Wes Roth

12.7K posts

Wes Roth banner
Wes Roth

Wes Roth

@WesRoth

FOLLOWS YOU. Artificial Intelligence, Automation & Optimism. Everything I say is 100% serious...

San Diego, CA Katılım Temmuz 2022
791 Takip Edilen33.2K Takipçiler
Sabitlenmiş Tweet
Wes Roth
Wes Roth@WesRoth·
Apple has quietly halted App Store updates for popular AI "vibe-coding" applications most notably the $9 billion startup Replit and mobile app builder Vibecode. After months of pushback, Apple is reportedly demanding major UX changes. Replit is being asked to force its generated app previews to open in an external web browser rather than natively inside its app. Vibecode was told it must completely remove the ability to generate software specifically for Apple devices.
Wes Roth tweet media
English
58
47
548
156K
Wes Roth retweetledi
Wes Roth
Wes Roth@WesRoth·
OpenArt has launched OpenArt Worlds, a new feature powered by World Labs' spatial AI that allows users to generate and navigate fully persistent 3D environments from a single text prompt or image. Instead of generating a flat 2D image, the AI builds a complete 3D environment that you can virtually "step inside" and walk through. At the exact moment of capture, you can prompt the AI to insert specific characters, objects, or cinematic elements into your carefully framed shot. The final captures can be exported as high-quality 2D images, used as perfectly consistent keyframes for video generation, or eventually exported as 3D meshes for professional software like Unreal Engine.
OpenArt@openart_ai

Today, we’re launching a new way to create with AI. With OpenArt Worlds, you can generate a fully navigable 3D environment from a single prompt or image, step inside it, and capture shots exactly the way you envision them. No more starting over. No more inconsistent scenes. You build the world once - and create inside it. • Move through your scene freely • Find your angles • Add characters and elements • Capture production-ready shots

English
4
15
56
5.5K
Wes Roth retweetledi
Wes Roth
Wes Roth@WesRoth·
Xiaomi has officially released MiMo-V2-Pro, a massive reasoning upgrade over its previous open-weights MiMo-V2-Flash model. While the previous Flash version was open source, MiMo-V2-Pro is currently available exclusively via Xiaomi's first-party API. MiMo-V2-Pro scored a 49 on the Artificial Analysis Intelligence Index, placing it at #10 overall. Built specifically for autonomous workflows, the model achieved a leading Elo of 1426 on the GDPval-AA (Agentic Real-World Work Tasks) benchmark. MiMo-V2-Pro dropped its hallucination rate to just 30% (down from 48% in the Flash version), earning it a competitive +5 on the AA-Omniscience Index. It features a massive 1-million-token context window but is currently restricted to text input and output only (no multimodality).
Wes Roth tweet media
Artificial Analysis@ArtificialAnlys

Xiaomi has released MiMo-V2-Pro, which scores 49 on the Artificial Analysis Intelligence Index, placing it between Kimi K2.5 and GLM-5 @Xiaomi's MiMo-V2-Pro is a new reasoning model and a significant upgrade over their prior open weights release, MiMo-V2-Flash (309B total / 15B active, MIT license), which scores 41 on the Intelligence Index. Xiaomi has not yet released the weights of this model and it is currently only available via Xiaomi's first-party API. Key takeaways: ➤ MiMo-V2-Pro scores 49 on the Artificial Analysis Intelligence Index behind GLM-5 (Reasoning, 50). It is ahead of Kimi K2.5 (Reasoning, 47) and Qwen3.5 397B A17B (Reasoning, 45). On the overall leaderboard, it places #10, just behind GPT-5.2 Codex (xhigh, 49) and ahead of Grok 4.20 Beta (Reasoning, 48) ➤ Leading Elo of 1426 on GDPval-AA (Agentic Real-World Work Tasks), ahead of peer models: On GDPval-AA, MiMo-V2-Pro places ahead of GLM-5 (Reasoning, 1406), Kimi K2.5 (Reasoning, 1283), and Qwen3.5 397B A17B (Reasoning, 1209). GPT-5.4 (xhigh) and Claude Sonnet 4.6 (Adaptive Reasoning, max effort) have an Elo of 1667 and 1633 respectively ➤ Competitive AA-Omniscience Index driven by low hallucination: MiMo-V2-Pro scores +5, ahead of GLM-5 (Reasoning, +2), Kimi K2.5 (Reasoning, -8), and Qwen3.5 397B A17B (Reasoning, -30). For context, Claude Opus 4.6 (Adaptive Reasoning, max effort, +14) and Gemini 3.1 Pro Preview (+33) remain ahead ➤ MiMo-V2-Pro is more token efficient than peers. It used 77M output tokens to run the Artificial Analysis Intelligence Index, significantly less than GLM-5 (Reasoning, 109M) and Kimi K2.5 (Reasoning, 89M) ➤ MiMo-V2-Pro costs $348 to run the Artificial Analysis Intelligence Index at $1/$3 per 1M input/output tokens. This is less expensive than GLM-5 despite scoring only 1 point lower on the Intelligence Index. For comparison, GPT-5.2 (xhigh) cost $2,304 and Claude Opus 4.6 (Adaptive Reasoning, max effort) cost $2,486 Key model information: ➤ Context window: 1M tokens ➤ Pricing: $1/$3 per 1M input/output tokens, for 256K token input and $2/$6 per 1M input/output tokens for 1M token input ➤ Availability: Xiaomi first-party API only ➤ Modality: Text input and output only (no multimodality)

