Image created with gemini-2.5-flash-image with claude-sonnet-4-5. Image prompt: Photorealistic Times Square at dusk with every billboard showing AI agents and autonomous assistants at work, central massive screen displays elegant humanoid agent with multiple arms multitasking, neon signs reading YOUR AI AGENT WORKS 24/7, vibrant blue and purple lighting, crowds looking up at the spectacle, cinematic depth of field
The jump from “”agents are nowhere close to working”” to “”okay, narrow agents for research and coding work pretty well”” to (very recently) “”general purpose agents are actually useful for a range of tasks”” has been quick enough (less than a year) so that most people have missed it.”” / X https://x.com/emollick/status/1972141975458796020
Sonnet 4.5 is the most important coding model release in a while. From our early-access evals, we estimate it’s roughly the same jump in capabilities between Claude 3.5 and 4. As a result, Devin is >2x faster and 12% better on our internal benchmarks.”” / X https://x.com/russelljkaplan/status/1972725070083838250
Sonnet 4.5 crushing GPT-5 high on ARC-AGI 2 https://x.com/scaling01/status/1973081750189334587
MCP Pointer is a local tool combining an MCP Server with a Chrome Extension. The extension lets you visually select DOM elements in the browser, and the MCP server makes this textual context available to agentic coding tools through MCP. https://x.com/firt/status/1970504677776044334
Now this is a good way to use MCPs. Chrome MCP is a great tool to debug and automate browser with Cursor. https://x.com/ozgrozer/status/1970548504167616541
🚨 Apple working on MCP support to enable agentic AI on Mac, iPhone, and iPad https://x.com/MCP_Community/status/1970428384024072657
Agentic Commerce https://developers.openai.com/commerce/
We’re introducing the Agentic Commerce Protocol, an open standard co-developed with @Stripe that enables programmatic commerce flows between users, AI agents, and businesses.”” / X https://x.com/OpenAIDevs/status/1972712933080920451
We have three cool announcements today: (1) @OpenAI is launching commerce in ChatGPT. Their new Instant Checkout is powered by @stripe. (2) We’re releasing the Agentic Commerce Protocol, codeveloped by Stripe and OpenAI. (3) @stripe is launching an API for agentic payments,”” / X https://x.com/patrickc/status/1972716417280860391
🚨 Google just dropped an ace 64-page guide on building AI Agents From ADK to AgentOps, Vertex AI Agent Engine to Agentspace, this guide is the clearest path yet from experimentation to scalable production 🔥 Download link (free!) in 🧵 ↓ https://x.com/DataChaz/status/1969844882299859416
Vibe working: Introducing Agent Mode and Office Agent in Microsoft 365 Copilot | Microsoft 365 Blog https://www.microsoft.com/en-us/microsoft-365/blog/2025/09/29/vibe-working-introducing-agent-mode-and-office-agent-in-microsoft-365-copilot/
ChatGPT Pulse — a background agent which delivers updates to you every day on topics of interest:”” / X https://x.com/gdb/status/1971267684609540583
Personalized just-in-time content is here. ChatGPT pulse is effectively a daily newsletter tuned to your interests. Cost of generation will only go down. NotebookLM style audio overview is an obvious next step. Eventually it’ll be a mf-ing YouTube video you can talk to. The https://x.com/bilawalsidhu/status/1971621232589324741
Now in preview: ChatGPT Pulse This is a new experience where ChatGPT can proactively deliver personalized daily updates from your chats, feedback, and connected apps like your calendar. Rolling out to Pro users on mobile today. https://x.com/OpenAI/status/1971259652684878019
Just had an interesting experience with ChatGPT Pulse Last night I asked ChatGPT a question and it gave me an answer that was just so-so This morning the first card in Pulse was a 3x better answer to the same question I’d be really curious how this works under the hood”” / X https://x.com/nbashaw/status/1972335882058473984
AI should do more than just answer questions; it should anticipate your needs and help you reach your goals. That’s what we’re beginning to build, starting with ChatGPT Pulse (rolling out now to Pro, with goal of making it available to everyone over time): https://x.com/fidjissimo/status/1971258542578663829
Today we are launching my favorite feature of ChatGPT so far, called Pulse. It is initially available to Pro subscribers. Pulse works for you overnight, and keeps thinking about your interests, your connected data, your recent chats, and more. Every morning, you get a”” / X https://x.com/sama/status/1971297661748953263
Building agents with the Claude Agent SDK \ Anthropic https://www.anthropic.com/engineering/building-agents-with-the-claude-agent-sdk
Claude Sonnet 4.5 runs autonomously for 30+ hours of coding?! The record for GPT-5-Codex was just 7 hours. What’s Anthropic’s secret sauce? https://x.com/Yuchenj_UW/status/1972708720527425966
Introducing Claude Sonnet 4.5 \ Anthropic https://www.anthropic.com/news/claude-sonnet-4-5
Anthropic asked 7 researchers to opine on the productivity boost they get from Claude Sonnet 4.5. Results: one “”qualitative answer”” +15% +20% +20% +30% +40% +100%; respondent indicated that his workflow is “”now mainly focused on managing multiple agents”” https://x.com/deredleritt3r/status/1972770139297767720
Claude https://claude.ai/new
Claude Sonnet 4.5 knows when it’s being tested https://www.transformernews.ai/p/claude-sonnet-4-5-evaluation-situational-awareness
Claude Sonnet 4.5 System Card https://assets.anthropic.com/m/12f214efcc2f457a/original/
Anthropic’s new Claude 4.5 Sonnet is now the #4 most intelligent model, beats 4.1 Opus, and places Anthropic in the top 3 in the race for frontier intelligence Claude 4.5 Sonnet offers a clear upgrade for Claude 4.1 Opus and Claude 4 Sonnet users, with greater intelligence at https://x.com/ArtificialAnlys/status/1972854742167761204
