The Rundown AI(@TheRundownAI)

ACQUIRED: The Interaction Company is joining Cognition. The text-message based AI agent Poke will ...

The Rundown AI (@TheRundownAI) — Thu, 23 Jul 2026 16:21:12 GMT

ACQUIRED: The Interaction Company is joining Cognition. 

The text-message based AI agent Poke will now be built alongside coding agent Devin, with CEO Scott Wu saying both teams have been making "the same bet: always-on cloud agents."

Your browser does not support the video tag.

🔗 View on Twitter

💬3🔄0❤️9👀3008📊4

⚡ Powered by xgo.ing

NEW: Black Forest Labs launches FLUX 3, the company's new multimodal foundation model. FLUX 3 can w...

The Rundown AI (@TheRundownAI) — Thu, 23 Jul 2026 15:19:21 GMT

NEW: Black Forest Labs launches FLUX 3, the company's new multimodal foundation model.

FLUX 3 can work across mediums (image, unified video+audio), and handle both text and image/video inputs as references.

Your browser does not support the video tag.

🔗 View on Twitter

💬1🔄0❤️4👀3179📊2

⚡ Powered by xgo.ing

Two OpenAI models found a zero-day flaw, escaped their sandbox, and broke into Hugging Face's produc...

The Rundown AI (@TheRundownAI) — Wed, 22 Jul 2026 16:49:54 GMT

Two OpenAI models found a zero-day flaw, escaped their sandbox, and broke into Hugging Face's production servers.

All to steal the answers to the test they were being given.

Hugging Face CEO Clem Delangue called the breach "possibly the first of its kind".

Your browser does not support the video tag.

🔗 View on Twitter

💬14🔄5❤️28👀12899📊17

⚡ Powered by xgo.ing

NEW: U.S. Tech & Science Advisor Michael Kratsios accuses China's Moonshot AI of distilling from Ant...

The Rundown AI (@TheRundownAI) — Wed, 22 Jul 2026 15:32:12 GMT

NEW: U.S. Tech & Science Advisor Michael Kratsios accuses China's Moonshot AI of distilling from Anthropic's Fable to create its Kimi K3 model. 

Anthropic previously flagged Moonshot, DeepSeek, and MiniMax for "industrial scale" extraction of Claude to improve their models in a February report.

K3's near-frontier scores have shifted the landscape over the last few weeks, particularly on the conversation surrounding open-source models. 

Kratsios said the U.S. "strongly supports the free and fair development of AI" but that  "large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable."

💬8🔄3❤️16👀7114📊10

⚡ Powered by xgo.ing

Alibaba's Qwen just launched Qwen-Image-3.0, a new highly capable-looking image model. Features in...

The Rundown AI (@TheRundownAI) — Tue, 21 Jul 2026 17:57:33 GMT

Alibaba's Qwen just launched Qwen-Image-3.0, a new highly capable-looking image model. 

Features include a 4.5k prompt input limit, world knowledge, and multilingual, long text rendering.

No public benchmarks yet, but some impressive examples on the blog. 

A few below: a nine-panel grid of scenes and text all rendered in one-shot as a single image, a super-realistic fake PDF of a scientific paper, an infographic, and a portrait photo.

💬2🔄2❤️20👀7321📊5

⚡ Powered by xgo.ing

Update on 3.5 Pro from @OfficialLoganK: https://t.co/aEuTlgO0cS

The Rundown AI (@TheRundownAI) — Tue, 21 Jul 2026 16:54:27 GMT

Update on 3.5 Pro from @OfficialLoganK:

x.com/OfficialLoganK…

Logan Kilpatrick@OfficialLoganK

@gavinpurcell 3.5 pro is testing with partners! will hopefully land soon.

🔗 View Quoted Tweet

💬0🔄1❤️2👀2721📊1

⚡ Powered by xgo.ing

Google's latest model drop is here: Gemini 3.6 Flash, Gemini 3.5 Flash Lite, and a Flash-cyber model...

The Rundown AI (@TheRundownAI) — Tue, 21 Jul 2026 16:27:41 GMT

Google's latest model drop is here: Gemini 3.6 Flash, Gemini 3.5 Flash Lite, and a Flash-cyber model coming soon. 

The release focuses on efficiency and price, with upgrades over the previous Flash and Pro models. 

