• Everyday AI
  • Posts
  • EP 636: Uber paying drivers $1 to train AI models? A sign of what’s next

EP 636: Uber paying drivers $1 to train AI models? A sign of what’s next

OpenAI releases agentic browser, Google drops huge vibe coding updates, Anthropic brings Claude Code to the web and more.

Author

Everyday AI
October 21, 2025

Sup y’all 👋

Don’t sleep on today’s #HotTakeTuesday show. It’s more than just the ‘why’—it’s what Uber’s $1 AI tasks mean for the future of jobs, data quality, and how you’ll work next. Catch the real story and what you need to do now.

What are your thoughts?

What are your thoughts on the new ChatGPT browser?

🗳️ Vote to see live results 🗳️

Login or Subscribe to participate in polls.

We’ll be covering it tomorrow for sure!

✌️

Jordan

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: Uber’s paying people as little as $1 to help train LLMs. We break down the bigger picture. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Qwen releases Deep Research update, Vibe coding might be shaky, Microsoft launches free AI global skills program and more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: OpenAI releases agentic browser, Google drops huge vibe coding updates, Anthropic brings Claude Code to the web and more. Read on for Byte Sized News.

💪 Leverage AI: Believe it or not, Uber’s Digital Tasks move greatly impacts future of jobs, data quality, and how you’ll work next  Keep reading for that!

↩️ Don’t miss out: Did you miss our last newsletter? We covered: OpenAI’s lopsided economics, AWS’s AI hiccups causes delays, the FTC’s lack of AI transparency and more and more. Check it here!

EP 636: Uber paying drivers $1 to train AI models? A sign of what’s next

Uber is paying its drivers as little as $1 to train LLMs. 😯

Smart business move or eery sign of what's to come?

On this Hot Take Tuesday episode, we uncover the trend of dirt cheap data labeling, why it's a good thing and a bad thing, and how this is actually a sign of what's next.

What's our hot take?

Also on the pod today:

•AI replacing 9-to-5 jobs? 🤖 
• Internet flooded with AI slop 🌐
• Autonomous vehicles data collection 🚕

It’ll be worth your 37 minutes:

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Maestro SFX turns simple text prompts into production ready sound effects, Director gets stuff done for you on the web using its own browser, and EraseVideo can remove AI video watermarks.

LLM Updates - Anthropic just launched Claude for Life Sciences to plug its AI into lab tools and speed up everything from literature reviews to regulatory reports. Want to see how AI could shave days of grunt work into minutes?

AI Deep Research — Qwen dropped a worthwhile Deep Research update:

AI Vibe Coding — AI is shaking up coding, but most engineers say it's not the game-changer CEOs claim—and sometimes it just creates more work.

AI Training — Adecco and Microsoft just launched a free global AI skills program for jobseekers, aiming to close the digital divide and boost employability

1. OpenAI Rolls Out Atlas Browser, Taking Aim at Google Chrome 💻

OpenAI just launched its own web browser, Atlas, on macOS, signaling a direct challenge to Google’s Chrome and potentially shaking up the search landscape, according to the Associated Press. Atlas offers built-in AI tools powered by ChatGPT, aiming to make online searches faster and more intuitive while giving users direct access to smarter answers.

With plans to expand to Windows, iOS, and Android soon, Atlas could change how professionals and businesses use the web for research, decision-making, and growth.

2. Google AI Studio Unveils Game-Changing Vibe Coding Update 🪄

Google AI Studio has just launched a major upgrade to its “vibe coding” platform, making it easier than ever for anyone to whip up and deploy AI-powered web apps in minutes, reports VentureBeat.

With a fresh interface, instant app generation powered by Gemini 2.5 Pro, and playful features like the “I’m Feeling Lucky” button, even absolute beginners can build, edit, and share live applications for free. The update drastically lowers the technical barriers that previously favored developer-focused rivals, putting advanced AI tools within reach of startups, professionals, and creative hobbyists alike.

3. Anthropic Rolls Out Claude Code Web App for Developers 🕸️

Anthropic just launched its Claude Code web app, making its viral AI coding assistant accessible to Pro and Max subscribers at claude.ai, according to TechCrunch. With this move, developers can manage multiple coding agents straight from their browser or iOS app, signaling a shift away from the old-school command line interface.

The company claims its AI models now write 90% of Claude Code’s own product, as CEO Dario Amodei predicts AI will soon handle most software engineering tasks.

4. Wikipedia Traffic Takes a Hit as AI Summaries Dominate Search 🔍

Wikipedia has seen an 8% drop in human pageviews over recent months, according to a new post from the Wikimedia Foundation, blaming the surge of AI-generated summaries in search results and the lure of social media.

