In partnership with
AI that works like a teammate, not a chatbot
Most “AI tools” talk... a lot. Lindy actually does the work.
It builds AI agents that handle sales, marketing, support, and more.
Describe what you need, and Lindy builds it:
“Qualify sales leads”
“Summarize customer calls”
“Draft weekly reports”
The result: agents that do the busywork while your team focuses on growth.
Beginners in AI
Good morning and thank you for joining us again!
Welcome to this daily edition of Beginners in AI, where we explore the latest trends, tools, and news in the world of AI and the tech that surrounds it. Like all editions, this is human curated, and published with the intention of making AI news and technology more accessible to everyone.
THE FRONT PAGE
What YouTube's Algorithm Really Wants You to Watch
TLDR: A new study found that 21% of YouTube Shorts shown to new users are AI-generated "slop," which is low-quality content churned out to farm views rather than inform or entertain.
The Story: Video editing company Kapwing analyzed 15,000 of YouTube's most popular channels and found 278 posting nothing but AI-generated junk. These channels have racked up 63 billion views and 221 million subscribers, earning an estimated $117 million annually. When researchers created a fresh YouTube account and watched the first 500 recommended Shorts, 104 were AI slop and a full third qualified as "brainrot." The top-earning channel, India-based Bandar Apna Dost, features an animated monkey in absurd scenarios and pulls in roughly $4.25 million a year.
To be clear, AI slop isn't the same as thoughtful AI-assisted content. Filmmakers and creators increasingly use AI tools to enhance their work, whether that's cleaning up audio, generating concept art, or speeding up editing workflows. Top film schools now offer courses on AI ethics and technique. The difference is intent and effort: AI-assisted creators still bring human judgment, storytelling, and craft to the process. Slop is the opposite.
Its Significance: YouTube's algorithm doesn't distinguish between human creativity and automated content farms. It optimizes for engagement, not quality, and AI slop is winning that game. Much of this content targets children through bright animations and nonsensical storylines, which raises concerns beyond wasted screen time. The platform's official position is that AI is just a tool and all content must follow community guidelines. But that response sidesteps the real issue. When one-fifth of what new users see is mass-produced digital junk, the algorithm isn't connecting people with quality content. It's connecting advertisers with eyeballs at any cost, and creators who actually put in the work are getting buried.
QUICK TAKES
The story: Meta is acquiring Manus, a Singapore-based AI agent startup that can browse the web and complete tasks on its own. Manus had over 2 million people on its waitlist and was making $125 million per year. The company raised $75 million at a $500 million valuation earlier this year from investors including Benchmark. The U.S. Treasury reviewed the deal because Manus's parent company is based in Beijing.
Your takeaway: Meta is betting big on AI agents that can actually do things for you, not just chat. This acquisition signals the race to build useful AI assistants is heating up fast.
The story: China released draft rules requiring AI chatbots and agents to tell users they're talking to a machine - both when they log in and every two hours during use. The rules also require AI to promote "core socialist values" and ban content about suicide, self-harm, gambling, and violence. Companies would need government security reviews before launching new AI tools.
Your takeaway: While the U.S. rolls back AI safety rules, China is moving in the opposite direction with strict requirements. These rules could shape how AI companies worldwide design products for global markets.
The story: Researchers built an AI model from scratch that mimics how real brains learn, without using any animal training data. The model learned to recognize objects exactly like lab animals do. More surprising: it discovered that about 20% of neurons fire before mistakes happen, essentially predicting errors before they occur. Scientists then found this same pattern in existing animal data that nobody had noticed.
Your takeaway: AI isn't just learning from brains anymore - it's helping us discover how brains actually work. This finding could lead to better treatments for learning disorders and more efficient AI systems.
The story: Researchers built artificial skin for robots that works like human skin, using "spiking" signals similar to our nervous system. The skin has sensors for heat, cold, pressure, and pain that send information the same way our neurons do. It can connect directly to energy-efficient AI chips, making robots that can feel and react much faster while using less power.