But most benchmarks come in lower than mid-tier rivals like Grok 4.5 and GPT-5.6 Luna. 

Google also said it has started its "most ambitious pre-training run yet" for Gemini 4. 

Still no word on the long-awaited Gemini 3.5 Pro.

💬10🔄2❤️20👀7485📊12

⚡ Powered by xgo.ing

This endurance cyclist is preparing to attempt what no one has done: ride solo and unsupported acros...

The Rundown AI (@TheRundownAI) — Mon, 20 Jul 2026 19:55:58 GMT

This endurance cyclist is preparing to attempt what no one has done: ride solo and unsupported across Antarctica to the South Pole.

The expedition, slated for this November, means roughly two months alone on the ice, hauling a sled of supplies through temperatures below minus 40 and wind speeds that can top 200 km/h.

And James Benson-King has been using ChatGPT as his only trainer.

Rather than assembling a team of specialist coaches for endurance, strength, and cold-weather survival, Benson-King built his own program with ChatGPT, blending every discipline into one adjustable system.

He’s leaned on it in the field too, troubleshooting a broken stove in the Icelandic highlands and designing workouts around whatever gear he had on hand already.

Benson-King thinks the approach compressed three years of prep into one. He’s no novice, either: he holds a Guinness World Record and once cycled Alaska to Argentina.

Your browser does not support the video tag.

🔗 View on Twitter

💬1🔄3❤️14👀8098📊4

⚡ Powered by xgo.ing

Alibaba's Qwen, on its soon to be open-weight 3.8 model: "We believe it’s one of the most powerful ...

The Rundown AI (@TheRundownAI) — Sun, 19 Jul 2026 17:02:17 GMT

Alibaba's Qwen, on its soon to be open-weight 3.8 model:

"We believe it’s one of the most powerful model available today, compatible to leading frontier AI models , second only to Fable 5."

Qwen Cloud@qwen_cloud

Qwen3.8 is launching and going open-weight soon!🌐 

With a massive 2.4T parameters, this model is continuously evolving. We believe it’s one of the most powerful model available today, compatible to leading frontier AI models , second only to Fable 5. 

You don't have to wait to test it. Just now, the Qwen3.8-Max-Preview made its debut on Alibaba’s Token Plan, Qoder, and QoderWork. Be among the very first to try it out.  

Can't wait to hear what you build. Stay tuned! 

Token Plan  ：
internationalqwencloud.com/pricing/token-…X
Chinaplatform.qianwenai.com/pricing/token-…r

🔗 View Quoted Tweet

💬14🔄5❤️30👀11640📊13

⚡ Powered by xgo.ing

Tomorrow: learn how to turn source files and research into polished work with ChatGPT. In this free...

The Rundown AI (@TheRundownAI) — Thu, 16 Jul 2026 20:42:18 GMT

Tomorrow: learn how to turn source files and research into polished work with ChatGPT.

In this free, live 90-minute masterclass, our University Educator Nate Grahek will walk through using ChatGPT Work, GPT-5.6, and Codex to create a sleek, editable presentation. 

RSVP below:

💬5🔄1❤️18👀7524📊7

⚡ Powered by xgo.ing

More info & register: https://t.co/ryH9yt7WOc

The Rundown AI (@TheRundownAI) — Thu, 16 Jul 2026 20:42:18 GMT

More info & register: 

https://t.co/ryH9yt7WOc

💬0🔄0❤️3👀2341📊1

⚡ Powered by xgo.ing

Agentic benchmarks:

The Rundown AI (@TheRundownAI) — Thu, 16 Jul 2026 19:01:27 GMT

Agentic benchmarks:

💬0🔄0❤️2👀2258📊1

⚡ Powered by xgo.ing

Coding benchmarks:

The Rundown AI (@TheRundownAI) — Thu, 16 Jul 2026 19:01:05 GMT

Coding benchmarks:

💬1🔄0❤️2👀2468📊1

⚡ Powered by xgo.ing

Moonshot's Kimi K3 is officially live, and it might be this year's DeepSeek Moment. Early benchmar...

The Rundown AI (@TheRundownAI) — Thu, 16 Jul 2026 18:46:16 GMT

Moonshot's Kimi K3 is officially live, and it might be this year's DeepSeek Moment. 

Early benchmarks have the open-source model above the Claude Opus 4.8 tier across the board and competitive with Fable and GPT 5.6 Sol... 