Bots have gotten sneakier, skewing traffic data while younger users flock to short-form video apps for info instead of the old open web. Pew Research confirms that Google's instant AI answers mean fewer clicks for publishers, leaving sites like Wikipedia worried about dwindling engagement and donations.

5. Google Unleashes Massive AI Learning Hub 📖

Google just launched "Google Skills," an all-in-one platform featuring nearly 3,000 courses and hands-on labs to supercharge your AI and tech know-how. Packed with content from Google Cloud, DeepMind, and more, it’s designed for everyone from students to company leaders and even offers gamified learning and direct job placement pathways with over 150 companies. With free access for many and instant coding tools, Google is making AI skills more accessible than ever.

🦾How You Can Leverage:

Sounds decent until you realize what's ACTUALLY happening here.

Uber’s just launched ‘Digital Tasks’ program grabbed a lot of headlines and conversation around the need for data, proper compensation and even the future of work.

In a nutshell, here’s how it works. Uber’s new program pays workers starting at $1 per micro task to train AI models. The tasks take minutes to complete and can be done while not driving, on tasks from menu translation to identifying objects in photos. Uber uses this data to improve their own services and sells it to other AI labs.

Here’s what it REALLY means.

AI labs are desperate for unique human data because the internet is becoming regurgitated AI slop. Synthetic content training synthetic content in endless loops, and human creators just kinda blindly copying-and-pasting whatever the outputs may be.

So while Digital Tasks may seem like an isolated experiment where a company is trying to gather higher quality human data to improve operations, it’s indicative of something much larger.

This isn't a gig economy experiment. It’s a sampling of the future of work.

That’s what we tackled on today’s episode of Everyday AI. Let’s dive in to see what it really means.

1 –Traditional Internet Training Data Died 🔥

AI labs scraped the entire internet already.

Dead Internet Theory is kinda real.

Now they're training on AI-generated content. That's a death spiral. GPT-5 launched in August 2024 with training data from a year earlier. By the time models hit shelves, much of the original training data is ancient and benchmark improvements collapsed.

Reddit makes $130 million annually selling human discussion data to AI labs.

Why? Unique human specialist knowledge can't be scraped anymore, and it’s the hottest commodity on the market.

Uber’s Digital Tasks experiment signals something way bigger. Companies like Scale AI, Appen, and Sama built billion-dollar businesses on crowdsourced training, but generic labor pools are worthless now.

Try This

Identify which employees in your company handle problems that don't exist in public documentation anywhere.

Those edge cases are training gold. Schedule 30-minute sessions with three specialists this week and have them narrate their decision-making process while solving complex scenarios.

Record those sessions. Extract the underlying logic patterns that separate expert performance from average in your specific domain.

Build a repository of these decisions before competitors realize this is the new competitive moat in your industry.

2 – Gig Work Replaces Traditional Careers by 2026 ⚡

Full-time 9 to 5 work won’t be the norm in 10 years.

We predicted this back in 2023. LinkedIn CEO Reid Hoffman said the same thing a year later. Most people thought it was wild.

Here's what's actually happening. Your job in 3 years might be training a model in non-technical ways for a few hours daily while juggling three other gig contracts.

Companies won’t need as many full-time employees for most roles anymore. They need specialists who can contribute domain expertise in short bursts to train categorical models. Universities are hemorrhaging enrollment and will hit a financial cliff within two years because they refused to teach AI skills companies actually want.

Try This

List your three strongest domain skills that aren't easily googleable right now.

Those are your future income streams when traditional employment starts to change. Start documenting how you solve problems in those areas with specific examples and decision frameworks.

3 –  Non-Technical Finetuning Teams Are the Future 🚀

Small categorical models trained on your company's unique workflows will dominate enterprise AI in 2026 and beyond.

Not generic large models. Fortune 100 companies are already building internal finetuning departments with non-technical employees. The massive opportunity is creating interfaces that let specialists train domain-specific models without writing code.

Uber's dollar task program is essentially a non-technical UI for model training. Drivers don't need to understand machine learning. They just complete tasks through a simple app interface.

That's the blueprint for enterprise success.

You're gonna have entire teams training models on first-party company data by late 2026. Not engineers. Your best domain experts currently spending 40 hours on their actual jobs. They'll spend 20 hours training models and 20 hours on specialized work that AI can't handle yet.

Companies building this internally will dominate. Everyone else will license generic solutions that don't understand their specific business context.

Try This

Pick one workflow in your company that new employees take months to learn properly.

Map how your top performer handles that process versus average employees. Document fifty real examples of decisions they make that require institutional knowledge.

This becomes your training dataset for a company-specific model. Partner with one domain expert and have them narrate their thinking while handling complex scenarios over the next two weeks.

You're not building AI infrastructure. You're capturing expertise that walks out the door when people leave or get promoted away from the actual work.

1 

Reply

or to participate.