Your takeaway: Robots are getting closer to human-like touch. This could mean safer robots in factories, better prosthetic limbs, and machines that can handle delicate tasks like surgery or elderly care.
The story: Chinese researchers created an AI chip called LightGen that uses light instead of electricity. It packs over 2 million artificial neurons onto a chip smaller than a quarter and can generate images and video more than 100 times faster than Nvidia's A100 GPU while using far less energy. The chip still needs bulky lasers to work, so it's not ready for everyday use yet.
Your takeaway: As AI's energy demands skyrocket, optical computing offers a path forward. If this technology leaves the lab, it could slash the power bills and environmental impact of AI data centers.
TOOLS ON OUR RADAR
📊 Fathom Analytics Paid: Track website visitors without cookies or consent banners, GDPR compliant by default.
⚡ Make (Integromat) Freemium: Visual automation builder with 2000+ apps and powerful branching logic.
📑 Coda Freemium: Merge documents, spreadsheets, and databases with 600+ app integrations, used by Uber and NYT.
✅ Super Productivity Free and Open Source: Track tasks and time with deep work focus features and developer integrations, alternative to Todoist.(One of my favorite new apps to test)
🎙️ PLAUD NotePin Freemium: Wearable AI voice recorder that transcribes in 112 languages with speaker labels and generates AI summaries.
TRENDING
2025: The Year AI Hype Met Reality - The year started with massive fundraising (OpenAI hit $300B valuation) but ended with disappointing model releases and 50+ copyright lawsuits. DeepSeek proved you don't need billions to build competitive AI.
Data Center Backlash Goes Mainstream - 142 activist groups across 24 states are now fighting data center projects. About $64 billion in developments have been blocked or delayed as communities push back against power consumption and rising electricity bills.
AI Agents Arrived in 2025 - With Mixed Results - New browsers from Perplexity, OpenAI, and others let AI act on your behalf online. But agents still make small mistakes too often to trust with important tasks, and one was misused in a cyberattack.
Indian Cinema Embraces AI - With Controversy - India's first fully AI-generated feature film is coming in 2025, and deepfakes are de-aging actors in Malayalam movies. But when one studio re-released a classic with an AI-altered ending, directors and actors condemned it.
AI Maps Nanocrystal Reactions Like a Subway System - Korean researchers used AI to visualize how semiconductor materials form, creating maps that look like subway routes. They discovered materials act as "traffic lights" that control which reactions happen.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, Gemini)
Sunk Cost Fallacy Detector: Stop throwing good money/time after bad by separating past investment from future value
Build me an interactive Sunk Cost Fallacy Detector as a React artifact that helps you decide whether to continue or quit based on future potential, not past investment.
The console should include these sections:
1. **Situation Input** - What are you stuck on?:
• Decision description (project, relationship, investment, career, education)
• How long invested: Months, Years, Decade+
• Resources invested so far:
- Money spent: $[amount]
- Time invested: [hours/months/years]
- Emotional energy (1-10)
- Opportunity cost (what you gave up)
• Current status: Struggling, Stagnant, Declining, Improving
• "Analyze Decision" button
2. **Sunk Cost Calculator** - What's already gone:
• Visual "money down the drain" metaphor
• Total investment breakdown:
- Financial: $X (unrecoverable)
- Time: X hours (can't get back)
- Effort: High/Medium/Low
• Emotional attachment meter (1-10)
• "This is gone regardless" reminder
• Comparison: "If starting fresh today, would you invest this?"
3. **Future Value Assessment** - What happens if you continue?:
• Looking forward only (ignore sunk costs):
- Additional money needed: $[amount]
- Additional time required: [estimate]
- Probability of success: [percentage]
- Expected outcome if successful
- Expected outcome if failed
• Net future value calculation
• "Is the future worth it?" score
• Best case, worst case, realistic case scenarios
4. **Fresh Start Test** - The critical question:
• **"If you hadn't already invested anything, would you start this today?"**
• Honest yes/no with reasoning space
• Reframe: "Imagine you just inherited this situation. Continue or quit?"