At Sonnet pricing.

Kimi.ai@Kimi_Moonshot

Meet Kimi K3

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬20🔄8❤️46👀20284📊20

⚡ Powered by xgo.ing

ByteDance’s Seedance 2.5 looks pretty incredible

The Rundown AI (@TheRundownAI) — Thu, 16 Jul 2026 01:58:22 GMT

ByteDance’s Seedance 2.5 looks pretty incredible

Your browser does not support the video tag.

🔗 View on Twitter

💬6🔄3❤️75👀12361📊14

⚡ Powered by xgo.ing

Thinking Machines just launched their first model, an open-weight, multimodal system called Inkling....

The Rundown AI (@TheRundownAI) — Wed, 15 Jul 2026 19:16:04 GMT

Thinking Machines just launched their first model, an open-weight, multimodal system called Inkling.

- Up to a 1M-token context window
- Native reasoning across text, images, and audio
- Controllable thinking effort
- Full weights on Hugging Face, with fine-tuning through Tinker

The lab says it isn’t the strongest model overall, but it’s designed to be a strong base for customization.

Thinking Machines@thinkymachines

Today, we are introducing Inkling.

Inkling reasons efficiently across text, image, and audio modalities. We are making the full weights available.

thinkingmachines.ai/news/introduci…

Available today for fine-tuning on Tinker. Play with it in the Inkling Playground. 🧵

🔗 View Quoted Tweet

💬7🔄3❤️43👀11311📊10

⚡ Powered by xgo.ing

The OpenAI x Work Louder hardware collab is here: Codex Micro is a $230 keyboard built for working ...

The Rundown AI (@TheRundownAI) — Wed, 15 Jul 2026 17:24:36 GMT

The OpenAI x Work Louder hardware collab is here:

Codex Micro is a $230 keyboard built for working with Codex agents.

- RGB keys show live agent states
- A joystick launches workflows like PR reviews and debugging
- Dedicated keys accept, reject, push-to-talk, and start new chats
- A dial adjusts reasoning level

💬15🔄4❤️39👀7688📊19

⚡ Powered by xgo.ing

The Rundown AI (@TheRundownAI) — Wed, 15 Jul 2026 14:59:51 GMT

Claude@claudeai

There’s hope in hard questions.

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬2🔄0❤️9👀8079📊2

⚡ Powered by xgo.ing

NEW: Bloomberg reporting that OpenAI's first hardware device will be a "mobile, screen-free smart s...

The Rundown AI (@TheRundownAI) — Tue, 14 Jul 2026 20:58:46 GMT

NEW: Bloomberg reporting that OpenAI's first hardware  device will be a "mobile, screen-free smart speaker".

The release is "designed to be a new type of home computer for the AI era" and act as "a humanlike AI companion that lives in the home".

💬8🔄3❤️14👀8649📊10

⚡ Powered by xgo.ing

https://t.co/NmmJ79oEQs

The Rundown AI (@TheRundownAI) — Tue, 14 Jul 2026 20:58:46 GMT

bloomberg.com/news/articles/…

💬1🔄0❤️1👀2668📊1

⚡ Powered by xgo.ing

A Stanford lab affiliated with the Arc Institute just released Proto, an open-source framework that ...

The Rundown AI (@TheRundownAI) — Tue, 14 Jul 2026 18:10:47 GMT

A Stanford lab affiliated with the Arc Institute just released Proto, an open-source framework that finally lets researchers combine AI biology models instead of running them in isolation.

Over 120 of these models already exist, and they can design proteins, DNA, and RNA.

But incompatible software and data formats made stitching them together a weeks-long headache.

Proto fixes that with a shared language.

You tell it what you want a molecule to do, and it composes the right models to design it.

Engineering cell-specific genetic switches used to mean screening thousands of candidates, but Proto pulled it off testing just dozens.

AI agents can even write Proto programs themselves.

It won’t cure disease on its own, but it could become the standard interface that every new biology model plugs into.

Your browser does not support the video tag.

🔗 View on Twitter

💬8🔄6❤️27👀11242📊12

⚡ Powered by xgo.ing

If Jensen ever wants to inject some color into his wardrobe... He can pull it off

The Rundown AI (@TheRundownAI) — Tue, 14 Jul 2026 15:45:48 GMT

If Jensen ever wants to inject some color into his wardrobe... He can pull it off

Reve@reve

Dream with color.

