OpenAI announced Defense Factory, a new continuous defense concept.
"Defender's window" is the gap between the cyber capabilities of open-weight models and the current security level.
Without continuous defense practices that rely on frontier models, this gap will close, enabling malicious actors to freely abuse vulnerable systems.
> "Agents can now conduct long-running cyber operations by abusing increasingly available open-weight models. In response, at OpenAI weβre building a Defense Factory. An automated defense operation to find, validate, and fix vulnerabilities continuously."
> "Defenders have two structural advantages. They can give agents direct access to their code and use frontier models to get a head start over attackers abusing broadly available open-weight models."
What does it mean?
βEverything that makes it successful is exactly what makes it dangerous.β
Internet won't be safe anymore π€
"Defender's window" is the gap between the cyber capabilities of open-weight models and the current security level.
Without continuous defense practices that rely on frontier models, this gap will close, enabling malicious actors to freely abuse vulnerable systems.
> "Agents can now conduct long-running cyber operations by abusing increasingly available open-weight models. In response, at OpenAI weβre building a Defense Factory. An automated defense operation to find, validate, and fix vulnerabilities continuously."
> "Defenders have two structural advantages. They can give agents direct access to their code and use frontier models to get a head start over attackers abusing broadly available open-weight models."
What does it mean?
βEverything that makes it successful is exactly what makes it dangerous.β
Internet won't be safe anymore π€
β€7π4π1 1
π¨ AI News | TestingCatalog
DAILY AI BRIEF π β Sept 3 GOOGLE π₯: > Gemini 3.8 Flash is live β third Flash in six weeks. Workhorse for coding and agents, 1M context. Intro price $0.75 / $3.75 through Dec 31. > Flash Cyber is the defender twin (vuln find + auto-patch). Gated to Fairwind:β¦
DAILY AI BRIEF π β Sept 4
OPENAI π₯:
> GPT-6 Astra is officially out β SOTA across computer use, coding, science, and cyber. 98% FrontierMath Tier 4, 63% / 99.9% on ARC-AGI-3 (standard vs adapter harness), 100% ExploitBench.
> Not in your ChatGPT dropdown yet. Rolling to a limited set of orgs / Daybreak first; Plus, Pro, Business, Enterprise, API, and AWS βover the coming days.β Hope is weekend. Until then, enjoy the banked limit resets.
MICROSOFT π₯:
> MAI-Transcribe-2 is live. Claims fastest + cheapest + most accurate speech-to-text: #1 on FLEURS across 60 languages (5.2% avg WER), 10Γ GPT-Transcribe, 7Γ Scribe v2, 5Γ Gemini 3.5 Transcribe. Foundry + MAI Playground + OpenRouter.
GOOGLE π₯: (for Ultra Trusted Testers?)
> Gemini Live can now use Workspace connectors β Gmail, Keep, Docs. Conversational inbox search, voice-first drafting, brain-dump-to-lists.
> New Labs experiment: Play with Putty β real-time collaborative vibe coding. Multiple people in one workspace building sites/tools live. Waitlist.
TESLA π₯:
> Cybercab is in the Austin robotaxi fleet. No steering wheel, no pedals, butterfly doors, public rides starting in limited zones. Elon: the future is here.
NVIDIA π₯:
> Confirmed $12.9B acquisition of Hugging Face. Jensen: platform stays open, multi-cloud, multi-accelerator, no NVIDIA compute required. 18M developers, 3M+ models. Largest Nvidia buy ever. Close targeted for 2027, subject to regulatory approval.
* Used Grok to compose this brief, cherry-picking the news and doing some post-editing.
OPENAI π₯:
> GPT-6 Astra is officially out β SOTA across computer use, coding, science, and cyber. 98% FrontierMath Tier 4, 63% / 99.9% on ARC-AGI-3 (standard vs adapter harness), 100% ExploitBench.
> Not in your ChatGPT dropdown yet. Rolling to a limited set of orgs / Daybreak first; Plus, Pro, Business, Enterprise, API, and AWS βover the coming days.β Hope is weekend. Until then, enjoy the banked limit resets.
MICROSOFT π₯:
> MAI-Transcribe-2 is live. Claims fastest + cheapest + most accurate speech-to-text: #1 on FLEURS across 60 languages (5.2% avg WER), 10Γ GPT-Transcribe, 7Γ Scribe v2, 5Γ Gemini 3.5 Transcribe. Foundry + MAI Playground + OpenRouter.
GOOGLE π₯: (for Ultra Trusted Testers?)
> Gemini Live can now use Workspace connectors β Gmail, Keep, Docs. Conversational inbox search, voice-first drafting, brain-dump-to-lists.
> New Labs experiment: Play with Putty β real-time collaborative vibe coding. Multiple people in one workspace building sites/tools live. Waitlist.
TESLA π₯:
> Cybercab is in the Austin robotaxi fleet. No steering wheel, no pedals, butterfly doors, public rides starting in limited zones. Elon: the future is here.
NVIDIA π₯:
> Confirmed $12.9B acquisition of Hugging Face. Jensen: platform stays open, multi-cloud, multi-accelerator, no NVIDIA compute required. 18M developers, 3M+ models. Largest Nvidia buy ever. Close targeted for 2027, subject to regulatory approval.
* Used Grok to compose this brief, cherry-picking the news and doing some post-editing.
1β€8 5 1
Claude Code on iOS will get support for / commands as well.
Does anybody have it already? π
Does anybody have it already? π
β€8π€£7π3
ANTHROPIC π₯: Claude Code usage limits have been reset!
