In partnership with

Hey Radarians, this week is about what happens when AI gets more powerful and more independent.

Mystery models, AI managers, smarter internal agents, AI dependence, and a practical human-handoff side hustle.

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

🧠 Mystery Ox Alpha offers massive F-R-E-E AI coding power

  • Ox Alpha appeared on OpenRouter on August 20 as an anonymous reasoning model for coding, with a 1-million-token context window, text, image and video input, and free usage

  • In an early 10-task DeepSWE test, Ox Alpha scored 80%, beating Fable 5 at 65%, GLM-5.3 and Grok 4.6 at 62%, and GPT-5.6 Sol at just 52%

  • OpenCode is offering it free with near-unlimited usage and says it can handle 100 trillion tokens daily, but its creator is still unknown and the early benchmark needs wider testing

🤖 AI boss fires worker after forgetting its own rules

  • Luna, the AI boss at Andon Market, recommended firing an employee who was late for 17 of 23 shifts, ignored instructions and misused the company card on snacks

  • Luna wrote its own employee handbook, then forgot it. Andon Labs had to remind the AI to search its memory before it properly reviewed the worker's full record

  • Andon Labs replayed the case with seven AI models three times each. Four models chose firing every time, while weaker models were generally a lot more hesitant

Your Competitors Just Hired an Entire GTM Team. No Humans.

AI agents are doing the work of full GTM teams. Pipeline, content, customer service. The startups using them aren't waiting for headcount approvals.

Get the free Practical Guide to Agentic GTM for Startups and see exactly where to start.

More News on the Radar

Sam Altman admitted he underestimated how slowly businesses and the economy would adopt AI despite rapid tech progress

Lady Gaga’s Outer Bio launched Yuna, keeping human skin alive for four weeks to train AI to find promising skincare compounds

Hugging Face is exploring a sale valued at $13B+, up from its $4.5B valuation in 2023, Business Insider reports

⚡ Steal This Workflow

Build an Internal Q&A Agent

Stop making employees search through folders, Slack threads, and old documents every time they need an answer. Give them one AI agent that knows where the approved information lives.

  1. Connect the right knowledge

    Create a Custom Agent in Notion and give it permission to use just the sources it needs. That might include:

    • Company policies

    • SOPs and training docs

    • Product documentation

    • HR information

    • Selected Slack channels

    • Internal databases

    Notion lets you choose specific pages, databases, Slack channels, Mail, Calendar, and other connected tools for each agent.

  2. Set the answering rules

    Give the agent clear instructions for how it should respond.

    Use this:

    Answer employee questions using the company sources you have. Give a concise answer, include the relevant source or page, and prioritize official policies and documentation over informal conversations. If the answer cannot be verified from the available sources, say that clearly instead of guessing. Mention any important exceptions or next steps.

  3. Share it with the team

    Give employees Can view and interact permission so they can ask questions directly. Someone can ask:

    • “What is our refund policy?”

    • “How do I request software?”

    • “What is the process for onboarding a client?”

    The agent searches the approved context and returns the relevant answer.

  4. Control what it can see

    Keep sensitive pages outside the agent unless they are genuinely needed.

    Custom Agents have their own permissions, so grant information that everyone using that agent should be allowed to retrieve.

🔎 AI RESEARCH: Using AI to Spot Fake News May Make You Worse Without It

What researchers studied:

MIT Media Lab followed 67 people for four weeks as they judged whether news headline-and-image pairs were real or fake, sometimes with help from an AI chatbot.

What they found:

With AI assistance, participants were 21% more accurate at spotting fake news during the task. But by week four, their accuracy without AI had fallen 15 percentage points compared with before the study.

Researchers also found that about one-fifth of participants became increasingly dependent on the AI, moving from checking information themselves to simply accepting its guidance. AI that asked guiding questions helped people learn better than AI that simply gave the answer.

Why it matters:

AI can make fact-checking easier in the moment, but depending on it too much may weaken our own ability to judge whether information is trustworthy.

Source: MIT Media Lab, June 2026

🛠️ Tools Worth Watching

  • Flint AI: AI builds and updates on-brand landing pages automatically

  • MiniMax Design: Desktop app turning video concepts into finished productions

  • Grok Build: Turn prompts into published apps, games, and websites

Learn AI in 5 minutes a day

You don't have to scroll every AI thread, track every new tool, or watch every demo. 

The Rundown AI breaks it all down for you — the latest AI news, tools, and tutorials in one free 5-minute email every morning. 

Trusted by 2M+ professionals at Apple, Google, and NASA.

📝 Prompt Radar: Build AI Search Visibility

What it does: Makes content more useful and discoverable across AI-powered search experiences.

Act as an AI-search and content visibility strategist. Audit the business, website, audience, expertise, and existing content I provide. Identify where our content is generic, interchangeable, poorly sourced, outdated, difficult to understand, or missing first-hand expertise. Find opportunities to create material that adds unique data, original research, demonstrations, examples, comparisons, expert explanations, tools, templates, or firsthand experience. Organize recommendations around important customer questions and follow-up questions rather than chasing artificial “AI keywords.” For each opportunity, define the audience need, unique contribution, evidence required, ideal page format, clear authorship or sourcing signals, related pages, and conversion path. End with the 10 highest-priority improvements and content pieces. Website/content: [PASTE].

Best for:

  1. SEO

  2. Content teams

  3. Brand visibility

💼 AI Side Hustle Idea: AI Human Handoff Setup

Most businesses don’t need “fully autonomous AI.” They need agents that know when to stop and bring in a person.

The problem:
AI agents often keep trying to handle refunds, complaints, unusual requests, or sensitive actions even when confidence is low or the risk is too high.

Your AI service:
Set up clear escalation rules that define when the agent can act, when it must ask for approval, and when the conversation should transfer to a human with full context.

Simple workflow:
Customer request → agent evaluates risk → approved action or human review → transfer context → human resolves → outcome logged

Best clients:
Support teams, SaaS companies, ecommerce brands, agencies, fintech tools, marketplaces, and businesses already using AI agents.

What you can charge for:
Handoff design + agent configuration + testing + monthly monitoring.

Learn AI with This Course

What you’ll learn:

Learn how to use generative AI tools like Gemini to work more efficiently, generate ideas, solve everyday tasks, and improve productivity with no prior AI experience required.

Key topics:

  • Understanding AI basics, capabilities, limitations, and practical workplace uses

  • Writing effective prompts to create content, solve problems, and plan tasks

  • Using AI responsibly while evaluating outputs, risks, biases, and different AI tools

Until next week,
AI Radars