Anthropic released a TON of updates today: • Sonnet 4.5 (with context awareness) • Claude Code 2.0 (+ new mascot 🦀) • Claude API: context editing + memory tool @mikeyk sat down for a special launch day chat about 4.5 and the @AnthropicAI developer roadmap! (and more: • VS https://x.com/latentspacepod/status/1973017487190139140
Piloting Claude for Chrome \ Anthropic https://www.anthropic.com/news/claude-for-chrome
Anthropic supremacy in the coding category of lmarena Claude 4.5 Sonnet on tied first with Claude 4.1 Opus https://x.com/scaling01/status/1973836516205134135
Claude can now create and use files \ Anthropic https://www.anthropic.com/news/create-files
Models are good enough now to start significantly speeding up researchers across fields. Sonnet 4.5 using the code execution and file creation feature in claude dot ai was able to replicate all the experiments in a published econ paper just from the raw data. https://x.com/alexalbert__/status/1972749073016132018
AI agents are now capable of doing real, if bounded, work. But that work can be very valuable. For example, the new Claude Sonnet 4.5 was able to replicate published economics research from data files & the paper. We need to figure out what to do with it: https://x.com/emollick/status/1972737754363752557
In real-world development scenarios, GLM-4.6 surpasses GLM-4.5 and reaches near-parity with Claude Sonnet 4, while clearly outperforming other open-source baselines. https://x.com/Zai_org/status/1973034644091392002
Anthropic experiments with real-time UI generation on Claude https://www.testingcatalog.com/anthropic-experiments-with-an-agent-for-gereating-ui-on-the-fly/
Subagents in Claude Code work like a coordinated team: one debugs, another tests, another refines. Each becomes an expert at its task, working in sequence to solve the problem at hand. https://x.com/claudeai/status/1971666134492696749
Enabling Claude Code to work more autonomously \ Anthropic https://www.anthropic.com/news/enabling-claude-code-to-work-more-autonomously
Kernel MCP’s search_docs + @AnthropicAI Claude Code = PBJ for vibe coding browser automations 🥜🫐 https://x.com/onkernel/status/1970929983460970825
I say it often, but the fact that Codex & Claude Code use UX that are accessible only to actual coders is a shame. They are genuinely powerful (& fun) tools for people who don’t touch code at all to generate useful (or creative) little applications. The barriers are unnecessary.”” / X https://x.com/emollick/status/1971784128581587048
We’re at an inflection point in AI’s impact on cybersecurity. Claude now outperforms human teams in some cybersecurity competitions, and helps teams discover and fix code vulnerabilities. At the same time, attackers are using AI to expand their operations. https://x.com/AnthropicAI/status/1974199155657748868
Apple’s ‘Veritas’ chatbot is reportedly an employee-only test of Siri’s AI upgrades | The Verge https://www.theverge.com/news/787046/apples-veritas-siri-ai-chatbot
Apple built its own ChatGPT-like app to test out new Siri AI revamp | Mashable https://mashable.com/article/apple-chatgpt-like-app-veritas-siri-ai-voice-assistant
Introducing Alexa+, the next generation of Alexa https://www.aboutamazon.com/news/devices/new-alexa-generative-artificial-intelligence
You can now buy products directly on ChatGPT. Launched with merchants like Etsy; Shopify and more coming soon. Merchants can integrate using the Agentic Commerce Protocol, a new open standard we built with Stripe: https://x.com/gdb/status/1972717815703683218
Cool thing launching today: you can now buy products directly from ChatGPT (starting with Etsy and expanding soon to over a million Shopify merchants). It’s powered by the Agentic Commerce Protocol, an open standard we built with Stripe. Merchants and developers who want Instant”” / X https://x.com/fidjissimo/status/1972707487238467914
ChatGPT already helps millions of people find what to buy. Now it can help them buy it too. We’re introducing Instant Checkout in ChatGPT with @Etsy and @Shopify, and open-sourcing the Agentic Commerce Protocol that powers it, built with @Stripe, so more merchants and developers https://x.com/OpenAI/status/1972708279043367238
Buy it in ChatGPT: Instant Checkout and the Agentic Commerce Protocol | OpenAI https://openai.com/index/buy-it-in-chatgpt/
We wrote about how Agentic purchasing would be the primary way that OpenAI would monetize free users on August 13th on GPT-5 set the stage for monetization https://x.com/SemiAnalysis_/status/1972714269839163901
I mostly buy stuff from ChatGPT now, so I am excited for this new feature!”” / X https://x.com/sama/status/1972993739074523239
Instant Checkout for merchants in ChatGPT https://chatgpt.com/merchants
Wall Street invited 300+ NYC engineers to builds LLMs for HFT and investing Trading is about to permanently change Here are the top 6 demos from the @Cerebral_Valley AI Fintech Hackathon (🧵): https://x.com/AlexReibman/status/1969847901737422955
Fast progress in training AI agents to interact with the world. Training on just 2,541 hours of Minecraft video, Google built an AI that runs on a single GPU & was able to mine diamonds offline (which takes an average of 24,000 clicks). The same approach may work for AI robots. https://x.com/emollick/status/1973385878195044444
AI Mode in Google Search updates: Visual exploration and discovery https://blog.google/products/search/search-ai-updates-september-2025/
AI as a research partner: Advancing theoretical computer science with AlphaEvolve https://research.google/blog/ai-as-a-research-partner-advancing-theoretical-computer-science-with-alphaevolve/
Google literally dropped an ace 64-page guide on building AI Agents: https://x.com/Meer_AIIT/status/1970889941417898384
