全部 AI 动态

Just pushed DFlash2 (@inco_ai) recipes to the Qwen3.8 27B cookbook⚡️ https://docs.sglang.io/cookbook/autoregressive/Qwen/Qwen3.8-27B The community has been seeing great results with NVFP4 + DFlash2, and these recipes should be some very good starting points to play with. More Qwen3.8 27B updates on the way 🫡

I Turned Grok Bot Into My CTO. I hand that bot the GitHub repo and tell it to run the show, so it spins up cloud agents off my machine, follows the PRs, and reaches for poteto-mode when a task actually needs the rigor. Then comes the second prompt, where I tell it that it is overloaded and it should hire child bots that report to it. After that the bots talk to each other instead of talking to me. One lands the PRs, one works the Convex backend, one owns auth, and I read the threads view-only. Two honest things before you copy this. It burns tokens. I went past two billion in a single day because I told my bot to use pstack for everything, and the fix was saving pstack for the hard tasks. And it does not replace engineers. It takes me out of the coordinating seat, which is a different claim. Timestamps 00:00 The Grok Bot CTO Workflow 01:23 Direct Responsible Agent Prompt 04:20 pstack Plugin and poteto Mode 08:46 Token Consumption Warning 11:49 Spinning Up a Team of Bots 17:20 Advanced GrokBot Use Cases

This is the easiest way to get started with Grok Build. Just go to https://grok.com/?mode=build, no scripts or download needed.

BREAKING: Muse Spark 1.2 by @AIatMeta takes 1st for Video-to-Website with an Elo rating of 1279 and impressive scores across all of our multimodal code categories. Muse Spark 1.2 also takes 2nd for Image-to-HTML with an Elo rating of 1252 and 3rd for Image-to-Frontend with Elo rating of 1272. At $1.25/1M input tokens and $4.25/1M output tokens, it lands on the price-preference Pareto frontier for all three categories. Congrats to the @Meta team on the achievement!

Grok 4.6 just tied for #1 on the Artificial Analysis Agentic Index • Grok 4.6 (high) — 59 • Claude Opus 5 (max) — 59 Outperforming Claude Fable 5 and GPT-5.6 Sol We’re entering the agentic era, and this is exactly the kind of benchmark that matters so much: tool use, planning, autonomy and complex problem solving Grok 4.6 is now sitting at the very top And that matters even more as Grok powers Grok Build and Grok Bot, where the model has to go beyond answering questions and actually take actions, use tools and complete real work Grok’s agentic capabilities are getting seriously powerful

Goodbye mental load....hellooooo @bot ! I know, I know. Grok Bot seems like it must be the next new thing for tech bros cold plunging in Patagonia vests, but hear me out moms (yes moms because disproportionately, the mental load falls more on moms than dads). Offload that mental load to AI and watch things just get done. Is it perfect? No. Is it amazing? Absolutely. Grok Bot is the new AI assistant from the team at Cursor + SpaceX (they're calling it SpaceXai). It's surprisingly useful and easy for non-technical people to use, and it can do a whole heck of a lot more than Claude and ChatGPT with little to no setup. Here are actual things that Grok Bot did for me this week (almost completely independently): - Looked through back to school emails and made sure I was prepared with everything that needed to be done for multiple kids. Mental load unlock: 100% I would've forgotten to pack a space change of clothes for my kindergartner on her first day. - Found 3 gift cards I somehow forgot to use from the last year. Oh right, I DID buy 2 Gap gift cards from Costco on black friday! Duh. Mental load unlock: Cha-ching! That's $150 in my pocket. - Scheduled multiple appointments for me. I've been putting off my annual physical exam because who has time to call the doctor. Grok Bot looked up options, pinged me with 3 dates/times, then booked it (I just had to confirm Zocdoc on the last step). It created a calendar reminder! Mental load unlock: Moms, we often put off our own health/wellbeing. Finally! AI can help. I also had it schedule a hair appt and massage for later this month. Let's go AI! - Listed a Facebook Marketplace item and is managing the sale. I had purchased a wagon for my kids years ago that's been collecting dust. Grok Bot found the old listing, used the photos, and asked me what $$ I want to list at. It's going to manage the rest (wish me luck here). Mental load unlock: Who has time to sell their old crap? I sure don't. But AI has infinite time. Notice, none of these are work examples. I will do those in a separate thread. All of these things are being done in the background, and I'm only being asked to approve or provide the AI direction. Don't be intimidated! Literally, if you've used WhatsApp or iMessage you can use Grok Bot with NO TRAINING. AMA I'm blown away by this tech. Photo is of my home assistant (her name is Helen) independently ordering my kindergartner a photo from Walgreens that mama somehow forgot about. Oops.

