🚨 AI News | TestingCatalog
7.94K subscribers
4.64K photos
777 videos
40 files
4.41K links
Latest AI News on AI Agents, Model Releases, Tools, Leaks, and Rumors πŸ—ž
Download Telegram
ANTHROPIC πŸ”₯: According to the latest report from METR, Anthropic is accelerating its internal AI model development, and this is attributable to Anthropic’s extensive use of AI for R&D

> Anthropic have accelerated 1.9x since October 2025.

> The chart categorizes Opus 5.5 as "non-frontier" and highlights internal models named Model 3 and Model 4, with substantial improvement achieved in a very short period.

RSI bench? πŸ‘€
6❀2πŸ‘2
Media is too big
VIEW IN TELEGRAM
META πŸ”₯: Muse Voice and expressive avatars are rolling out for Muse users in the US.

> Voice helps users think out loud, making tasks like brainstorming, trip planning, and interview prep easier.

> Your Muse’s avatar reacts as it listens and speaks, with expressions and gestures that match the conversation.

Additionally, Muse is rolling out in Mexico πŸ‡²πŸ‡½
πŸ‘3❀2
Meta rolls out Muse voice and expressive avatars in the US

Meta is bringing Muse voice conversations and expressive avatars to mobile, web and Mac. The rollout starts in the US, with real-time spoken replies, interrupt support, background task handling, and voice or avatar display options.

πŸ—ž #meta @testingcatalog
πŸ‘3❀2
Claude adds Personal Finance with read-only bank access

Anthropic is rolling out Claude Personal Finance in US beta for Pro, Max, and Ultra users. It uses read-only Plaid-linked account data to answer spending questions, flag unusual charges, and track bills without ads or financial products.

πŸ—ž #anthropic @testingcatalog
πŸ‘3❀21
OpenAI launches GPT-6 Cyber for Daybreak Red customers

OpenAI launched GPT‑6 Cyber for approved Daybreak Red users, pairing stronger vulnerability discovery and exploit validation with strict scope controls, monitoring and safeguards. Early results show major gains in kernel, web and Firefox research.

πŸ—ž #openai @testingcatalog
❀5πŸ‘31
🚨 AI News | TestingCatalog
Please open Telegram to view this post
VIEW IN TELEGRAM
❀53
Testing a different thing today πŸ‘€

I haven’t worked with Windows in a while (10+ years). Curious to see what the Surface Ultra offers.

> The Microsoft Surface Laptop Ultra uses an NVIDIA Blackwell RTX GPU built into the NVIDIA RTX Spark N1X superchip.

I am obviously checking Perplexity Computer and Hermes, powered by local models like Nemotron and others.

What else should I try to run there?

I’m looking for recommendations to stress-test its performance running local workloads.
❀6
ALIBABA πŸ”₯: Β«The becoming beginsΒ» teaser has been posted on the Qwen X account. We expect Qwen4-Max, Qwen4-Flash, Qwen4-Plus, and Qwen4-27B models to arrive soon.

Waiting for Qwen4-27B the most πŸ‘€
❀8πŸ”₯7πŸ‘2
META πŸ”₯: Muse and Meta AI apps now run on new specialized models trained with Muse Wonder.

Both Active Memory and Active Focus, introduced in Muse Wonder, are now available in Muse and Meta AI.

The Personal Superintelligence Index is up 9%, and Active Memory Eval is up 14%.
❀2
META πŸ”₯: Muse now has Spaces!

Users can ask Muse to generate an interactive page on a specific topic or visualize certain data.

These spaces work like interactive artifacts powered by data from sources Muse can access.

I've asked it to create a space for "testing" and got an interactive playground. I think that Meta should now find a way to make these spaces shareable; that would be big.

Testing time! πŸ‘€
❀1πŸ‘€1
Claude for iOS is getting another layout redesign with Chats, Projects and Code being separate tabs on a side nav bar.

Other options like Dispatch or Personal Finance got moved into the profile menu.

I feel like a bottom nav was better πŸ‘€
❀2