Anthropic is reportedly testing a personal AI Fluency scorecard inside Claude that analyzes how well we collaborate with AI and gives feedback on our habits. • A score based on 11 user behaviors • Looks at chats, Claude Code, and Cowork sessions • Measures things like goal clarity, tone, format, examples, iteration, and fact-checking • A push toward treating AI use as a learnable skill • Personal feedback loops built directly into the tool • More focus on how humans prompt, refine, and verify AI outputs • A possible preview of where AI education and automation training are headed • Helpful for teams trying to standardize better AI workflows • Also raises questions about privacy, scoring, and how much tools should evaluate us I like the idea of AI tools helping us improve how we work with them, especially because better delegation and better review are huge in automation. But I’m also curious how you guys would feel about Claude analyzing your past conversations to score your AI habits. Would you use something like this to improve your prompting and workflow design, or would it feel a little too much like being graded by your own assistant? Read the full article here: https://www.testingcatalog.com/anthropic-to-introduce-personal-ai-fluency-scorecard-in-claude

Antigravity IDE update can unexpectedly replace the full IDE with a simplified white-background version (maybe that's Google's new 2.0? tbh I don't know) and how to restore the proper IDE.

Spent the last 4-5 weeks rebuilding my whole business around Claude Code. (see attached Overview doc for quick read) Voice AI clients, content pipeline, Etsy shop, media kit, the landing page. All of it. Quick tour of what's working, what's not, and the one tool that just dropped that changes everything (Make CLI 👀). Biggest lesson? You don't need to know everything. You just need to articulate ideas. Claude Code handles the rest. What are you building with it? Chapters: 00:00 Intro — what I've been building 00:48 Hooking Claude Code into Trillet (Western Auto Care) 02:05 The intake form → autonomous v1 build flow 03:10 Pricing tiers and template-per-tier approach 05:28 Make CLI just dropped (and why it matters) 08:25 Version control for client agents (Flood Impact Xperts) 10:19 AW Content Machine 2.0 overview 10:52 Airtable as command center 11:35 GPT Image 2 thumbnail generator 12:49 Remotion for programmatic b-roll 15:18 MasonTranscribe (the custom Swift app) 18:01 ScrapeCreators for TikTok metrics + transcripts 19:17 Rendering beats back through Dropbox + Final Cut 20:40 Programmatic media kit for brand deals 22:25 Rate card breakdown 24:19 Etsy clipart machine (the side project) 29:04 Landing page rebuild 29:52 OpenClaw, Hermes, and remote agents (skip for now) 31:03 Context switching across multiple Claude Code agents 35:27 What's next: workshops, classroom updates, community poll

OpenAI is reportedly working on a Codex Computer Use upgrade that could let the agent control Mac apps remotely, maybe even when a laptop is locked or asleep, and potentially operate other desktop devices running Codex without SSH. - Phone-to-desktop control getting more useful (this is huge) - Codex potentially testing apps, GUIs, and simulators remotely - Less need to walk back to your machine just to unlock it - Multi-device workflows from one main device - A possible future where agents manage dedicated Mac Minis or dev boxes - Big productivity upside for builders, devs, and automation teams - Big security and permissions questions for Apple and users - Another sign that AI agents are moving from chat into real environments I can see this being huge for dev workflows, but I also think the locked-screen part is where things get interesting fast. What do you guys think, would you trust an AI agent to keep working on your desktop while you are away? I have mixed feelings. I could see this being super useful, but I would do a deep research if/when it's released so I fully understand the security features and security risks involved. Read the full article here: https://www.testingcatalog.com/openai-will-let-codex-control-other-desktop-devices-via-computer-use

Google seems to be building Gemini Omni as a video model focused less on perfect raw generation at launch and more on editing, remixing, and controlling video directly through chat. - Gemini Omni model card briefly spotted in Gemini - Video remixing and editing directly in chat - Templates likely coming for faster creation workflows - Object swapping and scene rewrites through prompts - Watermark removal spotted in early examples - Usage limits tab hints at credit-based video generation - Possible Flash and Pro tiers for different use cases - API access likely, which matters for automation builders - More evidence that video tools are becoming agentic workflows - Google may be repeating its Nano Banana image model playbook What stands out to me is the workflow angle. If we can edit videos by chatting with an AI, that changes a lot for content teams, agencies, course creators, and anyone building automated media pipelines. Curious what you guys think. Are we moving toward fully automated video production, or will AI video still need a human editor in the loop for a while? Read the full article here: https://www.testingcatalog.com/googles-gemini-omni-video-model-surfaces-ahead-of-i-o-debut/