things i use codex browser/chrome/computer for - closing the books our accounting software - clearing my linkedin queue - managing gmail / clearing inbox - setting up stripe radar rules - iterating on images in midjourney - browser based QA on every PR - build inspo library from other websites - completing security questionaires - signing up for procurement systems - research + outreach for vendors who only have crummy online forms - configuring saas when API is not available - registering for conferences - cancelling subscriptions - using iphone mirroring to use mobile apps from my desktop (look ma, no hands!) what else should I be doing?

grok bot is literally the closest to a real life Jarvis it can literally start a business, ship software and make you go viral all at once here's how you do it (in 4 mins) 1. install grok bot and make your chief of staff (tony soprano) 2.give it the 'Digital Chief of Staff Prompt' prompt 3.give it access to X, Stripe and Github 4. you are done just paste the article and give it to your chief of staff no matter what it will be the most productive thing you do this week

I am not a coder. I don’t work with CRM or play with spreadsheets. I’m about as removed from a “workflow” in my job as a kid flying a kite. I don’t really use AI beyond a super-Google level. So when I say Grok Bot is ABSOLUTELY GAME CHANGING, I don’t mean a better search tool. I mean it has fundamentally changed how I go about everything. And it did this in just a few hours. Let me explain. During the install, it asked if I wanted to connect it to several dozen programs. I don’t use a single one of them. So I thought, “Well, I’m not gonna get a lot of use out of this.” But I kept going. Then it asked if it could connect to my Gmail. I said yes. And that’s where the jaw-dropping began. It immediately figured out my life. It saw 40,000 emails and got a pretty good picture of me. So I asked it to do my taxes. It found all the docs needed in my email and did my taxes. I asked it to find first-class award tickets to Japan. It scoured the internet and logged into my airline accounts (with permission) and found 3 tickets to Japan in first class for 60K points. It would have booked them, but I’m still not sure of the dates I need so I stopped it. I created a few agents. One gives me an hourly political update based on key topics I need for work. One is cleaning up my Gmail inbox and organizing it. One is saving me money in my taxes. One makes sure it always checks with me on key decisions (like deleting an email). It’s absolutely insane. Think about it this way: If you could hire an assistant, what would you have them do? Start there, and have fun!

Inpainting using MiniMax H3 😎 https://github.com/drozbay/MaskVidExperiments

I made an AI influencer who got 1,200 followers and 1M views in a week - on less than $200. My stack: - Image gen/editing: @ChatGPT - Script / prompts: @ChatGPT - Sound effects: @ElevenLabs - Talking videos: @MiniMax_AI - Action videos: @imagine - Video gen: @krea_ai + @fal

We're continuing to improve cloud agents in Cursor. They pick up work from events, hold a goal until it's met, and stay on course through long sessions.

Grok Voice Think Fast 2.0 just beat GPT Realtime across both text and voice on VulcanBench Across 200 held-out questions, Grok achieved an insane 99% accuracy in text while also delivering stronger performance when the same questions were spoken And the voice result is what makes this especially impressive Grok preserves almost all of that performance when moving from typed questions to spoken ones, while still completing the full response dramatically faster So Grok is not just getting better at voice It’s becoming a seriously powerful realtime reasoning model that can listen, think and respond incredibly fast

Microsoft's MAI-Image-2.5-Pro debuts at #1 on the Artificial Analysis Image Editing Leaderboard, and takes the #7 spot in Text to Image MAI-Image-2.5-Pro is Microsoft AI's quality-focused image model, launched July 23 in preview on Microsoft Foundry. It joins MAI-Image-2.5 and MAI-Image-2.5-Flash in a family Microsoft positions as covering the quality-speed-cost curve, so builders can pick the point that fits their job. Pro sits at the quality end: Microsoft describes it as its highest-fidelity image model to date, aimed at hero imagery, detailed editing, and precise in-image text rendering. In the Artificial Analysis Image Arena, MAI-Image-2.5-Pro debuts at #1 on the Image Editing Leaderboard, surpassing Reve 2.1, GPT Image 2, and Microsoft's own MAI-Image-2.5. In Text to Image it lands at #7. On Microsoft Foundry, MAI-Image-2.5-Pro is priced per token: $5 per 1M text input tokens, $8 per 1M image input tokens, and $106 per 1M image output tokens, which works out to roughly $108.5 per 1k 1024x1024 images. That compares to $48 per 1k images for MAI-Image-2.5 and $20 per 1k for MAI-Image-2.5-Flash. MAI-Image-2.5-Pro is available in preview on Microsoft Foundry across seven global-standard regions, and can be tried out in the MAI Playground. Congratulations to @MicrosoftAI on the release! See below for comparisons between MAI-Image-2.5-Pro and other leading models in the Artificial Analysis Image Arena 🧵

