Hey Radarians, this week AI gets more local and practical.
OpenAI shakeups, Meta’s fresh agent, smarter file organization, AI creativity research, and useful tools worth watching.
Let’s dive in.
What’s Inside 📡
Radar News Brief
Steal This Workflow
AI Research
Tools Worth Watching
Prompt Radar
AI Side Hustle Idea
Learn AI with This Course
Radar News Brief
⚡ OpenAI’s longtime COO Brad Lightcap is leaving to start something new
Brad Lightcap, one of OpenAI’s longest-serving executives, is leaving the company after joining in 2018 and helping build major teams across finance, legal, partnerships and operations
After serving as CFO and later COO, Lightcap moved to special projects earlier this year and says he has been focused on what the world will need as AI enters its next phase
His departure adds to a wider leadership shakeup at OpenAI, following recent exits by Fidji Simo, Bill Peebles and Kevin Weil as the company prepares for a potential IPO
⚡ Meta launches 30B AI agent you can run on your own computer
Meta released Muse Glimmer, a roughly 30B open-weight model built for coding, reasoning and other agentic tasks, while remaining small enough to run on a Mac or PC
It can operate on a single consumer graphics card, giving developers a customizable local agent without depending entirely on cloud-hosted AI services
Glimmer is not superintelligence, but it fits Meta’s wider vision of putting capable AI directly in people’s hands instead of keeping advanced systems centralized
More News on the Radar
Anthropic plans invisible watermarks for Claude’s text, code and files to boost transparency and meet EU AI Act rules
SpaceXAI and Cursor have launched Grok Bot, a beta app that makes Grok act like an iMessage-style team of AI agents
OpenAI launched GPT-5.6-Cyber and is giving vetted defenders access through its expanded Daybreak security program
⚡ Steal This Workflow
Clean Up Google Drive With AI
Stop manually dragging dozens of random files into folders. Google Drive can use Gemini to suggest where everything should go.
Find the messy folder
Open Google Drive on the web and go to My Drive or any folder containing loose files.
If eligible files are detected, tap on Suggest file moves.
Let Gemini organize them
Gemini analyzes the files and recommends moving them into existing folders or creating folders for related content.
It can work with Docs, Sheets, Slides, PDFs, Office files, images, shortcuts, and videos with transcripts.
Tell it how you want things grouped
Refine the suggestions with natural-language instructions.
Try:
Organize these files by client and project. Keep contracts, meeting notes, invoices, research, and final deliverables separate. Create new folders when necessary and keep folder names short and clear.You can also ask Gemini to group files by project or year, organize meeting notes, or create more or fewer folders.
Review before anything moves
Check Gemini’s recommendations one by one or approve them in bulk.
Nothing gets moved without your permission.
Your messy Drive becomes an organized workspace without manually sorting every file yourself.
🔎 AI RESEARCH: Imperfect Training May Make Image AI Creative
What researchers studied:
Google researchers studied why diffusion models can create fresh images instead of simply copying examples from their training data.
What they found:
They found that neural networks naturally learn a slightly smoother version of their training data instead of learning it perfectly.
This smoothing helps the AI generate fresh results that fall between patterns it has already learned, producing images that can look original while still staying realistic.
Why it matters:
What looks like AI creativity may partly come from the imperfect way neural networks learn.
Source: Google Research, July 2026
🛠️ Tools Worth Watching
Orchid: Personal AI assistant managing everyday tasks through messages
Clera: AI talent agent matching candidates with suitable job roles
Thumbmagic: Optimized thumbnails powered by data from 200+ designs
📝 Prompt Radar: Create Email Nurture Sequence
What it does: Builds an email series that moves subscribers toward a useful next step.
Act as an email lifecycle strategist. Create a nurture sequence for the audience, lead source, offer, and desired action I provide. First identify what the subscriber already knows, what they want, why they joined, what may stop them from acting, and what trust must be built. Then design a sequence where every email has one clear job. For each email, provide the purpose, customer belief to change, main idea, subject-line options, opening direction, key points, proof required, CTA, and connection to the next email. Balance education, problem awareness, trust, product understanding, objection handling, and conversion. Avoid sending repeated sales messages disguised as education. End with timing recommendations, segmentation ideas, and performance metrics. Details: [PASTE DETAILS].Best for:
Lead nurturing
Course sales
B2B marketing
💼 AI Side Hustle Idea: AI Spreadsheet Cleanup Automation
Most businesses don’t need “another spreadsheet expert.” They need messy recurring data cleaned without repeating the same manual work every week.
The problem:
Excel and CSV exports arrive with duplicate rows, inconsistent names, broken formats, missing categories, and messy data that takes hours to fix before anyone can use it.
Your AI service:
Build a reusable workflow that cleans fresh files, standardizes entries, removes duplicates, classifies records, flags errors, and automatically creates a clean report.
Simple workflow:
Raw Excel/CSV → validate data → clean + standardize → classify → remove duplicates → generate report
Best clients:
Ecommerce stores, agencies, sales teams, accountants, recruiters, operations teams, and businesses handling regular data exports.
What you can charge for:
Automation setup + customization + monthly maintenance.
Learn AI with This Course
What you’ll learn:
Learn how to build and deploy AI agents with smolagents, LlamaIndex, and LangGraph through hands-on lessons, real-world assignments, and Hugging Face Spaces.
Key topics:
AI agent fundamentals with smolagents, LlamaIndex, and LangGraph
Building and deploying agents with Hugging Face Spaces
Real-world assignments, leaderboard projects, and agent observability
Until next week,
AI Radars