English
7
4
36
3.7K
Wes Roth retweetledi
Wes Roth
Wes Roth@WesRoth·
OpenAI has launched an open research competition called "Parameter Golf" as the first installment of its new Model Craft series. Participants must build the most efficient language model possible to minimize loss on a fixed FineWeb validation dataset. Submissions must adhere to a strict 16 MB limit for the entire artifact (model weights and training code combined) and must train in under 10 minutes on an 8xH100 GPU cluster. The leaderboard is fully automated via GitHub.
OpenAI@OpenAI

Are you up for a challenge? openai.com/parameter-golf

English
3
14
31
3.1K
Wes Roth retweetledi
Wes Roth
Wes Roth@WesRoth·
how Minimax M2.7 was made is absolutely INSANE it "evolved" 100+ times with zero human input They built a research agent using an early version of that same model soon it was handling 30 to 50 percent of their RL team's entire workflow. and then it got WEIRD
English
18
26
404
32.2K
Wes Roth
Wes Roth@WesRoth·
Google launched a massive upgrade to the "Build" mode inside Google AI Studio, turning it from a simple frontend prototyping sandbox into a full-stack application generator. The backend of AI Studio’s app generator is now powered by the "Antigravity" coding agent. This agent maintains deep project context, handles multi-file dependencies, and executes self-correcting logic across the stack. When the agent detects that your app needs persistent data or user accounts, it will automatically provision a Firebase backend. This allows developers to integrate Cloud Firestore (for databases) and Firebase Authentication (for "Sign in with Google") with a single click. The platform now natively supports server-side runtimes, enabling the creation of real-time collaborative workspaces and multiplayer applications directly from a natural language prompt. AI Studio now supports Next.js alongside React and Angular. The agent is also smart enough to proactively install modern UI libraries (like Shadcn or Framer Motion) via npm to polish the frontend aesthetics. Developers can safely close their browser tabs; the environment now saves your exact state so you can pick up the build session right where you left off.
Wes Roth tweet media
Google AI Studio@GoogleAIStudio

x.com/i/article/2034…

English
1
0
5
514
Riley Brown
Riley Brown@rileybrown·
That's it, i'm done with apple. We need to band together. Let's all switch to Android. Who's with me? - sent from my macbook, connected to my iPhone hotspot, while wearing air pods.
Wes Roth@WesRoth

Apple has quietly halted App Store updates for popular AI "vibe-coding" applications most notably the $9 billion startup Replit and mobile app builder Vibecode. After months of pushback, Apple is reportedly demanding major UX changes. Replit is being asked to force its generated app previews to open in an external web browser rather than natively inside its app. Vibecode was told it must completely remove the ability to generate software specifically for Apple devices.

English
28
4
103
17K
Tim Sweeney
Tim Sweeney@TimSweeneyEpic·
Apple needs to bring back Steve Wozniak to fix these idiotic missteps. It’s not only a terrible move against developers, it’s a move against kids learning to make apps and therefore a move against the future of Apple’s own platform.
Wes Roth@WesRoth

Apple has quietly halted App Store updates for popular AI "vibe-coding" applications most notably the $9 billion startup Replit and mobile app builder Vibecode. After months of pushback, Apple is reportedly demanding major UX changes. Replit is being asked to force its generated app previews to open in an external web browser rather than natively inside its app. Vibecode was told it must completely remove the ability to generate software specifically for Apple devices.

