Uber Drivers are Training for the Future

Hi everyone,

Uber is getting creative with its workforce, using driver data to help train its AI, an intriguing move, but what does it mean for privacy and innovation?

Meanwhile, Andrej Karpathy shares his thoughts on the future of AI agents, suggesting a slow approach. Could this be the path to more sustainable advancements?

Let's get right into it.

In this issue:

• 🤝 In Partnership: One plan for every app you need

• 🤿Deep Dive: Uber pays drivers to train AI

• 🤝Powered by: The brief for founders

• 🖼️AI Art: Examples of great and trending AI art

• 🤿Deep Dive: Karpathy predicts slow road to true AI agents

• 🤝Supported by: Stop fraud in its tracks

• ⚒️Tool Snapshots: Tools for AI, no-code, and productivity

🤝IN PARTNERSHIP WITH SETAPP

One membership. 250+ Mac & iPhone apps

Stop juggling tools. Start shipping results.

With Setapp, you get access to 250+ Mac and iOS apps that save you time, automate work, and keep projects moving. From workflow boosters to workspace organizers: everything is in one place.

Start with a 7-day trial, commitment-free.

Special for The Intelligent Worker readers: 20% off your 1st year of the Mac + iOS Annual plan. Redeem with INTELLIGENT20.

What readers say:

• “Setapp helped me discover new ways to be more productive and creative. It combines tools, tips, and hacks.”

• “Was able to save in yearly subscriptions & version upgrades by getting the exact same apps in Setapp. Bartender. BetterTouchTool. These two alone are worth the price of admission.”

New users only. Ends Nov 30.

Get 20% off the Mac + iOS Annual plan

🤿 DEEP DIVE

Uber and Uber AI Solutions Pilot Driver Data Tasks

Intelligence: Uber launched a US pilot that pays drivers and couriers to complete in‑app data tasks that help train AI models, with a broader rollout planned by the end of 2025.

• Tasks include uploading photos, recording short voice clips, and submitting documents for AI training and data labeling.

• Each task pays a fixed amount based on complexity and time, with payments appearing in driver balances within 24 hours, while US pay rates are not yet finalized.

• Task availability will vary based on client demand for data labeling work.

• The pilot was first tested in India and is set to expand across the US by the end of 2025.

• The program is separate from Uber’s autonomous ride efforts, such as the Waymo collaboration, and runs through Uber AI Solutions, which serves clients including Aurora and Tier IV.

🤝POWERED BY THAT START UP GUY

Founders need better information

Get a single daily brief that filters the noise and delivers the signals founders actually use.

All the best stories — curated by a founder who reads everything so you don't have to.

And it’s totally free. We pay to subscribe, you get the good stuff.

Get the brief

🖼️ AI ART

Examples of great and trending AI art

🤿 DEEP DIVE

Andrej Karpathy and OpenAI Agent Timelines

Intelligence: OpenAI cofounder Andrej Karpathy said on the Dwarkesh Podcast that functional AI agents are likely a decade away, arguing the industry is building tools beyond current model capabilities.

• Karpathy said today’s agents lack sufficient intelligence, are not multimodal enough, struggle with computer use, and do not have continual learning or reliable memory.

• He estimated about a decade to resolve these gaps and noted he is now building an AI-native school at Eureka Labs.

• He prefers human-AI collaboration where agents pull API docs, verify correct usage, ask when uncertain, and help users learn rather than dump large code outputs.

• He criticized visions of fully autonomous agents that render humans unnecessary and spread low-quality AI slop.

• In a follow-up post on X, he said the industry is overshooting tooling relative to present capability, while his timelines are five to ten times more conservative than common Silicon Valley hype but still optimistic compared to AI skeptics.

• Context from peers includes investor hype around 2025 as the year of the agent and Scale AI executive Quintin Au’s estimate of roughly a 20 percent error per action, implying about a 32 percent chance of getting five steps all correct.

🤝SUPPORTED BY CHARGEFLOW

Stop Fraud Before Fulfillment

Post-purchase fraud is rising, slipping past checkout tools and draining retail profits. Chargeflow Prevent blocks fraud after payment but before fulfillment, cutting disputes by 90% with <0.1% false positives.

Try Prevent Free

⚒️ TOOL SNAPSHOTS

Futuristic tools within AI, no-code, and productivity

• 📊 Rybbit - Simplify web analytics with GDPR-compliant, open-source tools. Free option available.

• 📂 Claude Skills - Packaging and standardizing workflows for task efficiency. Free to try.

• 🔧 Kibo UI Patterns - Streamline development with accessible open-source components. Free to use.

• 🎵 Daft Music - High-performance, elegant playback for Apple Music on macOS. Free option available.

• 💻 Compyle - Steering your coding process for efficient mechanical execution. Free to try.

ℹ️ ABOUT US

The Intelligent Worker helps you to be more productive at work with AI, automation, no-code, and other technologies.

We like real, practical, and tangible use-cases and hate hand-wavy, theoretical, and abstract concepts that don’t drive real-world outcomes.

Our mission is to empower individuals, boost their productivity, and future-proof their careers.

We read all your comments - please provide your feedback!