You can now Vibe Code voice AI agents and experiences in @GoogleAIStudio for free, with just a prompt 🔊 Earlier this week we announced our updated Gemini Live model, which excels at natural conversations. So easy to get started building! ai dot studio / build https://x.com/OfficialLoganK/status/1972013919683358966
Excited to be expanding access for this. Agentic capabilities in AI Mode for finding restaurant reservations are now available to all users opted into Labs in the U.S. Try it for yourself today → https://x.com/rmstein/status/1971617040193724661
Agentic capabilities in AI Mode – Search Labs
https://labs.google.com/search/experiment/43
Towards an AI-Augmented Textbook https://arxiv.org/pdf/2509.13348
👋 The all new GLM 4.6 from @Zai_org is available on OpenRouter. – GLM-4.6 achieves comprehensive enhancements across multiple domains, including real-world coding, long-context processing, reasoning, searching, writing, and agentic applications. – Context length: increased https://x.com/OpenRouterAI/status/1973037695774384352
Introducing GLM-4.6: Advanced Agentic, Reasoning and Coding Capabilities As our new flagship model, GLM-4.6 brings significant advancements across real-world coding, long-context processing (up to 200K tokens), reasoning, search, writing, and agentic applications. API: https://x.com/Zai_org/status/1973034639708344767
🔎 @Zai_org just released GLM-4.6, and according to Zhihu contributor toyama nao, it’s not chasing GPT-style fireworks — it’s optimizing for real-world agent use. ⚙️ Key Gains (Fig shown below): • Token efficiency = massive win → Reasoning model: 16K → 9K tokens 🔻 https://x.com/ZhihuFrontier/status/1973447038818762841
GLM-4.6 from @Zai_org is available Cline. 200K context (up from 131k), and completes tasks with 15% fewer tokens than GLM-4.5. >48.6% win rate against frontier models, making it one of the most capable open-source models. Available in Cline & the GLM subscription. https://x.com/cline/status/1973099598903386227
The model releases continue… we’ve just released support for GLM-4.6 under the @Zai_org provider! This improves upon the popular GLM-4.5 model, extending the context window from 128k to 200k and achieving higher scores on benchmarks. Try it out today!”” / X https://x.com/roo_code/status/1973022454298837294
GLM-4.6 实测! 最近都很流行更新一大截然后版本就增加0.1嘛? 直接说结论, 前端能力非常好, 效果甚至达到了炫技的程度, 不过随之而来的是复杂场景可能面临性能问题, 所以需要注意防止模型过度发挥或者偶尔留意下代码设计和架构. 不过 python 能力没看到太多变化, 考虑到 python 就一个测试, https://x.com/karminski3/status/1973353334796140716
GLM 4.6 runs quite fast on an M3 Ultra with mlx-lm even at higher precision. Pretty remarkable that it benchmarks competitive to the just-released Sonnet 4.5. Hope those benchmarks hold-up in day-to-day use. Here’s a run using 5.5 bpw quantized model, generating 5.3k tokens at https://x.com/awnihannun/status/1973063906341114327
GLM-4.6 is out on Hugging Face Longer context window: The context window has been expanded from 128K to 200K tokens, enabling the model to handle more complex agentic tasks Superior coding performance: The model achieves higher scores on code benchmarks and demonstrates better https://x.com/_akhaliq/status/1973068098539593932
Cursor can now control your browser. Agent can take screenshots, improve UI, and debug client issues. Try our early preview with Sonnet 4.5. https://x.com/cursor_ai/status/1972778817854067188
My favorite thing about the new college grads that I’ve hired is that they don’t ask you how to do stuff They just put it in ChatGPT and fucking try even if wrong So much better than new grads asking you how to do shit without trying Chat is breeding agency into kids”” / X https://x.com/dylan522p/status/1971425552902082941
Wow – Qwen3-Coder-30b AWQ (4bit) on a single 3090, 115 tokens per second. It just zero-shat Pac-Man. It’s no GLM4.5-Air – but, it runs on a single 3090! https://x.com/QuixiAI/status/1971427136977453184
Former OpenAI and DeepMind researchers raise whopping $300M seed to automate science | TechCrunch https://techcrunch.com/2025/09/30/former-openai-and-deepmind-researchers-raise-whopping-300m-seed-to-automate-science/
AI Agents get smarter with memory—learning, adapting and improving over time. Want to build self-improving agents? Our latest AI Agent for Beginners lesson breaks it all down. ⬇️ https://x.com/msdev/status/1970896017072566646
We’ve updated function calling to support files and images as tool call outputs. You can now call functions like `generate_chart` or `load_image` and return those files back to the model, rather than just JSON or text. 🌠 https://x.com/OpenAIDevs/status/1971618905941856495
Terence Tao + AI for solving hard math problems🤯 In this example the insight comes from Terence, and the muscle—via an hourlong conversation with GPT-5 and some python code it wrote—comes from AI. “”Here, the AI tool use was a significant time saver… Indeed I would have been”” / X https://x.com/kevinweil/status/1974161952260624459
Terence Tao: “I was able to use an extended …” – Mathstodon
https://mathstodon.xyz/@tao/115306424727150237
Yet more evidence that a pretty major shift is happening, this time by Scott Aaronson https://x.com/SebastienBubeck/status/1972368891239375078
A 100% open-source alternative to n8n! Sim is a drag-and-drop open-source platform to build and deploy Agentic workflows. – Runs 100% locally – Works with any local LLM I used it to build a finance assistance app & connected it to Telegram in minutes. The workflow is simple: https://x.com/_avichawla/status/1972912060230164732
One aspect of agents that needs to be considered more is context window contamination. As one agent carries out work, its ability to be “unbiased” decreases as its context window fills with its previous work & decisions. Sometimes analysis requires a fresh agent with a new prompt”” / X https://x.com/emollick/status/1972405830768267668