• Comparison: Your actual choice vs. outsider perspective
• "What would you tell a friend?" prompt
• Emotional vs. logical reasoning separator
5. **Exit Cost Analysis** - What does quitting cost?:
• Costs of continuing:
- More money
- More time
- Opportunity cost (what else you can't do)
- Stress and energy drain
• Costs of quitting:
- Write-off amount
- Ego hit (admitting mistake)
- Others' opinions
- Short-term pain
• Side-by-side comparison
• "Which cost is actually worse?" analysis
• Recovery timeline if you quit
6. **Bias Detector** - Are you falling for it?:
• Sunk cost warning signs:
⚠️ "I've already invested so much"
⚠️ "I can't let it all be for nothing"
⚠️ "Just a little more and it'll work"
⚠️ "I'll look like a failure if I quit"
⚠️ "I've come too far to stop now"
• Which phrases you're using
• Severity: Low, Medium, High, Extreme
• Related biases (escalation of commitment, loss aversion)
• "Search Sunk Cost Examples" for case studies
7. **Decision Recommendation** - Clear verdict:
• Overall assessment:
- Future value score (1-10)
- Sunk cost influence level (how biased you are)
- Recommendation: Continue, Quit, Pivot, Take a break
• Key factors in recommendation
• If quit: Recovery plan and next steps
• If continue: Success criteria and quit triggers
• "Decide by [date]" deadline setter
• "How will you know when to quit?" checklist
Make it look like an analytical decision tool with:
• Weight scale visual (sunk costs vs. future value)
• Before/after comparison layouts
• Warning symbols for bias indicators
• Clean, objective design (remove emotion)
• Professional color scheme (grays, blues, red for warnings)
• Visual metaphors (drains, scales, paths)
• Data-driven aesthetic
• Clear verdict display
When I click "Search Sunk Cost Examples," use web search to find real-world examples of sunk cost fallacy (business failures, project cancellations), success stories of quitting, and decision-making frameworks for cutting losses.What this does: Separates past investment (which is gone forever) from future potential—forcing you to evaluate whether continuing makes sense based on what happens next, not what you've already spent, with clear bias detection and a quit/continue recommendation.
What this looks like:
WHERE WE STAND (based on today’s stories)
✅ AI Can Now: Predict when your brain is about to make a mistake - a biomimetic model discovered "incongruent neurons" that fire before errors happen, a pattern scientists then found hiding in real animal data.
❌ Still Can't: Complete a multi-step online task reliably enough to trust AI agents launched this year but small mistakes mean humans still need to supervise every action.
✅ AI Can Now: Generate video 100x faster than top GPUs using light instead of electricity, but only in a lab with bulky lasers attached.
❌ Still Can't: Build data centers without community backlash - 142 activist groups across 24 states have blocked or delayed $64 billion in projects over power consumption and rising electricity bills.
✅ AI Can Now: Feel pressure, heat, cold, and pain through artificial skin that sends signals like human neurons - the missing piece for robots that handle delicate tasks.
❌ Still Can't: De-age an actor or alter a film's ending without sparking public outrage - Indian cinema's AI experiments show audiences still reject unauthorized changes to beloved works.
RECOMMENDED LISTENING/READING/WATCHING
May lives in a near-future where climate change is rampant and AI is everywhere. She's struggling financially, so she sells her face to a tech company for quick cash. Then she takes her family on vacation to the last forest preserve, trying to escape the surveillance and AI assistants that dominate daily life.
Phillips writes a dystopia that feels five minutes away from our current world. The AI’s are helpful until they're invasive. The environmental collapse is terrifying because it's incremental. May's attempt to protect her children while being broke and desperate rings true. This is science fiction as a mirror held up to exactly where we're heading if we don't course-correct.
Thank you for reading. We’re all beginners in something. With that in mind, your questions and feedback are always welcome and I read every single email!
-James
By the way, this is the link if you liked the content and want to share with a friend.
Some links may be affiliate links. This helps support the newsletter at no extra cost to you.