English
66
34
388
57K
Wes Roth
Wes Roth@WesRoth·
Xiaomi has detailed the capabilities of MiMo-V2-TTS, its new flagship speech synthesis model, which launched last night alongside its new MiMo-V2-Pro and Omni text/vision models. The company is pitching the system as a massive leap from traditional text-to-speech, focusing on "performance" rather than simply reading words aloud.
Xiaomi MiMo@XiaomiMiMo

MiMo TTS actually reads the text — not just the words. CAPITALIZED? It shouts. A dash — natural pause. Ellipsis... it lingers. No special tags. The model reads formatting the way you do. 🎧 3/n

English
2
0
9
789
Wes Roth retweetledi
Wes Roth
Wes Roth@WesRoth·
Major tech companies and enterprise customers are refusing to cut ties with Anthropic, despite the Pentagon and the Trump administration officially designating the AI startup a national security "supply chain risk." Rather than treating the company as a liability, corporate customers are viewing Anthropic's refusal to compromise its ethical guardrails as a massive trust signal.
Wes Roth tweet media
English
13
11
61
3.3K
Wes Roth retweetledi
Wes Roth
Wes Roth@WesRoth·
Google has officially lifted the veil on the "biggest update ever" for its AI design tool, Stitch. The update transforms the platform from a simple prompt-to-UI generator into a spatial, multimodal, and highly autonomous design environment. The 5 Major Upgrades: AI-Native Infinite Canvas: Stitch is abandoning the traditional flat workspace for a node-based spatial canvas. It can now ingest text, code, product requirement documents (PRDs), and reference images simultaneously. It also introduces an Agent Manager to orchestrate multiple design tasks in parallel (and finally adds a Light Mode). Smarter Design Agent: The AI now possesses full "canvas context." It can mix mobile and desktop screens in the same workspace, instantly swap specific assets (like logos or profile pictures) across multiple screens, and even reverse-engineer a product brief based on the UI you are building. Voice Live Mode (Preview): Designers can now converse directly with the tool. Because Stitch can "see" the canvas and the user's cursor, you can ask for real-time critiques, verbally command it to navigate to specific screens, and queue up multiple concurrent design updates hands-free. Instant Prototypes: By clicking 'Play,' the AI automatically maps the logical order of your screens to create an interactive prototype. It even features an "imagine next screen" function that generates missing user flows on the fly when you click a dead end, complete with shareable URLs and mobile QR codes. DESIGN.md & Design Systems: To solve the massive issue of AI hallucinating inconsistent UI elements, every new project now automatically starts with a unified design system. Google introduced the DESIGN.md file format, allowing developers to easily export design tokens to their codebase or import existing brand guidelines (even extracting them directly from a live URL).
Stitch by Google@stitchbygoogle

Meet the new Stitch, your vibe design partner. Here are 5 major upgrades to help you create, iterate and collaborate: 🎨 AI-Native Canvas 🧠 Smarter Design Agent 🎙️ Voice ⚡️ Instant Prototypes 📐 Design Systems and DESIGN.md Rolling out now. Details and product walkthrough video in 🧵

English
9
10
102
15K
Wes Roth retweetledi
Wes Roth
Wes Roth@WesRoth·
Runway has unveiled a massive breakthrough in AI video generation at NVIDIA's GTC event, showcasing a new model capable of generating high-definition video in genuine real-time. The new model boasts a "time-to-first-frame" of under 100 milliseconds. This means HD video begins generating and playing almost the instant a command is given. This research preview was co-developed with NVIDIA and runs on their newly announced Vera Rubin architecture. This real-time capability is a foundational step for Runway's General World Model (GWM-1). It pushes the technology beyond simple video generation and into the realm of live, interactive environments.
Runway@runwayml

A breakthrough in real-time video generation. As a research preview developed with @NVIDIA and shared at @NVIDIAGTC this week, we trained a new real-time video model running on Vera Rubin. HD videos generate instantly, with time-to-first-frame under 100ms. Unlocking an entirely new creative paradigm and bolstering the foundations of our General World Model, GWM-1. Real-time generation opens a fundamentally different design space for video models and world simulation. We're investing in co-designing our models alongside advances in hardware to keep pushing this frontier.

English
24
94
663
78.8K
Wes Roth retweetledi
Wes Roth
Wes Roth@WesRoth·
Microsoft is considering suing Amazon and OpenAI over their massive $50 billion cloud partnership, arguing that the deal violates Microsoft's exclusive Azure hosting agreement with the AI startup. The looming legal clash hinges entirely on the technical definitions of "stateful" versus "stateless" access. Under the original agreement, Microsoft claims all access to OpenAI's models must be routed exclusively through its Azure cloud infrastructure. Amazon and OpenAI are currently building a "Stateful Runtime Environment" (SRE) on AWS's Bedrock platform to host OpenAI's new "Frontier" product, which requires persistent memory and context. OpenAI argues this does not violate the contract because it isn't providing backdoor access to its "stateless" models or offering them primarily as APIs.
Financial Times@FT

Microsoft weighs legal action over $50bn Amazon-OpenAI cloud deal ft.trib.al/6LZe39E

English
5
6
23
2.8K
Wes Roth retweetledi
Wes Roth
Wes Roth@WesRoth·
Blink New has launched Blink Claw, a fully managed cloud platform designed to let users deploy unlimited, specialized AI "employees" that run continuously 24/7.
Blink.new@blinkdotnew