The most powerful coding model, right at the core of Replit Agent 3. Just rolled out to all our users! ↓🧵 https://x.com/pirroh/status/1972805643691266111
N8n AI Agents are INSANE!! #n8n https://x.com/NetworkChuck/status/1970166755633893668
Building with Cursor (public) https://www.notion.so/cursorai/Building-with-Cursor-public-273da74ef0458051bf22e86a1a0a5c7d
The workflow-for-building-workflows: https://x.com/cline/status/1971436086217122213
How I stopped repeating myself to Cline – Cline Blog
https://cline.bot/blog/how-i-stopped-repeating-myself-to-cline
Cline + LM Studio: the local coding stack with Qwen3 Coder 30B – Cline Blog
https://cline.bot/blog/local-models
Cursor 1.7 is now available! As you type a prompt, suggestions now appear. Press Tab to accept. Also new: custom hooks, deeplinks, team-wide rules, menubar support, and more. https://x.com/cursor_ai/status/1973093532282851837
Introducing Lovable Cloud & AI, a new chapter for vibe coding. Anyone can now build apps with complex AI and backend functionality, just by prompting. 100k+ new ideas, tools, and sites are built on Lovable daily. Today, we’re redefining what’s possible: https://x.com/lovable_dev/status/1972680165378650391
Use exa-code: Fast, efficient web context for coding agents | Exa Blog https://exa.ai/blog/exa-code
VibeGame: Exploring Vibe Coding Games https://huggingface.co/blog/vibegame
Cognition | Rebuilding Devin for Claude Sonnet 4.5: Lessons and Challenges https://cognition.ai/blog/devin-sonnet-4-5-lessons-and-challenges
We rebuilt Devin for Claude Sonnet 4.5. Available starting today as an Agent Preview that’s over 2x faster and 12% better on our Jr. Developer Evals. https://x.com/cognition/status/1972709846937157704
Model Context Protocol (MCP) might sound abstract, but Damian Brady, @damovisa, breaks down why it matters for developers. Hint: it’s about giving tools like GitHub Copilot the power to actually do things, not just suggest code. Learn more: https://x.com/docsmsft/status/1970880434515943826
Announcing our public preview of Chrome DevTools MCP! Experience the full power of DevTools in your AI coding agent → https://x.com/ChromiumDev/status/1970505063064825994
We’re open-sourcing Scribe, an MCP-based system that lets agents actually run notebook cells and receive Jupyter outputs – text, errors, images, etc. This is the agent scaffolding that gave us the results above! (6/7)”” / X https://x.com/GoodfireAI/status/1973789154174754877
I’ve read the docs of new Laravel MCP: https://x.com/PovilasKorop/status/1969699384943071715
New episode of the VS Code Insiders Podcast is out, and we’re diving deep into the @github MCP Registry! https://x.com/code/status/1972800559489876276
This might be the coolest MCP integration I’ve seen: ableton-mcp It exposes tools for controlling playback, changing the tempo, writing midi clips, renaming tracks (use an LLM to help you organize and rename), searching samples, chosing an instrument etc For live jamming as a https://x.com/nsthorat/status/1970281878126239928
Figma MCP server, now with design context anywhere you work → Remote access → Connection with Figma Make → New Code Connect UI components https://x.com/figma/status/1970532828166398328
The “”90% of code written by AI”” thing is increasingly true for very good independent coders, not just a thing the labs say about the future. (The remaining 10% is still very important and very human, though)”” / X https://x.com/emollick/status/1972710280137257214
Today, we are launching “Help Me Choose” in @yupp_ai – a new product feature where multiple AIs critique each other and debate among themselves to help users synthesize diverse perspectives and get the best answer out of their own “AI council”. https://x.com/lintool/status/1973874173157257485
What do Superhuman, Baseten, and Ricky Bobby all have in common? An obsession with speed. If you’re a Superhuman user, you know their email app lives and dies by how fast their users can get through all things email. https://x.com/basetenco/status/1971683977242259623
The sudden viability of economically valuable AI agents has been surprising. This partially as this has happened very quickly & very recently. But also because we had bad assumptions about the limits of agents, which are more self-correcting than expected https://x.com/emollick/status/1973160975290544469
Learning just leveled up: eligible college/uni students can claim extra access to Copilot Podcasts, Deep Research, and Vision w/a Microsoft 365 Personal subscription, free for 12 months. 1 year, 1 subscription, countless lightbulb moments Claim by 10/31: https://x.com/mustafasuleyman/status/1973432724120559823
I accidentally ran Codex in a totally unrelated repository again. It found the correct repository from the error message alone (no GitHub link was given), found the source code, cloned it, and started fixing the issue. The commit was 15.3 thousand lines of code”” / X https://x.com/Sauers_/status/1970727099162861788
Today, @ekindogus and I are excited to introduce @periodiclabs. Our goal is to create an AI scientist. Science works by conjecturing how the world might be, running experiments, and learning from the results. Intelligence is necessary, but not sufficient. New knowledge is https://x.com/LiamFedus/status/1973055380193431965
We’re leading a $300M founding round for @periodiclabs. Science today is bottlenecked by slow, manual experimentation. Periodic is changing that. They’re building AI scientists paired with autonomous laboratories that can hypothesize, experiment, and iterate at speeds impossible https://x.com/a16z/status/1973057267286098131
Periodic Labs Today, we introduce Periodic Labs. Our goal is to create an AI scientist.
