The Rundown AI(@TheRundownAI)

Cognition's new SWE-2 (a Kimi K3 post-train) coming in with some powerful coding benchmarks. "On F...

The Rundown AI (@TheRundownAI) — Thu, 10 Sep 2026 16:44:04 GMT

Cognition's new SWE-2 (a Kimi K3 post-train) coming in with some powerful coding benchmarks. 

"On FrontierCode, SWE-2 achieves a score of 50.00%... matching Fable 5.1 at 64% lower cost."

Cognition@cognition

Introducing SWE-2, our closest model yet to the frontier.

On leading evals, it scores on par with recent frontier models – at up to 70% lower cost.

We scaled RL to multiple trillions of parameters, with a refined recipe that pushes the Pareto curve on both capabilities & cost.

🔗 View Quoted Tweet

💬6🔄2❤️11👀7218📊5

⚡ Powered by xgo.ing

DeepSeek V4.1 Flash is here, and DeepSeek calls it "the smallest model in our new architecture famil...

The Rundown AI (@TheRundownAI) — Thu, 10 Sep 2026 15:26:27 GMT

DeepSeek V4.1 Flash is here, and DeepSeek calls it "the smallest model in our new architecture family," with a V4.1-Pro still to come.

It beats GPT-5.6 Sol and Claude Opus 5 on several agentic, coding, and cyber benchmarks.

Flash is a third the size of DeepSeek's V4-Pro, but beats it on almost every benchmark in the table.

Cost: $0.15 per million input tokens, $0.60 per million output.

Pretty wild scores at that size and bottom-barrel prices, and the bigger models in the family are still on the way.

💬7🔄2❤️30👀7177📊9

⚡ Powered by xgo.ing

Anthropic disclosed a fourth case of Claude breaking into real systems during cyber testing

The Rundown AI (@TheRundownAI) — Wed, 09 Sep 2026 19:37:02 GMT

Anthropic disclosed a fourth case of Claude breaking into real systems during cyber testing

Anthropic@AnthropicAI

We’re sharing our alignment assessment of incidents in which Claude models gained unauthorized access to real systems during third-party cybersecurity evaluations mistakenly connected to the internet.

METR will also conduct an independent investigation, with wide-ranging access, including to transcripts beyond the window in which the incidents occurred, and to Anthropic employees permitted to share confidential information. Our initial agreement runs for eight weeks, and we intend to give METR as much time as it deems necessary to complete a thorough investigation. anthropic.com/research/align…

🔗 View Quoted Tweet

💬5🔄2❤️14👀7395📊5

⚡ Powered by xgo.ing

Full research: https://t.co/xIDVd2K8Xz

The Rundown AI (@TheRundownAI) — Wed, 09 Sep 2026 17:54:53 GMT

Full research:

anthropic.com/institute/econ…

💬0🔄1❤️2👀1409📊1

⚡ Powered by xgo.ing

Yesterday, Anthropic's alignment science lead put a >10% chance on AI killing all humans within a de...

The Rundown AI (@TheRundownAI) — Wed, 09 Sep 2026 17:52:49 GMT

Yesterday, Anthropic's alignment science lead put a >10% chance on AI killing all humans within a decade.

For some lighter reading, Anthropic's economics team just published what happens to the U.S. economy if it doesn't.

Three scenarios for 2030, GDP measured against a world without AI:

- Modest: AI's impact is "something like the internet's." GDP +1.6%.

- Substantial: AI can do half of all knowledge work, most of it on its own. GDP +8.3%. Knowledge-worker wages "essentially flat."

- Extreme: AI does nearly all knowledge work. GDP +32.4% (!) to $44.4T, growth hits 15% a year, and the economy doubles every 4.5 years.

💬6🔄1❤️14👀7015📊7

⚡ Powered by xgo.ing

Warner Music Group went from a legal fight with Suno to helping develop its new AI music models. Th...

The Rundown AI (@TheRundownAI) — Wed, 09 Sep 2026 15:42:51 GMT

Warner Music Group went from a legal fight with Suno to helping develop its new AI music models.

The two settled in November 2025. Today, Suno launched v6 with Warner, BMG and Believe as development partners.

Some new abilities with the v6 models: 

- Rewrite a single lyric or change a chorus without replacing the rest of the song

- Take vocals from one of their songs and drums from another, then combine them

- Pull out specific elements of a song and build around around them

- Use text, audio, images or video as the starting references for music

The three new versions: flagship v6 for precision and control, v6-wild for more experimental results, and a v6-mini free tier. The first two require Pro ($8/mo) or Premier ($24/mo) subscriptions.

