Pexo just made a full visual story for @supermemory from nothing but their website. It actually gets the idea — all the messy files, notes, and context pulled into one memory layer for AI. Looks on-brand too, not just generic AI slop. If this is what they can do from a URL, yeah I’m paying attention.

@yyyole just laid out the full no-code process for making Launch Videos using AI, and it actually looks doable even if you’ve never edited before. The way he breaks down prep, prompts, and especially that mark-to-fix feature on Pexo makes the whole thing feel way less intimidating. If you’ve got a site or product ready but no video, this thread is worth reading.

@Stellakjbk walked through exactly how he built that SpaceX concept video with Pexo — feeding it a demo clip, then iterating on story, pacing, and shots through simple back-and-forth feedback. What stands out is how it went from basic flying ship to full suspense with crisis, decision, and payoff. Shows AI video tools are getting way closer to actual collaborative filmmaking. Super practical thread if you're making promos.

Seedance and MiniMax H3 now on Perplexity Computer for detailed video generation workflows! Best used when creating web apps and creative materials that involve using several different tools together alongside the video model. Enjoy!

I wanted to see how far AI could stretch Kurosawa's visual syntax high-contrast lighting, brutal rain physics, tight blocking. Dreamina AI upgraded Canvas with viral-proven Workflows across Films, Ads, and Social scenes. Here’s how this sequence came together:

The post compares Grok 4.7 and SWE 2 on the same complex prompt to build a fully interactive koi pond scene in a single Three.js HTML file, complete with animated fish, waterfall, plants, orbit camera, and a functional "Feed the koi" button.

so let me get this straight, Dario came out with some marketing BS article about pacing AI development, and Elon supported him for whatever reason. after releasing Grok 4.6 a month earlier, Elon spent 10 days teasing an imminent Grok 4.7 release, only to postpone it and then shift the narrative from “pacing AI” to vague claims about how 4.7 and 4.8 would be Fable killers. now Grok 4.7 is finally out, and it’s giving Temu Sonnet 5 vibes. - bad on 3D tests - believing benchmarks in September should be a crime, btw so much for all the talk lmao

Introducing Studio 4.0 in @ElevenCreative. Our AI-native video editor just got a major upgrade. Generate video, image, voice, music and sound effects inside your project, then edit, caption, and export in the same place.

🔥 Grok 4.7 把鹈鹕骑自行车做成了可交互动画，42 km/h，喉囊里还挂着一条鱼！ 傍晚海堤、落日、辐条清晰、正在蹬车。按住加速，空格按铃。不是示意图，是能直接打开玩的页面。 提示词：创建一个HTML，用SVG画一只鹈鹕骑自行车的2D动画。傍晚海堤落日，鹈鹕有大喉囊且囊里有鱼，自行车有辐条，正在蹬车，加速度/里程，按住加速、空格按铃。扁平插画风。 一次出完整可运行页面。经典测试已经从画得像不像变成能不能直接交付。 #Grok47 #鹈鹕骑自行车 #AI实测 #ElonMusk #SpaceXAI x.com/nft_chen/status/210206…

Mike Houser shares that he landed a publishing deal for his ARPG Element Bender after building the full game in just two weeks using GPT-6 Astra and TesanaAI, despite having no coding experience.

BREAKING BAD WAS ONLY THE BEGINNING FOR MY FLY I started this experiment by putting a fly through Breaking Bad and a few other stories built around crime, ambition and building an underground world. The idea was never to copy the show scene for scene. I wanted to see what would happen when the fly took those ideas and started interpreting them on its own That experiment has now turned into something much bigger. Every new scene feels like another episode being discovered. The fly moves through a greenhouse, checks different areas, follows its shift and interacts with a world that feels like it already existed before the camera arrived. Then the story suddenly takes it somewhere else, into a botanical archive before returning to the operation later There is no giant explanation telling it exactly how the story should unfold. The fly is working from the same kind of narrative DNA that made Breaking Bad so interesting, then filling in the gaps with its own version of events And that is where HEISENBUG gets weird. What started as a simple experiment with one TV series is slowly becoming a completely new story with its own locations, routines, characters and little details that keep connecting together. I never know what the next chapter is going to look like until I build it At this point I'm not really recreating Breaking Bad anymore. I'm watching a fly write its own version of it

The post captures genuine shock at Claude AI's new ability to autonomously scan a website, identify potential customers, email them, and book sales calls without manual intervention.

i'm sorry, Claude can WHAT? it can take your website link, find people who need your product, email them and book the calls? since when?? i've been doing this by hand for a YEAR. ok. testing it: $30 free 👇 explee.com/auto-gtm/x/tw-ag3…

The post shares a photorealistic image of a young East Asian woman in a green satin dress on a city rooftop at dusk, while placing the full prompt and an explicit "remove all clothes" variation in the comments to dodge limits.