https://periodic.com/
Designing agentic loops https://simonwillison.net/2025/Sep/30/designing-agentic-loops/
Real AI Agents and Real Work – by Ethan Mollick https://www.oneusefulthing.org/p/real-ai-agents-and-real-work
FYI we’ve rebranded the Claude Code SDK to the Claude Agent SDK to reflect that it’s the best way to build any general purpose agent, not just coding agents. Especially now when paired with Sonnet 4.5.”” / X https://x.com/alexalbert__/status/1972718342197981194
LangChain just showed how you can turn Claude Code into a domain specific AI agent. Claude Code + highly-condensed context + MCP tool to access more information as needed. https://x.com/Saboo_Shubham_/status/1969449238708064552
We just shipped a bunch of updates to Claude Code, making it a more capable agent than ever. And on top of it: it’s now powered by Claude Sonnet 4.5.”” / X https://x.com/_catwu/status/1972711105157054772
new claude code vscode extension is pretty slick https://x.com/gallabytes/status/1972805892610617466
Live in Cline: Claude Sonnet 4.5 What’s different from Sonnet 4: > more terse, less narration > maintains state across sessions via context files > enhanced capability across long tasks 200k or 1M context window & the same $3/$15 pricing (increased > 200k tokens) https://x.com/cline/status/1972708023232852309
We also shipped a ton of new Claude Code features today, including: – checkpointing – a new VSCode extension – tab to think – a new mascot, we call him Clawd 🦀 https://x.com/trq212/status/1972784970054893877
Claude Sonnet 4.5 vs 4.0 (measured on coding tasks): App from scratch → 4.5: ~20min, functional build. → 4.0: 40+min, lower quality. Simple code change → 4.5: 3 tool calls, correct edit. → 4.0: longer chain, same result. Complex refactor → 4.5: faster + tests passed. → https://x.com/augmentcode/status/1973431992097308983
Claude Sonnet 4.5 continues the tradition of Claude verbal cleverness. For fun, I gave it this very random prompt: “Mash these up into a fine paste: [I quote the final line of 100 Years of Solitude] And: 10 PRINT “HELLO WORLD” 20 GOTO 10” Lots of smart bits in the answer. https://x.com/emollick/status/1972839189181006114
Managing context on the Claude Developer Platform \ Anthropic https://www.anthropic.com/news/context-management
Imagine with Claude https://claude.ai/imagine/
My favorite thing about Claude Sonnet 4.5 so far is how aware it is of its context. When coding and seeing it go “”woow that’s 66k tokens”” it’s very funny. I’ve never seen a model do that before. I wonder if it has any impact on performance.”” / X https://x.com/skirano/status/1973026387528458451
Claude Sonnet 4.5 Is A Very Good Model – by Zvi Mowshowitz https://thezvi.substack.com/p/claude-sonnet-45-is-a-very-good-model
Download Claude https://claude.ai/download
Automated Behavioral Audit Scores shows improvements across all categories for Claude 4.5 Sonnet less misaligned behaviour, less cooperation on malicious requests, less sycophancy https://x.com/scaling01/status/1972713543775682654
I polled the Claude Code team and everyone now uses Sonnet 4.5 as a their daily driver! We think it’s the strongest all-around coding model and are excited for you to use it. https://x.com/_catwu/status/1973524717899489340
Today vibe coding goes pro. Introducing Bolt v2: → World’s best agents (Claude Code, Codex) → Built in backend (hosting, DB, storage, …) → No error loops, no setup nightmares Now anyone can build without boundaries. https://x.com/boltdotnew/status/1973063093849567591
🅰️ Claude Sonnet 4.5 + LangSmith 🦜🛠️ ICYMI – @AnthropicAI just dropped Claude Sonnet 4.5! LangSmith just shipped both cost-tracking and prompt playground support, allowing you to quickly discover how this powerful new model impacts your agents. Get started with docs below 👇 https://x.com/Hacubu/status/1972811123176186131
We’re running a “Built with Claude Sonnet 4.5” challenge. We want to see the coolest things you can build with 4.5 in the next week. Four winners will receive one year of Claude Max 20x and $1k in Claude API credits. https://x.com/alexalbert__/status/1973071320025014306
Prior to the release of Claude Sonnet 4.5, we conducted a white-box audit of the model, applying interpretability techniques to “read the model’s mind” in order to validate its reliability and alignment. This was the first such audit on a frontier LLM, to our knowledge. (1/15) https://x.com/Jack_W_Lindsey/status/1972732219795153126
Claude Sonnet 4.5 is now available in the OpenHands LLM provider and OpenHands Cloud! If you’re using the CLI, GUI, or are a cloud subscriber just select it at the appropriate dropdown. Overall, seems like a very good model 😃 We’re still working on evals! https://x.com/OpenHandsDev/status/1973485506714652998
Introducing Claude Sonnet 4.5—the best coding model in the world. It’s the strongest model for building complex agents. It’s the best model at using computers. And it shows substantial gains on tests of reasoning and math. https://x.com/claudeai/status/1972706807345725773
so we analyzed millions of diff edits from cline users and apparently GLM-4.6 hits 94.9% success rate vs claude 4.5’s 96.2%. to be clear, diff edits are not the end-all-be-all metric for coding agents. but what’s interesting is three months ago this gap was 5-10 points. open https://x.com/nickbaumann_/status/1973846157886697771
🧵 behind the scenes of our first Claude Code product outside the terminal: @claudeai for VS Code ! https://x.com/nateparrott/status/1972717967415582796
🚨 Leaderboard shakeup in the top slot! Claude Sonnet 4.5 now tied for #1 in the Text Arena, matching Claude Opus 4.1! 🏆 Quick reminder: the Arena rankings are powered by tens of thousands of real human votes which have put @AnthropicAI’s Claude Sonnet 4.5 joins the very top https://x.com/arena/status/1973828836510085385
