What's new? Issue #219 🗞️

Greetings!

It looks like the quiet period is over - we’re finally seeing clear signs of major upcoming releases in the AI world!

Starting beyond the usual giants (OpenAI and Google), last week brought several notable launches from smaller labs and some interesting announcements. Most notably, Kimi K2 Thinking is now available. It’s impressive that they managed to train such a high-performance model at a low cost, putting pressure on OpenAI, Google, and others to step up their own releases to maintain their performance advantage. Not only is Kimi K2 Thinking cheaper, but Moonshot, the company behind it, has a strong consumer focus with feature-rich apps, unlike DeepSeek. Their platform already includes agent capabilities with computer use, and more advanced agentic features are promised soon. This approach is quite different from last year’s DeepSeek, and Kimi has real potential to challenge OpenAI’s dominance in the market.

We’ve also seen the Perplexity Comet browser for Android rolled out to early adopters. It appears to bring agentic capabilities similar to its desktop version, including the ability to operate web pages autonomously—something worth testing soon.

Additionally, Telegram has announced its entry into the AI space with Cacoon, a decentralised platform that enables developers to build AI-powered apps and features. We might see these features integrated natively into Telegram soon, which would be exciting. Notably, Telegram plans to let users exchange TON cryptocurrency for compute, a forward-thinking move that’s likely to spur innovation.

Turning to the major players, OpenAI is feeling pressure from both sides: from below by Kimi K2 Thinking, which offers comparable performance at lower prices, and from above by Google’s anticipated Gemini 3 launch, which is expected to set a new standard for model performance. OpenAI’s response appears to be the upcoming GPT-5.1 family. Evidence points to at least three models (including Thinking and Pro variants), and possibly upgraded Codex as well.

There’s also a stealth model, Polaris Alpha, available for public testing on OpenRouter; while unconfirmed, it almost certainly belongs to OpenAI and already demonstrates strong results in creative writing and other tasks. It’s not yet clear how much compute budget is assigned to the OpenRouter version, so expectations for GPT-5.1 Pro remain high; its full capabilities have yet to be revealed. The GPT-5.1 models are expected to become available on November 24, at least for enterprise users, with a possibility of an earlier launch.

On the Google side, fresh traces of Gemini 3 indicate a release in the coming weeks. The Nano Banana 2 model has also been making waves. Recently, it became accessible on a third-party platform, where users have shared numerous sample outputs confirming that it represents a significant upgrade over Nano Banana 1, delivering capabilities that the previous model couldn’t handle. We’ve compiled a large gallery of these new images, so feel free to browse through them.

All in all, the next two weeks promise to be very eventful. It’s fair to call this period “AI Shipvember.” Let’s break down what’s coming next.

Newsletter Guide

Here's your guide on how to make the most out of this newsletter:

🤳 Gear up with your mobile phone and desktop devices.
📩 Scan through the newsletter, picking out the apps that pique your interest.
🔥 Keep an eye on the "hot" emoji to identify the hottest user experiences.
👀 Dive deeper into our TestingCatalog posts to discover how to put these apps to the test.
📲 Roll up your sleeves and try them out yourself!

Anthropic

Apple

Apple will use Google's 1.2T model for Siri and develop its own 1T model.

Baidu

ERNIE-5.0-Preview-1022 achieves a high ranking on LMArena with 1432 points.

Cognition

Windsurf is getting Codemaps for AI-assisted coding – What's new? Codemaps offers AI-generated code maps using SWE-1.5 and Claude Sonnet 4.5 on Windsurf; It integrates with DeepWiki and Cascade for just-in-time code navigation;

Figma

Figma acquires Weavy, launches Figma Weave – What's new? Figma Weave, from Figma's acquisition of Weavy, adds AI media generation and editing; it uses AI models and node-based tools for cinematic video and realistic visuals;

Google

Early look at images generated by Nano Banana 2 – What we know so far: Nano Banana 2 is expected to launch on November 11 with 2k native output and huge improvements across many areas; It is expected to be based on Gemini 3.0 Pro.

Google test annotation and background removal on Mixboard – Google's Mixboard to debut image creation on a canvas with new viewing modes, background removal and annotations for refined visuals.