Suno also said it is developing artist-specific experiences where artists can choose whether to participate and get paid if they do.

💬5🔄0❤️17👀6891📊7

⚡ Powered by xgo.ing

Can paralysis be undone... even a little? Keith Thomas broke his neck diving into a pool. Five el...

The Rundown AI (@TheRundownAI) — Tue, 08 Sep 2026 21:54:19 GMT

Can paralysis be undone... even a little?

Keith Thomas broke his neck diving into a pool. 

Five electrode arrays and an AI model later, he can lift an egg without crushing it -- with some results remaining even with the system switched off:

Your browser does not support the video tag.

🔗 View on Twitter

💬7🔄2❤️18👀7494📊8

⚡ Powered by xgo.ing

OpenAI just released ChatGPT Images 2.5, calling it its new "state-of-the-art image model." What's...

The Rundown AI (@TheRundownAI) — Tue, 08 Sep 2026 18:53:28 GMT

OpenAI just released ChatGPT Images 2.5, calling it its new "state-of-the-art image model." 

What's new:

- Up to 50% faster generation than Images 2.0
- Edits only what you ask for, holds earlier edits
- Sketch: draw in ChatGPT, get a finished image

Your browser does not support the video tag.

🔗 View on Twitter

💬8🔄16❤️72👀9678📊20

⚡ Powered by xgo.ing

BREAKING: OpenAI says it has a solution to the Navier-Stokes Millennium Prize Problem, a $1M math pr...

The Rundown AI (@TheRundownAI) — Tue, 08 Sep 2026 17:38:17 GMT

BREAKING: OpenAI says it has a solution to the Navier-Stokes Millennium Prize Problem, a $1M math problem unresolved for roughly 90 years.

OpenAI said it was produced by about 10,000 coordinating AI agents over 88 hours on an unreleased model it calls "significantly more capable than GPT-6 Astra."

OpenAI also addressed this morning's accusation from NYU mathematician Tristan Buckmaster, who said the company pressured him over credit for a related result:

"We congratulate Levent Alpöge and Tristan Buckmaster on their remarkable mathematical work. We (the researchers and the agents) did not see any of their work through any means until they released it publicly — in particular, no specific user data was accessed in order to solve this problem."

OpenAI@OpenAI

We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.

The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra.

The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.

🔗 View Quoted Tweet

💬4🔄0❤️15👀7443📊4

⚡ Powered by xgo.ing

The first teaser for "Artificial" is out, the movie covering OpenAI's 2023 board drama. Amazon drop...

The Rundown AI (@TheRundownAI) — Tue, 08 Sep 2026 16:04:21 GMT

The first teaser for "Artificial" is out, the movie covering OpenAI's 2023 board drama.

Amazon dropped the film in June after investing $50B in OAI.

Andrew Garfield (who previously played Eduardo Saverin in "The Social Network") stars as Sam Altman:

Your browser does not support the video tag.

🔗 View on Twitter

💬4🔄4❤️33👀7593📊10

⚡ Powered by xgo.ing

Every day we feature a top community workflow in The Rundown AI newsletter. Share what you've built...

The Rundown AI (@TheRundownAI) — Mon, 07 Sep 2026 17:02:40 GMT

Every day we feature a top community workflow in The Rundown AI newsletter.

Share what you've built and 2M+ readers might see it next: app.therundown.ai/community

💬0🔄0❤️1👀1598

⚡ Powered by xgo.ing

The taste decoder Tony couldn't put his design taste into words, so he stopped trying -- he had AI...

The Rundown AI (@TheRundownAI) — Mon, 07 Sep 2026 17:02:39 GMT

The taste decoder 

Tony couldn't put his design taste into words, so he stopped trying -- he had AI mock up three wildly different versions of a real report, picked one, said why, and repeated.

Three rounds later it knew his taste better than any prompt could, and he had it write reusable design guidelines for next time.

Credit to Tony Ojeda: app.therundown.ai/community/post…

💬1🔄0❤️1👀2021📊2

⚡ Powered by xgo.ing

2M+ people read The Rundown. More of them are building with AI every day. In this week's top 5 from...

The Rundown AI (@TheRundownAI) — Mon, 07 Sep 2026 17:02:36 GMT

2M+ people read The Rundown. More of them are building with AI every day.

In this week's top 5 from our Community Workflow Hub:  a laser-engraving shop's homegrown CRM, a meal tracker you text like a friend, a blood pressure app built on doctor's orders, and more:

💬6🔄0❤️16👀7491📊7

⚡ Powered by xgo.ing

6 Astra is starting to show up in the model picker 👀

The Rundown AI (@TheRundownAI) — Fri, 04 Sep 2026 18:58:06 GMT

6 Astra is starting to show up in the model picker 👀

The Rundown AI@TheRundownAI

GPT-6 Astra is here, with OpenAI calling it “a generational leap in capability”.
 