The Magic of Claude Code https://www.alephic.com/writing/the-magic-of-claude-code
Box tested Claude Sonnet 4.5 for data extraction accuracy with Box AI on 40,000 fields across 1,500+ documents and images across a range of industries and use-cases. Overall, the model performed 4 percentage points better than Sonnet 4, and it saw major gains especially in areas https://x.com/levie/status/1972777693377949830
🚨 New Model Update @AnthropicAI’s Claude Sonnet 4.5 Thinking 32k is now live in the Arena! https://x.com/arena/status/1973491519857549618
GLM-4.6 is now live in Kilo Code, and the economics are wild: 48.6% win rate against Claude Sonnet 4 $0.60/$2.20 per million tokens (vs Claude’s $3/$15) 30% more token efficient 200K context window This changes the math on AI-assisted development. 🧵 https://x.com/kilocode/status/1973396250729877904
anthropic’s thinking campaign is just so damn tasteful… it feels like a warm room. the aesthetic is comforting as hell, i kinda get a sense of claude as ai helping you be you from it. that’s why we’re weaving claude into everything we build. i’ll post more about sonnet 4.5 https://x.com/signulll/status/1973828026761695439
Vibe Check: Claude Sonnet 4.5 https://every.to/vibe-check/vibe-check-claude-sonnet-4-5
Anthropic works on customizable Skills for Claude https://www.testingcatalog.com/anthropic-works-on-customizable-skills-for-claude/
Sonnet 4.5 is now available in Cursor! We’re excited to hear what you think.”” / X https://x.com/cursor_ai/status/1972713190074261949
Sonnet 4.5 is now the primary model in Amp. I used it a lot this weekend while coding, and we saw incremental yet meaningful improvements that make it an obvious upgrade. Based on our eval/integration work and my personal experience: • It Just Works; keep doing what you were https://x.com/sqs/status/1972748958813626829
Yes, we built an MCP server for Spotify during the podcast recording 🤯 @ShrekOverflow shows us how to vibe code while breaking down the mysteries of OAuth Turns out, OAuth doesn’t have to be scary when you’ve got AI agents building with you! https://x.com/positiveblue2/status/1970906422998450356
🚨 Leaderboard Update: we have a four-way tie for #1 in the Arena! 🏆 The very top tier is now tied across the strongest models in the world: 🏆 Claude Sonnet 4.5 32k Thinking 🏆 Claude Sonnet 4.5 standard 🏆 Claude Opus 4.1 🏆 Gemini 2.5 Pro All separated by just a few Arena https://x.com/arena/status/1974215622474293262
My review of Sonnet 4.5 based on ~30 hours of Claude Code use is that it’s basically the same as Opus 4.1. Which is quite good! But not as good as GPT-5 (codex thinking=high). Claude Code is still much more polished than Codex. But I find GPT-5 much stronger as a model.”” / X https://x.com/finbarrtimbers/status/1973922679418974298
I have an enterprise plan for ChatGPT ($35/month?) and a Claude 20x Max plan ($200/month). I get equal value from them. The ChatGPT plan is high value for the price. Claude Max is not. I still get >$200 of value from it, but ChatGPT is way better utility/$.”” / X https://x.com/finbarrtimbers/status/1973923264524398687
Sonnet 4.5 was just released – it takes the #1 spot on our finance and programming benchmarks! We had the chance to test the new @claudeai model ahead of launch and found it to be exceptionally capable, especially for agentic applications. (1/6) https://x.com/_valsai/status/1972707249748582454
Announcing @ChromeDevTools MCP! 🚀 Connect your AI coding agent to Chrome’s powerful automation & debugging capabilities with ease. Key features: ✅ Reliable automation: It can programmatically handle clicks, form fills, dialogs, and page navigation with ease. ✅ Performance https://x.com/addyosmani/status/1970503277621256263
Now sounds like OpenAI Codex might be is beating out Claude Code.. Spent some time reverse engineering Codex CLI. Let’s see what decisions @OpenAI made that @AnthropicAI didn’t TAKEAWAY: The model already understands shell. Just use that It even simpler than Claude Code. A https://x.com/imjaredz/status/1973035370041532685
Today we’re expanding Microsoft 365 Copilot with the addition of Anthropic’s Claude models. Customers can now use both OpenAI and Claude — starting in Researcher and Copilot Studio, and coming to more experiences soon. Our multi-model approach goes beyond choice. It’s all about https://x.com/satyanadella/status/1970884338993778855
Claude Sonnet 4.5 and 4.5 Thinking are now available for Perplexity Pro and Max subscribers. https://x.com/perplexity_ai/status/1972751588629545329
Turn design into code with Claude Code + @figma. Through MCP, Claude sees your mockup at the data level—component hierarchies, design tokens, auto-layout rules—and translates it into production-ready code. https://x.com/claudeai/status/1970541285615264071
We’ve focused on improving Claude’s skills in defensive cybersecurity. The results of this are visible in Claude Sonnet 4.5, which is comparable or superior to Opus 4.1 in cybersecurity tasks—yet both faster and cheaper. Read more: https://x.com/AnthropicAI/status/1974199158929305738
Grok Code Fast is right up there. Higher diff edit success rate than Claude 4.5 and GPT-5 Codex and much much cheaper if I may add. Try it out, don’t stop using it and keep sending all that feedback our way. It’s only going to get better.”” / X https://x.com/gauravisnotme/status/1974001009778115066
Sonnet 4.5 ranking #1 on Deep Research Bench https://x.com/scaling01/status/1973088829138460987
New on the Anthropic Engineering Blog: Most developers have heard of prompt engineering. But to get the most out of AI agents, you need context engineering. We explain how it works: https://x.com/AnthropicAI/status/1973098580060631341