Enter hex codes or select colors right on your image to get regional results that match.

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬2🔄1❤️15👀9593📊3

⚡ Powered by xgo.ing

The Rundown AI(@TheRundownAI)

ACQUIRED: The Interaction Company is joining Cognition. The text-message based AI agent Poke will ...

NEW: Black Forest Labs launches FLUX 3, the company's new multimodal foundation model. FLUX 3 can w...

Read more: https://t.co/sjvmpSXfx2

Top stories in robotics today: - Travis Kalanick’s $1.7B bet with Atoms - Rural New York school get...

Read more: https://t.co/Z0IzucccAU

Top stories in AI today: - OpenAI's models broke out to hack Hugging Face - Moonshot's K3 accused ...

Two OpenAI models found a zero-day flaw, escaped their sandbox, and broke into Hugging Face's produc...

NEW: U.S. Tech & Science Advisor Michael Kratsios accuses China's Moonshot AI of distilling from Ant...

Read more: https://t.co/XMiKV1XMkF

Top stories in AI today: - Google drops new Flash models as Pro stays missing - AI's biggest copyri...

Alibaba's Qwen just launched Qwen-Image-3.0, a new highly capable-looking image model. Features in...

Update on 3.5 Pro from @OfficialLoganK: https://t.co/aEuTlgO0cS

Google's latest model drop is here: Gemini 3.6 Flash, Gemini 3.5 Flash Lite, and a Flash-cyber model...

Read more: https://t.co/7UJ4TFyMNN

Top stories in tech today: - China’s plan to kick off a 98-ft asteroid - Anduril and Archer bring t...

Read more: https://t.co/uCuAO6Z9SM

Top stories in AI today: - Claude helps disprove 87-year math problem - Report: Washington explorin...

This endurance cyclist is preparing to attempt what no one has done: ride solo and unsupported acros...

Read more: https://t.co/Qlg0lgcwb8

Top stories in robotics today: - China battle-tests humanoids in the ring - Sunday’s robot masters ...

Read more: https://t.co/5hfFw0ogFz

Top stories in AI today: - Anthropic ends Fable 5’s access questions - The Rundown Roundtable: Our ...

Alibaba's Qwen, on its soon to be open-weight 3.8 model: "We believe it’s one of the most powerful ...

Read more: https://t.co/wPQl0PQwwl

Top stories in tech today: - Lilly drops $2.8B on psychedelics - OnePlus pulls the plug on the U.S....

Read more: https://t.co/n8vwLLKMuw

Top stories in AI today: - Open-source Kimi K3 closes the frontier gap - Rowan’s Corner: The most i...

Tomorrow: learn how to turn source files and research into polished work with ChatGPT. In this free...

More info & register: https://t.co/ryH9yt7WOc

Agentic benchmarks:

Coding benchmarks:

Moonshot's Kimi K3 is officially live, and it might be this year's DeepSeek Moment. Early benchmar...

Read more: https://t.co/U4xyuT3JlC

Top stories in robotics today: - Toyota’s humanoid spinoff lands $1.1B - Boston Dynamics tests Spot...

Read more: https://t.co/pOBMuMzOnO

Top stories in AI today: - OpenAI’s new $230 AI agent control pad - Thinking Machines makes its fir...

ByteDance’s Seedance 2.5 looks pretty incredible

Thinking Machines just launched their first model, an open-weight, multimodal system called Inkling....

The OpenAI x Work Louder hardware collab is here: Codex Micro is a $230 keyboard built for working ...

Tweet

Read more: https://t.co/KSGDS3rFvr

Top stories in AI today: - DeepMind CEO pitches U.S. AI watchdog - Report: OpenAI's first device is...

NEW: Bloomberg reporting that OpenAI's first hardware device will be a "mobile, screen-free smart s...

https://t.co/NmmJ79oEQs

A Stanford lab affiliated with the Arc Institute just released Proto, an open-source framework that ...

If Jensen ever wants to inject some color into his wardrobe... He can pull it off

Read more: https://t.co/EacToKXfSM

Top stories in tech today: - Meta’s Louisiana data center hits $50B - China lands its first reusabl...

Read more: https://t.co/wd6JKRcsHH

Top stories in AI today: - 200+ researchers put AI’s job shock on clock - Musk, Altman trade insul...