Digg

Digg lets developers earn rewards from the AI coding data they already create.

Developers use Claude Code, Codex, Cursor and other coding agents every day. These sessions contain valuable training signals such as prompts, tool calls, code edits, failures, corrections, tests and final outcomes.

With Digg, users choose which sessions they want to contribute. We structure those sessions into coding trajectories and datasets that can be licensed to AI labs and companies working on post-training, RL and coding agents.

Our consumer-facing idea is simple:

Code with AI → contribute approved sessions → earn AI stock rewards.

Why Now

AI labs need better data for training coding agents beyond public repositories and static code.

They increasingly need real execution data showing how developers:

• understand a task
• navigate a repository
• use tools
• make code changes
• fail and recover
• test and verify the final result

We have already spoken with a few companies and startups working in RL, post-training and AI data. The feedback has been positive, especially around demand for high-quality coding trajectories.

A few are also interested in connecting us with AI labs once we build a meaningful dataset.

Product

Our first product is Digg Code.

It works alongside AI coding agents and allows developers to contribute approved coding sessions.

From these sessions, Digg can build:

• Coding trajectories
• SFT demonstrations
• Correction and preference data
• Tool-use traces
• RL tasks and verifiers
• Coding-agent evaluations

Users remain in control of which sessions they contribute.

Business Model

Digg builds structured datasets from approved contributor sessions and licenses them to AI labs, model companies and AI data platforms.

Contributors are rewarded for the data they provide.

Our goal is to build a network where developers are not only users of AI, but also participate in the value created from the data used to improve AI models and agents.

Roadmap & Milestones

September 2026

• Launch Digg Code public beta
• Launch $DIG on Futard on September 11
• Start onboarding the first public developer contributors
• Bootstrap contributor rewards

Next Milestone

• Reach 3,000–5,000 contributors
• Build our first large coding trajectory dataset
• Validate dataset quality and contributor economics
• Start pilots and dataset discussions with AI labs and AI data companies

Next Phase

• Expand support across more coding agents and developer tools
• Launch Digg Light for broader computer-use workflows
• Build more RL-ready datasets, evaluations and verifiers

Market & Differentiation

Most AI data companies create training data by hiring people to perform tasks specifically for a dataset.

Digg starts with work that is already happening.

Developers are already spending hours every day working with Claude Code, Codex, Cursor and other AI agents.

These sessions contain more than just the final code. They capture how developers and agents:

understand → explore → act → fail → correct → verify → complete

Digg turns approved real-world coding sessions into structured training assets while allowing the people creating that data to participate in the upside.

Real coding work → structured trajectories → better AI agents → rewards for contributors.

Links

• 🌐 Website: digg.sh
• 𝕏 X / Twitter: @Digglabs
• 💬 Telegram: @karthikdigg
• 📄 Pitch Deck: View Deck
• 🎥 Product Demo: Watch on X