.
🌐 Project page: sensenova.ai/has-gpt-6-astra…
📄 Report: sensenova.ai/gpt-6/assets/do…
💻 Code & model: github.com/OpenSenseNova/Sen…
Poonam Soni retweeted
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.
Poonam Soni retweeted
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.
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.
Only 1.3B parameters… and THREE benchmark wins 🤯
#1 on olmOCR-Bench, OmniDocBench and DocParsingBench
Fastest open-source parser in the test ⚡
100 pages, 101 seconds, one RTX PRO 6000
15.43× Chandra-OCR-2's throughput
Handwritten notes, complex tables, even chemical structures
Open source, replaceable layout model
That's OXR. Which document would you throw at it first?
github.com/SoMarkAI/OXR
🤝 Paid partnership
Poonam Soni retweeted
🚨 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.
Poonam Soni retweeted
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.
🤝 Paid partnership
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.
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.
🤝 Paid partnership
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.
Poonam Soni retweeted
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.
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.
What I like about how Voice Arena ranks these: native speakers, blind A/B pairs, no self-reported numbers. Voxtral from Mistral comes in at #6 of 20 in Hindi.
🤝 Paid partnership
Mistral AI's Voxtral-mini-tts-3-0 debuts at #6 of 20 on the Voice Arena Hindi 🇮🇳 TTS Leaderboard at 1048 Elo, ahead of Maya Research's Maya-2-Native, ElevenLabs' Eleven v3 and Cartesia's Sonic-3.5.
Voxtral-mini-tts-3-0 is the latest TTS model from @MistralAI , part of the Voxtral 3 release, with support for 35+ languages and 50+ dialects. Every ranking on Voice Arena comes from blind, head-to-head votes by vetted native speakers, scored with Bradley-Terry Elo.
Only Google DeepMind, ElevenLabs and Cartesia place models above it, which makes it the highest-ranked model on the Hindi board from any other lab, in a tight group where Maya-2-Native and Eleven v3 sit within 4 points.
Results backed by 600+ blind, head-to-head listener votes from native Hindi speakers.
Voxtral-mini-tts-3-0 is also being evaluated on our US English board, with results coming soon.
Congratulations to @MistralAI on the release!
See below for the exact audio raters judged and unedited listener verdicts 🧵
🚨 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.
Poonam Soni retweeted
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.
🤝 Paid partnership
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!