I gave ChatGPT Codex an article & screenshot from a famous, lost Maxis simulation, SimRefinery, and asked it to create it for me. It built a playable prototype for me while I did other stuff, never touching any code once, instead just occasionally poking Codex for small changes. https://x.com/emollick/status/1972043547550666796
Voice Assist in Action | Try Our AI Voice Assistant | CallRail https://www.callrail.com/voice-assist-in-action
Veo3 high quality on Freepik 😀. Again I use Grok4 +Kimi K2 to write my prompts if you are not using them you are missing big time 👀 Veo3 on Flow refuses to generate the prompt for violating the safety, while the API and Gemini App did not have the problem 🤔🤔🤔 { https://x.com/IamEmily2050/status/1951248894035402973
We’re launching a new feature at Yupp: Help Me Choose. Don’t know which AI’s response is better? Help Me Choose lets a third AI give its own analysis. Then, the two models that answered you analyze each others’ responses – and their own. Final step? You pick your favorite! https://x.com/yupp_ai/status/1973882910907470237
Most companies aren’t seeing a return on AI investments. This tech CEO wants to change that | CNN Business https://www.cnn.com/2025/09/30/tech/scale-ai-making-money-meta
IBM is back! They just joined Hugging Face Enterprise & released Granite 4.0 in open-source with a new hybrid Mamba/transformer architecture that reduces memory requirements without reducing accuracy much. This set of models is great for agentic workflows like tool calling, https://x.com/ClementDelangue/status/1973798540389355903
AMD is using Cline as their coding agent for local models. After testing 20+ models, they found what actually works: > 32GB RAM: Qwen3-Coder 30B (4-bit) > 64GB RAM: Qwen3-Coder 30B (8-bit) > 128GB+ RAM: GLM-4.5-Air 10-minute setup with @lmstudio + Cline, linked below”” / X https://x.com/cline/status/1973035211379310708
Some relevant details about the RL phase for DeepSeek’s latest models: – finetune separate specialist models with RL for mathematics, competitive programming, general logical reasoning, agentic coding, and agentic search, which are used to generate data for final training – https://x.com/iScienceLuvr/status/1972627894636814585
DeepSeek-V3.1-Terminus launches with improved agentic tool use and reduced language mixing errors | VentureBeat https://venturebeat.com/ai/deepseek-v3-1-terminus-launches-with-improved-agentic-tool-use-and-reduced
DeepSeek has launched V3.2 Exp with their new DeepSeek Sparse Attention (DSA) architecture that claims to reduce the impact of the quadratic scaling of compute with context length We’ve independently benchmarked V3.2 Exp as achieving similar intelligence to DeepSeek V3.1 https://x.com/ArtificialAnlys/status/1973230103854456993
Getting ready to try DeepSeek-V3.2-Exp from @deepseek_ai ? vLLM is here to help! We have verified that it works on H200 machines, and many other hardwares thanks to the hardware plugin mechanism. Check out the recipes https://x.com/vllm_project/status/1972664010702221399
DeepSeek-V3.2 shows: – Chinese chips are rising: Day-0 support for Huawei Ascend & Cambricon; – ML compiler: DeepSeek uses TileLang, letting you write Python → compile to optimized kernels on diverse hardware. E.g., 80 lines of Python can reach 95% of FlashMLA’s (CUDA written https://x.com/Yuchenj_UW/status/1972698599554490856
How does @deepseek_ai Sparse Attention (DSA) work? It has 2 components: the Lightning Indexer and Sparse Multi-Latent Attention (MLA). The indexer keeps a small key cache of 128 per token (vs. 512 for MLA). It scores incoming queries. The top-2048 tokens to pass to Sparse MLA. https://x.com/vllm_project/status/1972617272901644345
As @_xjdr says this is not yet profiled fully but we can provisionally accept that DSA had just enabled 1M contexts for ≈everyone. Some “”prototype””, hah. They always underhype their work. Btw, you can drop “”sparse””. Just DeepSeek Attention, it’s cleaner. https://x.com/teortaxesTex/status/1973209017603465324
DeepSeek V3.2 breakdown 1. Sparse attention via lightning indexer + top_k attention 2. Uses V3.1 Terminus + 1T continued pretraining tokens 3. 5 specialized models (coding, math etc) via RL then distillation for final ckpt 4. GRPO. Reward functions for length penalty, language https://x.com/danielhanchen/status/1972613546119991791
Some notes on DeepSeek’s v3.2’s sparse attention mechanism DSA can be thought of as a noncontiguous sliding window where each token only attends to 2048 other tokens. This means that both memory to load and FLOPs is constant at O(2048) (for decode at least). https://x.com/nrehiew_/status/1973193918662713510
💻 API Update 🎉 Lower costs, same access! 💰 DeepSeek API prices drop 50%+, effective immediately. 🔹 For comparison testing, V3.1-Terminus remains available via a temporary API until Oct 15th, 2025, 15:59 (UTC Time). Details: https://x.com/deepseek_ai/status/1972604777474458033
Official release of DeepSeek-V3.2-Exp with DeepSeek Sparse Attention + massive price cuts! DeepSeek Sparse Attention (DSA) makes inference cheaper (especially long-context) by learning which past tokens matter for each new token and running full attention only on those. DSA https://x.com/scaling01/status/1972613330209485282
🚀 Introducing DeepSeek-V3.2-Exp — our latest experimental model! ✨ Built on V3.1-Terminus, it debuts DeepSeek Sparse Attention(DSA) for faster, more efficient training & inference on long context. 👉 Now live on App, Web, and API. 💰 API prices cut by 50%+! 1/n”” / X https://x.com/deepseek_ai/status/1972604768309871061
🛠 Open Source Release 🔗 Model: https://x.com/deepseek_ai/status/1972604780469239843
DeepSeek-V3.2-Exp/DeepSeek_V3_2.pdf at main · deepseek-ai/DeepSeek-V3.2-Exp https://github.com/deepseek-ai/DeepSeek-V3.2-Exp/blob/main/DeepSeek_V3_2.pdf
