# This Week in All Things AI - Week 36-2025

> Sunday 31st August 2025 to Saturday 6th September 2025

**Published by:** [This Week in All Things AI](https://paragraph.com/@twiata/)
**Published on:** 2025-09-07
**URL:** https://paragraph.com/@twiata/this-week-in-all-things-ai-week-36-2025

## Content

The upcoming week shall have Apple's most anticipated hardware launch of the year. Feel free to share this publication link to those in your family, friends and acquaintances who have a curiosity in knowing more about AI tools/services/use cases which are rapidly evolving. Below is my personal website which aggregates links to many of my socials as well as content and community that I curategoolamabbas | Instagram, TikTok | LinktreeChief Knowledge Officer @ Animoca Brands. Also assists in Growth Marketinghttps://linktr.eeContinuing to use a tabular format to show the digest of posts on days which have had multiple posts to isolate them individually. The following messages were posted on the 'All Things AI' Telegram group from Sunday 31st Aug 2025 to Saturday 6th Sep 2025Sunday 31st August 2025Monday 1st September 2025Props to Ben Chevalier for pointer to the below Nano Banana best practices from the Google team themselves. Philipp Schmid @_philschmid Gemini 2.5 Flash Image (Nano Banana) best practices - Be hyper-specific: The more detail you provide, the more control you have. Instead of "fantasy armor," describe it "ornate elven plate armor, etched with silver leaf patterns, with a high collar and pauldrons shaped 2,040 11:12 PM • Aug 30, 2025 Tuesday 2nd September 2025Wednesday 3rd September 2025For those in the group using Replit or thinking about using ReplitUntil now, using Replit Agent meant choosing between two paths: Start with one of their curated app types. For example, HTML, CSS, React, ShadCN UI, built to work seamlessly with Agent. Or go fully custom—without Agent support.From now on, Replit Agent now works with any framework! Replit ⠕ @Replit Big news: Replit Agent now works with any framework! What's now possible for advance builders: → Import existing projects and get Agent support → While Agent previously supported popular languages, it can now handle even more with Java, Rust, Go, C#, Angular, Vue, 454 2:32 AM • Sep 3, 2025 Amjad Masad @amasad Replit Agent is great at starting projects scratch, but it couldn't run repos or build custom stacks. Announcing General Agent: Import, build & run anything! Watch it run ComfyUI by just adding "repl•new/" to the URL. 747 6:53 AM • Sep 3, 2025 Replit - Agent is now available to build apps with any frameworkWe are excited to announce that Replit Agent is now available to assist in building apps across all development frameworks. This allows more advanced users to be able to build with the Agent with the framework of their choice. Why does this matter?https://blog.replit.comThursday 4th September 20251Perplexity rolling out Comet to all students worldwide. Aravind Srinivas @AravSrinivas Comet is now available for all students. Should be a no brainer to switch from Chrome for the following: - Comet Assistant - Flash Cards - Ad Block - Study Mode Perplexity @perplexity_ai We are rolling out Comet to all students worldwide. Ask Comet to manage your schedule, order textbooks, or prepare for exams with Study Mode. 651 2:12 AM • Sep 4, 2025 https://www.perplexity.ai/grow/comet/students2via Chris Mendez A long-form article from his experience trying out 3 IDE’s Three Faces of AI Coding: Cursor, IntelliJ, and ChatGPTTwo weeks, three AI dev tools: Cursor the reckless junior, IntelliJ the cautious senior, and ChatGPT the strategist. Together they reveal not just code-but how I code.https://www.chrisjmendez.com Friday 5th September 20251Poster: Robby Yung Best use of AI today2Perplexity's agentic browser Comet is coming soon to mobile and is now available for pre-orders on Android Play Store https://play.google.com/store/apps/details?id=ai.perplexity.comet Reply-by: Leeny I still think that ChatGPT 5.0 is doing a good job too 👍 I’m experimenting among perplexity Gemini ChatGPT and grok.3EmbeddingGemma is a 300M parameter, state-of-the-art for its size, open embedding model from Google, built from Gemma 3 (with T5Gemma initialization) and the same research and technology used to create Gemini models. EmbeddingGemma produces vector representations of text, making it well-suited for search and retrieval tasks, including classification, clustering, and semantic similarity search. This model was trained with data in 100+ spoken languages.4TIL (Today I learned) that elements of LottieFiles uses Groq (with a Q) under the hood.Ideation and Animation at Human SpeedThe animator can try new ideas or suggest tweaks, get back the LottieFiles result, and accept or discard it, all in the matter of a couple of seconds. The back and forth can be as fast as the creator wants.https://groq.comLottieFiles: Download Free lightweight animations for website & apps.Effortlessly bring the smallest, free, ready-to-use motion graphics for the web, app, social, and designs. Create, edit, test, collaborate, and ship Lottie animations in no time!https://lottiefiles.comPS. If you haven't tried Groq and Cerebras and experienced their 'speedy gonzales/road runner' level of inference speed via their chat interface then I strongly encourage you to try it outCerebrasCerebras is the go-to