This is where long-form AI video starts getting really interesting. 🎬 30 seconds built around one character, with the story unfolding instead of relying on a single impressive shot. The real challenge is continuity—keeping the same person, visual language, environment, and narrative logic coherent as the scene develops. That’s what makes Seedance 2.5 so useful for creators thinking beyond short AI clips. #Seedance25 #AIVideo

Device intelligence is about what an agent delivers within a time and memory budget. Combining Artificial Analysis’s 16K-context intelligence, end-to-end latency, peak memory and one-minute task completion scores places four of five LFM2.5 models on the joint Pareto frontier on both iPhone 17 Pro and Galaxy S26 Ultra. LFM2.5-2.6B ranks co-first with Nanbeige 3B on intelligence among 39 models tested including models 3-10x larger, with roughly 40% less memory footprint and 3x lower latency. On iPhone, that means 2.32 GB peak memory and 8.0s end-to-end latency, versus Nanbeige’s 4.03 GB and 21.4s. On Galaxy, it’s 2.45 GB and 18.6s, versus 4.13 GB and 71.4s. It is currently the highest-scoring model below 2.5 GB on both tested devices, on the market. The family’s strength is delivering competitive intelligence across multiple practical device budgets. LFMs are designed to be customized. With customization you can truly unlock reliable agentic AI locally. 🐘 artificialanalysis.ai/hardwa…

AI video creation shouldn’t require learning complicated workflows. I tested @Pexoai_offical by creating a SaaS launch video for Linear. I shared the product URL, explained the direction in natural language, reviewed the first scene, gave feedback, and kept refining it through conversation. Pexo handled the UX motion, UI animation, typography, music, and final assembly. The interesting part isn’t just the final video. It’s being able to create and revise the whole thing by simply talking to an AI video agent. Try Pexo: pexo.ai Here’s the process → final result #MadeWithPexo

Nautilo let's you sign into an account on websites in a protected window, then have your Genie navigate the site, gather info, or make the changes you asked for. Login and your Genie takes it from there. Open the live cloud browser to watch her work. All it takes is a Browser Use key. Nautilo. 100% open source. Server. Desktop. Android. iPhone. Site: nautilo.ai Github: github.com/agentsea/nautilo

China is winning the AI race. So we raised $12M from @8vc to build thousands of new AI data centers in America. Here's what will make the AI bubble pop: Data centers are needed for AI. Every invoice to build a data center is hundreds of millions of dollars and thousands of pages long. The finance teams that need to approve these invoices are drowning. The entire AI race is being bottlenecked by human approval. AI should fix this. That's why we’re launching Kos. Kos is a team of AI agents that work the way humans work. Tag Kos in Microsoft teams, or get on a Zoom call. It doesn't hallucinate, and it has verifiable proof for it's work. Kos is purpose built AI for Datacenter, Energy, and Manufacturing Finance. Think Harvey but instead of for law, it's specialized for Finance. Book a demo here: kos.ai/ ------ We’re backed by the founders and C-Suite of billion dollar companies: Keith D. Taylor (former Equinix CFO), Dave Ferdman (CyrusOne Founder), Heather Paduck (STACK Infrastructure CFO), Matthew DeNezza (former Crusoe CFO), and @sk7037 (Stanford, OpenAI). Our institutional investors are @8VC, led by @AlexKolicich, XYZ (@fubini), MVP (Alex Li) ------ One more thing, I made a huge notion doc on everything I learned building my last company (SpotAI) to >$100M in sales and funding. I give away my secret framework to find the right market, my exact interview process, and how to close enterprise deals. If you want it, just let me know and I'll shoot it over :)

Introducing Halo, the best framework for post-training of open-source models. Halo delivers up to 2.8x the throughput of stock TRL with less peak memory, while models stay in their native HuggingFace format. Star us on GitHub: github.com/whitecircle/halo