🔥 Google rolls out Opal AI Agent builder in 160+ countries – What's new? Google rollout Opal, a no-code platform for AI mini-apps now in 160+ countries; adds web deployment and Google Sheets integration;

Google Cloud unveils Ironwood TPU and Axion instances – Google Cloud unveils Ironwood, its seventh-gen TPU for large AI training and inference, with Axion N4A preview and upcoming C4A metal instances.

🔥 Google connects Gemini Deep Research with Workspace data – What's new? Gemini Deep Research merges data from Gmail, Drive and Chat with web content; it is available on desktop with mobile access coming soon;

Google prepares a new Image Agent for Stitch – Google refines Stitch for design teams with Image Agent mode marked by a banana icon, detailed project summary and new export links.

Google plans to release Gemini 3 Pro Preview in November – Gemini 3 Pro in VertexAI code signals a November reveal with a 1M token context window for diverse use cases.

NotebookLM will get custom Video Overview styles soon – NotebookLM will get a custom style feature in its video overview settings, letting creators define visuals via tailored prompts with potential Nano Banana 2 support.

Google is preparing Nano Banana 2 for the upcoming release – Google prepares GEMPIX 2 from the Nano Banana lineup with Gemini UI announcement cards signalling a launch next week.

Google plans to expand Opal with advanced Agent capabilities – Leaked MCP and Agent updates extend Opal's reach by linking Gmail, Calendar, Drive and local files into unified custom workflow creation.

Manus AI

Support for slide templates to follow brand guidelines.

Microsoft

Microsoft prepares Copilot Shopping updates ahead of Black Friday – Microsoft Copilot to get a shopping tab for tracked prices, order history and future prices, plus review insights to power AI commerce.

Microsoft prepares Copilot to get Sora 2 video generation – Microsoft's Copilot update will add video generation via Sora 2 with a new Videos tab and refreshed settings available to all users.

Microsoft to broaden Copilot Portraits with new use cases – Microsoft trials new Copilot Portraits in Labs, using AI avatars to mimic facial cues for job prep, study, and public speaking practice.

Moonshot AI

🔥 Moonshot AI unveils Kimi K2 Thinking with 44.9% achievement on HLE – Moonshot AI unveils Kimi K2, a trillion-parameter model offering advanced reasoning and precise problem-solving that outperforms GPT-4.1.

OpenAI

OpenAI readies GPT-5.1 Thinking model ahead of Gemini 3 Pro – GPT-5.1 Thinking debuts on ChatGPT with refined multi-step reasoning and variant models amid competitive pressures before Gemini 3 Pro.

OpenAI, AWS sign $38bn cloud deal to power OpenAI with GPUs – What's new? OpenAI partners with Amazon Web Services to run AI on advanced EC2 UltraServers with NVIDIA GPUs; models are live on Amazon Bedrock for enterprise use;

You can now test upcoming GPT-5.1 on OpenRouter – OpenAI readies GPT-5.1 models with Reasoning and Pro variants for enterprises, as leaked code hints at a November 24 release date. Try GPT-5.1 already now on OpenRouter.

OpenRouter

🔥 New stealth model Polaris Alpha available with 256k context.

Opera

Opera launches deep research agent for Opera Neon founders – What's new? Opera launches Opera Deep Research Agent in Opera Neon using multi-agent reasoning with GPT & Gemini for reports; founders get fast access while others join waitlist;

Perplexity

🔥 Perplexity starts testing Comet browser for Android in beta – Perplexity starts a limited beta of its Comet Android browser, mirroring its desktop design with dark mode, ad blocker, and tab management.

🔥 Perplexity rleases virtual Try On Avatar for shopping on iOS – Perplexity debuts a virtual try-on tool on its US iOS app that lets users create digital avatars from full-body photos and preview garments.

Comet update centralizes data controls with Privacy Snapshot – Perplexity debuts Privacy Snapshot for Comet, adding a homepage widget that groups privacy settings to block tasks and disable web actions.

Perplexity Assistant will be able to join your meetings soon – Perplexity is preparing to update its email assistant with a possibility to auto-record sessions and send recaps to participants.