A few of the headline benchmarks:

- ARC- AGI 3 – 98.6% (!), compared to 7.8% for 5.6 Sol
- Exploit-Bench (cybersecurity) - 100% (!)
- FrontierMath Tier 4 (v2) - 97.6%, vs. 87.8% for Fable 5.1
- DeepSWE v1.1 - 74.1%, vs. 67.4% for Fable 5.1

New highs for the industry across math, science, health, computer use, coding, cybersecurity, and more.

OpenAI's Greg Brockman: "Welcome to the AGI era."

🔗 View Quoted Tweet

💬9🔄2❤️20👀8437📊8

⚡ Powered by xgo.ing

GPT-6 Astra is here, with OpenAI calling it “a generational leap in capability”. A few of the head...

The Rundown AI (@TheRundownAI) — Thu, 03 Sep 2026 18:07:36 GMT

GPT-6 Astra is here, with OpenAI calling it “a generational leap in capability”.
 
A few of the headline benchmarks:

- ARC- AGI 3 – 98.6% (!), compared to 7.8% for 5.6 Sol
- Exploit-Bench (cybersecurity) - 100% (!)
- FrontierMath Tier 4 (v2) - 97.6%, vs. 87.8% for Fable 5.1
- DeepSWE v1.1 - 74.1%, vs. 67.4% for Fable 5.1

New highs for the industry across math, science, health, computer use, coding, cybersecurity, and more.

OpenAI's Greg Brockman: "Welcome to the AGI era."

💬54🔄90❤️1549👀142314📊229

⚡ Powered by xgo.ing

The GPT-6 hype teasers have started!

The Rundown AI (@TheRundownAI) — Thu, 03 Sep 2026 15:21:37 GMT

The GPT-6 hype teasers have started!

OpenAI@OpenAI

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬2🔄0❤️38👀9977📊5

⚡ Powered by xgo.ing

The heavy week of AI releases continues with Meta and Muse Spark 1.3. Big jumps across the board o...

The Rundown AI (@TheRundownAI) — Wed, 02 Sep 2026 19:36:56 GMT

The heavy week of AI releases continues with Meta and Muse Spark 1.3. 

Big jumps across the board over 1.2, with Mark Zuckerberg calling it "frontier performance almost too cheap to meter."

💬6🔄3❤️28👀7744📊9

⚡ Powered by xgo.ing

NEW: NYC Public Schools are banning generative AI from pre-K through eighth grade for the entire 202...

The Rundown AI (@TheRundownAI) — Wed, 02 Sep 2026 17:45:00 GMT

NEW: NYC Public Schools are banning generative AI from pre-K through eighth grade for the entire 2026-27 school year, impacting over 600k students.

Mayor Zohran Mamdani said the tech industry "wants us to believe that A.I.-powered early education is not only inevitable, but necessary," adding: "We do not see it that way."

💬7🔄2❤️22👀7575📊9

⚡ Powered by xgo.ing

U.S. Commerce Secretary Howard Lutnick told Axios "We trust Anthropic," adding the company has "done...

The Rundown AI (@TheRundownAI) — Wed, 02 Sep 2026 17:19:38 GMT

U.S. Commerce Secretary Howard Lutnick told Axios "We trust Anthropic," adding the company has "done what we asked" and is "back on the right side."

The reversal follows a summer in which Commerce restricted Fable and Mythos under export controls and the Pentagon branded Anthropic a supply-chain risk, a designation usually reserved for foreign adversaries.

💬6🔄2❤️17👀7936📊8

⚡ Powered by xgo.ing

Is Google back? Gemini 3.8 Flash already appears to be live for some and is expected to be officia...

The Rundown AI (@TheRundownAI) — Wed, 02 Sep 2026 15:32:57 GMT

Is Google back? 

Gemini 3.8 Flash already appears to be live for some and is expected to be officially released today. 

Benchmarks looking strong, especially with the same Flash price point:

💬8🔄1❤️25👀7342📊9

⚡ Powered by xgo.ing

Claude Fable 5.1 has arrived

The Rundown AI (@TheRundownAI) — Tue, 01 Sep 2026 18:04:33 GMT

Claude Fable 5.1 has arrived

💬4🔄1❤️15👀7511📊5

⚡ Powered by xgo.ing

Senator Bernie Sanders is escalating the AI pause conversation, with a new op-ed in Fox News arguing...