platform for fast and effortless AI training. Learn more at cerebras.ai.https://www.cerebras.aiGroq is fast inference for AI buildersThe LPU™ Inference Engine by Groq is a hardware and software platform that delivers exceptional compute speed, quality, and energy efficiency. Groq provides cloud and on-prem solutions at scale for AI applications.https://groq.comDon't ask questions which require websearch, ask questions which can be answered by whatever is in their training data What you want to be seeing is 'tokens per second' which may be rendered on their respective websites as T/sSaturday 6th September 20251Moonshot AI just dropped Kimi K2-0905 - 32B activated parameters and 1T total parameters; - SOTA performance in frontier knowledge, math, and coding among non-thinking models; - Meticulously optimized for agentic tasks; - Upgraded from 128K to 256K to provide better support for complex, long-running tasks; - Supports output speeds of up to 60-100 tokens per second. Kimi.ai @Kimi_Moonshot Kimi K2-0905 update - Enhanced coding capabilities, esp. front-end & tool-calling - Context length extended to 256k tokens - Improved integration with various agent scaffolds (e.g., Claude Code, Roo Code, etc) Weights & code: huggingface.co/moonshotai/Kim… Chat with new Kimi 2,481 11:13 AM • Sep 5, 2025 Philip Kiely @philip_kiely The Kimi team did an incredible job iterating on developer feedback with K2 0905. - Fixed the issues: agents, functions, context - Compounded on strengths: code, style, writing Cheers to the Kimi Moonshot team! Kimi.ai @Kimi_Moonshot Kimi K2-0905 update - Enhanced coding capabilities, esp. front-end & tool-calling - Context length extended to 256k tokens - Improved integration with various agent scaffolds (e.g., Claude Code, Roo Code, etc) Weights & code: huggingface.co/moonshotai/Kim… Chat with new Kimi 22 12:22 PM • Sep 5, 2025 Kimi K2-0905: Moonshot's Latest Open-Source Model is Live in Cline - Cline BlogKimi K2-0905 doubles the context window to 256k tokens and delivers some of the highest reliability tool calling we've seen in an open-source model. Built for coding agents that actually work.https://cline.bot2Qwen-3-Max-Preview released from Alibaba - with over 1 trillion parameters - beats their previous best model Qwen3-235B-A22B-2507. - broader knowledge, better at conversations, agentic tasks & instruction following Below is a quote from their launch X postScaling works — and the official release will surprise you even more. Qwen @Alibaba_Qwen Big news: Introducing Qwen3-Max-Preview (Instruct) — our biggest model yet, with over 1 trillion parameters! Now available via Qwen Chat & Alibaba Cloud API. Benchmarks show it beats our previous best, Qwen3-235B-A22B-2507. Internal tests + early user feedback confirm: 2,645 11:43 PM • Sep 5, 2025 Rohan Paul @rohanpaul_ai Alibaba’s Qwen team just launched its BIGGEST model yet. Qwen3-Max-Preview, a 1T+ parameter model now live in Qwen Chat and the Alibaba Cloud API. Says it beats Qwen3-235B-A22B-2507 on internal benchmarks and early trials, with stronger conversation, instruction following, Qwen @Alibaba_Qwen Big news: Introducing Qwen3-Max-Preview (Instruct) — our biggest model yet, with over 1 trillion parameters! Now available via Qwen Chat & Alibaba Cloud API. Benchmarks show it beats our previous best, Qwen3-235B-A22B-2507. Internal tests + early user feedback confirm: 6 2:35 AM • Sep 6, 2025 OpenRouter @OpenRouterAI Qwen3-Max, @Alibaba_Qwen's most powerful model is live on OpenRouter: Higher accuracy in math, coding, logic, and science tasks Stronger instruction following & reduced hallucinations Optimized for RAG + tool calling (no “thinking” mode) 577 11:46 PM • Sep 5, 2025 3Available now to logged-in chatGPT users on the web By popular request: you can now branch conversations in ChatGPT, letting you more easily explore different directions without losing your original thread. OpenAI @OpenAI By popular request: you can now branch conversations in ChatGPT, letting you more easily explore different directions without losing your original thread. Available now to logged-in users on web. Micah @MicahJanke Let us branch off from chats at specific points without losing the node we branched off from, sometimes I want to persue multiple possible branches to a conversation or sometimes I want to go on like a 1-3 side prompt conversation without it muddying the context of the original 22.8K 4:13 AM • Sep 5, 2025 I am continuing the experiment of participating in the onchain creator/curator economy by tokenizing aka coining each of these posts and pairing it with my Zora creator coin https://zora.co/@ymgabbas If readers find value and want to indicate their support they could do so by clicking/tapping on the 'Support' button at the top of this email/post or they could also support by acquiring my Zora creator coin Ticker is 'YMGABBAS' CA (on Base): 0x4e24c96c0215ad8fff692504910b25470144f89b I welcome feedback not only on this newsletter but across all the various surfaces where I curate content Regards, Yusuf

## Publication Information

- [This Week in All Things AI](https://paragraph.com/@twiata/): Publication homepage
- [All Posts](https://paragraph.com/@twiata/): More posts from this publication
- [RSS Feed](https://api.paragraph.com/blogs/rss/@twiata): Subscribe to updates
- [Twitter](https://twitter.com/yusufg): Follow on Twitter