Someone just built a directory with 170+ ready-to-use Grok Bots. Marketing, sales, productivity, ops… everything. Just copy the prompt → paste into Grok → done. And it’s completely open source. People are building around Grok extremely fast.

Grok 4.6 just took the #1 spot on MedAgentBench....one of the most interesting benchmarks for real-world agentic healthcare tasks and Grok now has two generations sitting in the top 3 • #1 Grok 4.6 — ~95.9% • #2 GPT-5.6 Sol — ~94.7% • #3 Grok 4.5 — ~93.4% MedAgentBench goes far beyond answering medical questions It tests AI agents on 300 clinically derived tasks across 10 categories inside a realistic electronic health-record environment, requiring the model to reason, use tools and actually execute multi-step clinical workflows Grok 4.5 was already one of the strongest medical agents tested. Grok 4.6 just pushed it to another level Grok is rapidly becoming one of the strongest AI model families for agentic healthcare work

We're making Git hosting more reliable, performant, and scalable. This post traces 20 years of Git infrastructure and explains how that history led us to design and operate our Git storage, Origin, as if it were a database. https://cursor.com/blog/git-at-any-scale

OMG. I asked my Chief Engineer on Grok @bot to bring up the feature and it's done. 😳 I can text my @maticrobots now through my Grok @bot—anywhere and anytime.

Grok Bot is going through 90,000 emails in my two gmail accounts and purging the bullshit. Something I've never dared to pursue myself. Good luck in there bud, do whatever you want. Just clean it tf up.

On @harvey's Legal Agent, it is tied with Fable 5 at 11.3pts and is ahead of Kimi K3, Qwen 3.8 Max, and DeepSeek V4.

gemini 3.7 flash vs deepseek v4 pro 0813 vs muse spark 1.2 – on voxel city dioramas three models each built three crossy road-style 3d scenes – a construction site, a nyc intersection, a river with a drawbridge – as single self-contained html files the setup: @nousresearch's hermes agent cli on @openrouter, three.js skills preloaded, identical prompts per scene tasks: 1. construction site – tower crane on a working lift loop, paver laying fresh road, roller compacting it behind 2. nyc crossing – four-way intersection with a traffic light state machine, queuing cars, pedestrians crossing on the walk signal 3. river drawbridge – double-leaf bascule that lifts for tall boats, cars queuing at the barriers, animated water every scene: @threejs r185, box geometry only, a locked 20-color palette, four camera presets, and a day/night mode with bloom. one file, no build step, no assets models: @googledeepmind gemini 3.7 flash, @deepseek_ai v4 pro 0813, @aiatmeta muse spark 1.2 muse and gemini finished every scene in two to three minutes. deepseek took 15 to 41 minutes per scene - build time, all three scenes #1 gemini 3.7 flash – 6m 43s #2 muse spark 1.2 – 7m 20s #3 deepseek v4 pro – 91m 25s - total tokens #1 muse spark 1.2 – 440,279 #2 gemini 3.7 flash – 713,855 #3 deepseek v4 pro – 20,957,568 - total price #1 muse spark 1.2 – $0.53 #2 gemini 3.7 flash – $0.56 #3 deepseek v4 pro – $4.57 - agent calls across the three builds #1 muse spark 1.2 – 12 #2 gemini 3.7 flash – 18 #3 deepseek v4 pro – 143 observations: • muse won two of the three scenes on looks with the smallest files in the test – 887 to 1,042 lines against gemini's 1,934 to 2,377. cheapest, fastest to a good frame, and shortest turned out to be the same column • deepseek burned 20.96m tokens – 29x gemini, 48x muse – across 143 agent calls. prompt caching is the only reason that cost $4.57: the cache discount absorbed roughly $30 of resent context • gemini was the only model whose files needed zero fixes to render – and the only one whose night mode is cosmetic. the sky never darkens and one camera button does nothing. clean code for a scene it never looked at follow @thehypedotnews for 24/7 ai news, analysis and breakdowns