OpenAI announces GPT-5.5-Cyber (new) model update, which scores 85.6% on CyberGym benchmark in comparison to 81.9% in its early version.

Codex got a new Security plugin too 👀

Sakana AI releases Fugu Ultra system to rival top AI labs

Sakana AI launched Fugu Ultra, a public OpenAI-compatible orchestration model for complex engineering, research, cybersecurity, and data analysis.

🗞 #ai @testingcatalog

OpenAI launches new security tools and updates GPT-5.5-Cyber

OpenAI expands Daybreak from bug discovery to patch delivery with Codex Security, limited GPT-5.5-Cyber access, partner distribution, and Patch the Planet for open-source projects, targeting validated fixes across enterprise, government, and OSS.

🗞 #chatgpt @testingcatalog

Flashcards are now editable on NotebookLM 👀

Users can adjust the text of questions and answers, plus add new cards to the stack.

FlashcardLM ⚡️

Google tests literature review matrix tool for NotebookLM

Google is developing a NotebookLM “Lit Review” artifact that turns uploaded sources into a literature review matrix. Aimed at research-heavy reading, it may connect with Play Books, but launch timing and citation reliability remain unclear.

🗞 #notebooklm @testingcatalog

Anthropic prepares Cowork support for mobile apps

Anthropic’s iOS app hints at Cowork shifting from desktop-tethered Dispatch to cloud and web, enabling mobile task runs and a unified scheduled-actions view. App code also points to a selectable voice model, suggesting a coming refresh.

🗞 #claude @testingcatalog

BREAKING 🔥: OpenAI is preparing "Bidi 1" for the upcoming web release!

> A new voice model will be available in settings, alongside standard and advanced options.
> Voice mode bubble will have a Yellow color instead of blue.

How soon? 👀

BYTEDANCE 🔥: Seedance 2.5 has been officially announced, along with an updated Seedance 2.0.

- Seedance 2.0 now supports 4k output
- Seedance 2.5 will be able to generate 30-second videos in one go
- ByteDance also announced a new AI copyright commercialization platform

This video ad is stunning 👀

BREAKING 🔥: First tests of "Bidi 1", an upcoming bidirectional voice model from OpenAI. This upgrade will arrive in ChatGPT and, potentially, in Codex soon as well.

> Bidi 1 can speak over while you are talking and keep listening.
> Bidi 1 can switch between tasks back and force mid-sentence.
> Bidi 1 is much better at handling interruptions and pauses.
> Bidi 1 can better keep and memorize the context while you speak.

There is still a cap on how long it can keep speaking, which is expected, but it easily counted to 23 without pausing.

* Bidi 1 is not available yet, but given all the recent preparations, we will get it very, very soon.

OpenAI prepares bidirectional voice mode for rollout on ChatGPT

OpenAI is testing Bidi 1, a bidirectional voice model for ChatGPT that can listen and speak at once, handle interruptions, retain context, and reduce unwanted cut-ins. A wider web and mobile rollout may begin soon.

🗞 #chatgpt @testingcatalog

OPENAI 🔥: An upcoming Bidi 1 voice model will be able to translate in real-time!

This will unlock a huge pile of use cases to be built on top of when it lands on the APIs.

OPENAI 🔥: Bidi 1, an upcoming voice model from OpenAI, can sing and generate different sounds too.

A rap sample 👀

Mistral AI launched OCR 4 👀

> Win rates averaging 72%, alongside the top overall score on OlmOCRBench (85.20).
> Alongside the extracted text, OCR 4 returns bounding boxes, typed-block classification, and inline confidence scores.
> OCR 4 is an ingestion component of Search Toolkit, Mistral's open-source, composable search framework.
> Support for 170 languages across 10 language groups.
> OCR 4 is compact enough to run in a single container.

Anthropic launched Claude Tag for Team and Enterprise users.

Claude Tag works in Slack and can tackle more complex tasks, break them down into smaller milestones, and integrate with connected tools.

A new AI coworker 👀

ClickUp rolls out Brain² AI with deep workspace context

ClickUp relaunched Brain² as a context-aware AI coworker that acts across workspace data, picks models per task, creates deliverables and agents, cites sources, and supports secure, permission-aware work management.

🗞 #sponsored @testingcatalog

Meta announced a new series of Meta Glasses in partnership with EssilorLuxottica.

> Compatible with prescription lenses.
> 26 styles across a range of colors, lenses, and frames.
> Launching with Meta AI powered by Muse Spark from day one.

While my Meta HSTN still didn't get Muse Spark, tho

Mistral launches OCR 4 for multilingual document extraction

Mistral OCR 4 targets enterprise document understanding with structured extraction, block classification, confidence scoring, and support for 170 languages. It posts strong benchmark results, lower cost, faster processing, and flexible deployment.

🗞 #mistral @testingcatalog

Anthropic launches Claude Tag on Team and Enterprise plans

Anthropic launched Claude Tag, a Slack-based team agent for Claude Enterprise and Team users. It works in shared channels, uses channel-scoped permissions and admin controls, and shifts Slack work billing from individuals to organizations.

🗞 #claude @testingcatalog

Meta launches AI glasses with three new styles from $299

Meta launched AI-powered Meta Glasses with EssilorLuxottica, offering 26 prescription-compatible styles from $299. Features include open-ear audio, hands-free capture, 8-hour battery life, and Meta AI with live translation support.

🗞 #ai @testingcatalog