deepseek-ai/DeepSeek-V3.2-Exp · Hugging Face https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Exp
Ollama’s cloud models just got an update! ollama run kimi-k2:1t-cloud Welcome Kimi K2 by @Kimi_Moonshot for the 1 trillion parameter model! ollama run deepseek-v3.1:671b-cloud DeepSeek V3.1 Terminus by @deepseek_ai for the updated updated that is better in language”” / X https://x.com/ollama/status/1971750071483167010
Thank you @NVIDIADC for supporting vLLM for @deepseek_ai model launch. Blackwell is now the go to release platform for new MoEs, and we cannot do it without the amazing team from NVIDIA.”” / X https://x.com/vllm_project/status/1973238773090754735
(1/6) It was a huge week of launches for @julesagent, our AI coding agent! 🐙 New updates like File selector and Memory make Jules smarter and more personal, and other updates like Jules Tools and the experimental API make it easier to integrate into your workflow, helping you”” / X https://x.com/GoogleLabs/status/1974206675859984531
Big updates for @julesagent, the async coding agent that works wherever you do. 💻 Integrate Jules into your local development loop with Jules Tools 🔌 Build on top of Jules with the new Public API for custom workflows 🔒 More control and trust with simpler environment setup and https://x.com/googledevs/status/1974217899536474565
Meet Jules Tools: A Command Line Companion for Google’s Async Coding Agent – Google Developers Blog https://developers.googleblog.com/en/meet-jules-tools-a-command-line-companion-for-googles-async-coding-agent/
Google ADK 🤝 AG-UI Excited to welcome ADK into the AG-UI ecosystem. Bring your ADK agents into Fullstack applications and allow them to collaborate with users in real time. Includes: – Agent UI (headless UI or pre-built components) – State synchronization (between agent <> https://x.com/CopilotKit/status/1971257034407543125
Jules just cd-ed into your terminal. 🤖 Introducing Jules Tools, a new lightweight command line interface that lets you do everything you can do in the browser, available now in your terminal. You can get it via npm install -g @google/jules Get the details in the blog → https://x.com/julesagent/status/1973812188977508755
And of course, the documentation for getting started: https://x.com/julesagent/status/1974179029726159274
Saving the best for last 😊. Integrate Jules’s capabilities directly into your own systems, applications, and CI/CD pipelines. The API is designed to make Jules a programmable member of your engineering team. Read the blog for more → https://x.com/julesagent/status/1974178592683954252
We’re not done yet. 😉 Here’s a recap of everything we’ve launched for Jules this week, plus a little hint about what’s coming tomorrow. Read more here: https://x.com/julesagent/status/1973898898067632212
YouWare now supports GLM-4.6 @Zai_org Clear gains vs GLM-4.5: +27% in coding performance Lower credit usage Try it now. https://x.com/YouWareAI/status/1972990785428811923
📢 New Model Drop: GLM 4.6 is now on Yupp! This new model from @Zai_org offers enhanced coding, long-context processing, reasoning, and writing. We tried it out with some prompts: https://x.com/yupp_ai/status/1972994220144427266
We released a patch this week (v3.32.3) which includes GLM-4.6. The model hits 94.9% success rate on diff edits in our testing. That’s within 1.3 points of Sonnet 4.5 at 10% the cost.”” / X https://x.com/cline/status/1973870619013136850
I’ve been playing with the new Excel agent and it seems like Microsoft is taking a leap past copilots It feels still in development, but it does autonomous Excel work much better than Microsoft’s Copilot approach, which it effectively kills (with unclear implications for work)”” / X https://x.com/emollick/status/1972902686942212586
You can now talk to a Copilot Portrait in real-time! We’ve heard from some users that they’d feel more comfortable talking to a face when using voice. So in the US, UK + Canada, we’re rolling out a new Copilot Labs experiment where you can talk with animated portraits. https://x.com/mustafasuleyman/status/1972740299840782684
Updated TLDR (after hammering Sonnet 4.5 for 24h); GPT-5-Codex is good and fast Sonnet 4.5 is great and fast GPT-5-High is great and slow Opus 4.1 is no longer needed”” / X https://x.com/iannuttall/status/1973460871574856019
You can just run qwen3-coder on a macbook w/ @lmstudio”” / X https://x.com/cline/status/1971591597080064121
Cool post on integrating Qwen3 Coder + Cline + LM Studio for high-quality local coding: https://x.com/awnihannun/status/1971603427131351218
Why is everyone hyping legged humanoids when wheels move faster, cheaper, and longer? While billions flow into bipedal humanoid startups, researchers are exploring hybrids: ❗️humanoids that use wheels for transport and legs for manipulation. One example: a humanoid riding a https://x.com/IlirAliu_/status/1971995762952151257
Agentic Memory & Context Engineering Hackathon https://cerebralvalley.ai/e/mongoDB-hackathon
ReasoningBank: memory for self-evolving LLM agents • Distills strategies from both successes & failures • Enables agents to learn, reuse, and improve over time • Outperforms prior memory methods on web & SWE tasks (+34.2% eff., –16% steps) https://x.com/arankomatsuzaki/status/1972870229463355677
We tested whether coding agents with CLI tools are all you need for complex document search and analysis tasks 🧪 Our new SemTools benchmark used 1000 @arxiv papers to compare agents with and without semantic search capabilities: 🔍 Agents with semantic search provided more https://x.com/llama_index/status/1973783798044307741
New features. New workflows. A new form of video. (…Did someone say Video Agents? 👀) Check out everything we announced at Synthesia 3.0 – our biggest event ever! 👉 https://x.com/synthesiaIO/status/1973688529818620193