The Rundown AI (@TheRundownAI) — Tue, 01 Sep 2026 16:09:17 GMT

Senator Bernie Sanders is escalating the AI pause conversation, with a new op-ed in Fox News arguing CEOs must "immediately pause the development of increasingly powerful AI."

Sanders says the pause must be international, and wants the U.S. and China to negotiate an AI agreement when Trump and Xi meet in Washington later this month.

💬3🔄1❤️10👀7432📊5

The Rundown AI(@TheRundownAI)

Read more: https://t.co/5mpcEk0Zju

Top stories in tech today: - Apple Watch to listen and take notes - Air taxis are taking off in Tex...

Read more: https://t.co/QnFPqvZ7fg

Top stories in AI today: - Anthropic opens the files on global Claude misuse - Rowan’s Corner: I us...

Cognition's new SWE-2 (a Kimi K3 post-train) coming in with some powerful coding benchmarks. "On F...

DeepSeek V4.1 Flash is here, and DeepSeek calls it "the smallest model in our new architecture famil...

Read more: https://t.co/WPiPqXjtm1

Top stories in robotics today: - Inside Tesla’s messy Cybercab debut - Ex-Boston Dynamics exec want...

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

Top stories in AI today: - Anthropic employee’s exit post sparks AI doom debate - Nate's Notebook: ...

Anthropic disclosed a fourth case of Claude breaking into real systems during cyber testing

Full research: https://t.co/xIDVd2K8Xz

Yesterday, Anthropic's alignment science lead put a >10% chance on AI killing all humans within a de...

Warner Music Group went from a legal fight with Suno to helping develop its new AI music models. Th...

Read more: https://t.co/xOhg6Qe1BL

Top stories in AI today: - OpenAI claims a Millennium Prize problem - Meta debuts new Muse personal...

Can paralysis be undone... even a little? Keith Thomas broke his neck diving into a pool. Five el...

OpenAI just released ChatGPT Images 2.5, calling it its new "state-of-the-art image model." What's...

BREAKING: OpenAI says it has a solution to the Navier-Stokes Millennium Prize Problem, a $1M math pr...

The first teaser for "Artificial" is out, the movie covering OpenAI's 2023 board drama. Amazon drop...

Read more: https://t.co/XA6QlkSffJ

Top stories in tech today: - Apple gets ready to ‘unfold’ the iPhone - Australia is coming for the ...

Read more: https://t.co/HN3lJeZzLX

Top stories in AI today: - OpenAI opens the books on its AI research intern - Insilico's AI-designe...

Every day we feature a top community workflow in The Rundown AI newsletter. Share what you've built...

The taste decoder Tony couldn't put his design taste into words, so he stopped trying -- he had AI...

2M+ people read The Rundown. More of them are building with AI every day. In this week's top 5 from...

Read more: https://t.co/sk0HNv4qMI

Top stories in robotics today: - Robots on the runway at IFA 2026 - Figure locks up 100K GPUs to t...

Read more: https://t.co/6aSD9J78rJ

Top stories in AI today: - Another OpenAI agent swarm surfaces - The Rundown Roundtable: Our AI use...

6 Astra is starting to show up in the model picker 👀

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

Top stories in tech today: - A surgeon just operated in Apple Vision Pro - Ray Kurzweil backs a bra...

Read more: https://t.co/r0Px1XAK0B

Top stories in AI today: - OpenAI’s “generational leap” with GPT-6 Astra - Rowan’s Corner: Where th...

GPT-6 Astra is here, with OpenAI calling it “a generational leap in capability”. A few of the head...

The GPT-6 hype teasers have started!

Read more: https://t.co/sv4HZDv3RP

Top stories in robotics today: - Uber and Wayve bring robotaxis to London - Altman says OpenAI ‘wil...

Read more: https://t.co/wT6jRBpM4m

Top stories in AI today: - Meta, Google join the September AI launch party - Nate's Notebook: Tech ...

The heavy week of AI releases continues with Meta and Muse Spark 1.3. Big jumps across the board o...

NEW: NYC Public Schools are banning generative AI from pre-K through eighth grade for the entire 202...

U.S. Commerce Secretary Howard Lutnick told Axios "We trust Anthropic," adding the company has "done...

Is Google back? Gemini 3.8 Flash already appears to be live for some and is expected to be officia...

Read more: https://t.co/jCkUTwmi8a

Top stories in AI today: - Fable 5.1 ends the frontier's quiet period - Bernie Sanders escalates A...

Claude Fable 5.1 has arrived

Senator Bernie Sanders is escalating the AI pause conversation, with a new op-ed in Fox News arguing...