Thatβs new π
AI space feels like a battleground these days tbh.
Thatβs new π
AI space feels like a battleground these days tbh.
π12β€7π1π1
OpenAI launches GPT-6 Astra across ChatGPT and API
OpenAI launched GPT-6 Astra, a computer-use and software model rolling out across ChatGPT, API, Azure, and AWS.
π #openai @testingcatalog
OpenAI launched GPT-6 Astra, a computer-use and software model rolling out across ChatGPT, API, Azure, and AWS.
π #openai @testingcatalog
TestingCatalog AI News
OpenAI launches GPT-6 Astra across ChatGPT and API
GPT-6 Astra is rolling out to ChatGPT paid plans and major cloud APIs with faster computer use, new Codex context memory and stronger safeguards.
β€4π3
Astra testing time π
> This is GPT-6, a new generation of intelligence. Astra is state-of-the-art in coding, computer use, science, and professional work. Give it a hard problem, a half-formed idea, or anything you've been meaning to build. See where it takes you.
* Got it on a Business (not Business Premium) plan already.
** Plus users should have it as well!
> This is GPT-6, a new generation of intelligence. Astra is state-of-the-art in coding, computer use, science, and professional work. Give it a hard problem, a half-formed idea, or anything you've been meaning to build. See where it takes you.
* Got it on a Business (not Business Premium) plan already.
** Plus users should have it as well!
π¨ AI News | TestingCatalog
DAILY AI BRIEF π β Sept 4 OPENAI π₯: > GPT-6 Astra is officially out β SOTA across computer use, coding, science, and cyber. 98% FrontierMath Tier 4, 63% / 99.9% on ARC-AGI-3 (standard vs adapter harness), 100% ExploitBench. > Not in your ChatGPT dropdownβ¦
DAILY AI BRIEF π β Sept 5
OPENAI π₯
ANTHROPIC π₯
GOOGLE π₯
META π₯
* Used Grok to compose this brief, cherry-picking the news and doing some post-editing.
OPENAI π₯
- Astra is rolled out to all Plus and Business β not just Pro / Enterprise. API live at $10 / $50. GPT-6 Pro in Chat is Astra.
- Full banked reset today for Plus, Pro, and Business. Lands EOD. Create or upgrade before 8pm PT and you get it too.
- Tibo: internal Astra pulled some shipping plans 6 months forward to DevDay.
ANTHROPIC π₯
Claude Code
weekly limits reset
Friday β
Max plans only
. Fable 5.1 + long weekend. Pro/Team did not get this flush.
GOOGLE π₯
Lyria 3.5
is in Gemini, AI Studio, and the API. Better vocals, richer arrangements, short or full tracks. Tools > Create music.
META π₯
Muse Spark 1.3 max
is public in Muse Code and the Meta Model API. Wang: use max even if you already tried high / xhigh.
* Used Grok to compose this brief, cherry-picking the news and doing some post-editing.
β€6 6π5
OpenAI is testing a new Writing Style feature for ChatGPT.
> Users can connect documents, emails and messaging apps as reference examples.
> ChatGPT will be able to write text in βuserβs voiceβ, based on provided references.
> Availability is limited to a small group of users.
> Users can connect documents, emails and messaging apps as reference examples.
> ChatGPT will be able to write text in βuserβs voiceβ, based on provided references.
> Availability is limited to a small group of users.
β€8π₯6 4π3
Google keeps transforming Gemini desktop into superapp
Google is testing major Gemini desktop upgrades: Ask and Assign modes, folder-scoped agent tasks, remote computer links, local app control, Obsidian support, customization options, Finder integration, and prep for future models.
π #google @testingcatalog
Google is testing major Gemini desktop upgrades: Ask and Assign modes, folder-scoped agent tasks, remote computer links, local app control, Obsidian support, customization options, Finder integration, and prep for future models.
π #google @testingcatalog
TestingCatalog AI News
Google keeps transforming Gemini desktop into superapp
Closed tests of Gemini desktop add Ask and Assign modes, Obsidian connections and a Finder action. Public release timing remains unclear.
β€4π2
Media is too big
VIEW IN TELEGRAM
GOOGLE π₯: Gemini desktop app keeps being transformed into a superapp with "Ask" and "Assign" modes, computer use, and more!
> Nano Banana 2.5 Flash is in the works.
> Integration with Obsidian spotted.
> Users will also be able to "Ask Gemini" directly from Finder.
> Nano Banana 2.5 Flash is in the works.
> Integration with Obsidian spotted.
> Users will also be able to "Ask Gemini" directly from Finder.
β€9 8π3
Grok Imagine Agent got an upgrade and is now using Image 2.0 image model underneath.
> "Make me an ad banner for @testingcatalog"
That's quite a good result π
> "Make me an ad banner for @testingcatalog"
That's quite a good result π
β€4π1
SpaceXAI did a usage reset for all Grok Bot users.
I was planning to properly test those for a while. Looks like it the time since Astra ate all my banked resets already.
I was planning to properly test those for a while. Looks like it the time since Astra ate all my banked resets already.
β€6π3
GPT-6 Astra from OpenAI claimed a top spot on WebDev Arena, surpassing recently released Claude Fable 5.1 by 35 points.
> βIt also reshapes the Pareto frontier as the best-performing model at $40/Mtoken, which matches the latest Claude model pricing.β
Are we set to hit 2k by the end of this year?
> βIt also reshapes the Pareto frontier as the best-performing model at $40/Mtoken, which matches the latest Claude model pricing.β
Are we set to hit 2k by the end of this year?
β€11π2 1