Introducing Blink Claw - the first platform to hire unlimited AI employees that run your business 24/7. 180+ AI models included. Gmail, Slack, LinkedIn, HubSpot - one-click connect. No API keys, no $600 Mac Mini. Reply "Claw" + RT. Your first agent is on us. ($50 - 200 creds)

English
2
11
56
6.8K
Riley Brown
Riley Brown@rileybrown·
@WesRoth Damn we made it too easy to ship to the App Store… luckily we’re working on an update they can’t restrict. @vibecodeapp_ will prevail
English
6
1
56
6.6K
Wes Roth
Wes Roth@WesRoth·
Apple has quietly halted App Store updates for popular AI "vibe-coding" applications most notably the $9 billion startup Replit and mobile app builder Vibecode. After months of pushback, Apple is reportedly demanding major UX changes. Replit is being asked to force its generated app previews to open in an external web browser rather than natively inside its app. Vibecode was told it must completely remove the ability to generate software specifically for Apple devices.
Wes Roth tweet media
English
58
47
548
156K
Wes Roth
Wes Roth@WesRoth·
Anthropic introduced Channels for Claude Code, a new feature that lets developers control their local AI coding sessions remotely using messaging platforms via the Model Context Protocol (MCP). The feature debuts with official support for Telegram and Discord.
Thariq@trq212

We just released Claude Code channels, which allows you to control your Claude Code session through select MCPs, starting with Telegram and Discord. Use this to message Claude Code directly from your phone.

English
6
3
22
1.8K
Wes Roth
Wes Roth@WesRoth·
Lovable, an AI platform originally designed strictly for building full-stack applications, has announced an update that transforms it into a generalized, agentic workspace. The platform is expanding its capabilities to handle data science, business analysis, and marketing asset creation directly within the same chat interface used to build software. Unlike standard chatbots that just generate text, Lovable’s AI agent can now write and execute Python scripts in a secure environment to run hard math, analyze massive datasets, and process files. The platform can now take raw data (like a messy CSV or a Slack channel export) and automatically generate formatted PDFs, PowerPoint pitch decks, Word documents, and branded invoices. Users can prompt the agent to process and generate visual assets. This includes resizing and watermarking image batches, converting 2D product images into 3D animations, and generating marketing videos directly from text descriptions. Lovable can natively connect to external tools. For example, it can pull recent client call transcripts from Granola to instantly build a working software prototype, or scrape a Slack #feedback channel to rank feature requests and run sentiment analysis. You can now upload a bloated Excel tracking sheet, a PDF spec, or even a static screenshot of a competitor's product, and Lovable will convert it into a fully functional web app with a real database and user logins
Anton Osika – eu/acc@antonosika

Introducing Lovable for more general tasks. Lovable has always been for building apps. Today it also becomes your data scientist, your business analyst, your deck builder, and your marketing assistant. This is a big step toward what Lovable is becoming: a general-purpose co-founder that can do anything. See examples below.

English
2
2
8
1.2K
Wes Roth
Wes Roth@WesRoth·
Perplexity has announced Perplexity Health, a new integration suite for its cloud-based autonomous agent, Perplexity Computer. This update transforms the AI from a general-purpose researcher into a highly personalized health and wellness analyst. The platform natively connects to Apple Health, electronic health records (EHRs) from over 1.7 million care providers, and major fitness wearables like Fitbit, Ultrahuman, and Withings (with ŌURA integrations expected soon). Rather than looking at isolated metrics, the AI can cross-reference your live wearable data (like a sudden spike in resting heart rate) with your historical lab results and medical records simultaneously to provide highly contextualized answers. To combat the unreliability of SEO-optimized health blogs, Perplexity Health strictly grounds its answers in premium medical literature, clinical guidelines, and peer-reviewed journals, providing direct citations for every claim.
Perplexity@perplexity_ai

Perplexity Computer now connects to your health apps, wearable devices, lab results, and medical records. Build personalized tools and applications with your health data, or track everything in your health dashboard.

English
2
2
10
1.2K
Wes Roth retweetledi
Wes Roth
Wes Roth@WesRoth·
The incident reportedly took place at a Haidilao hotpot branch in San Jose, California. According to early reports, this was allegedly a Zootopia-themed entertainment robot doing a promotional dance that suffered a severe software glitch. It supposedly got stuck mid-dance routine literally throwing the "forbidden steps" and refused to power down, leading to the plate-smashing wrestling match with the staff.
ElBuni@therealbuni

SE RE PUDRIO EN UN RESTAURANTE CHINO Un robot se canso de laburar, rompió la vajilla y se puso a tirar los pasos prohibidos Lo tuvieron que contener entre varios

English
9
3
39
8.1K

Keşfet

@rileybrown @TimSweeneyEpic @vibecodeapp_ @elonmusk @BarackObama @taylorswift13 @cristiano @BillGates