Sahara AI, a decentralized-AI startup that pays people to gather and label AI training data, said it produced what it calls the "longest real-world human-to-human conversation framework ever collected for AI research" for Snapchat, which runs a consumer chatbot.
The dataset, called REALTALK, began as a customer request from Snap, the parent of Snapchat, according to a blog post Sahara published Thursday.
"Snap's researchers needed matched pairs of real people talking naturally, every single day, for three straight weeks, with enough structure behind it to measure what an AI can remember and how well it reads emotion," according to Sahara.
The methodology, according to the project:
Sahara AI recruited ten participants, all native English speakers aged 18 to 25, and paired each person into two separate conversations with two different partners. Because every speaker appears in two chats, the data shows how the same person shifts tone, closeness, and emotional register depending on who they're talking to. Researchers need that to study whether an AI can hold a consistent personality across different relationships.
And the result:
The final corpus contains nearly 160,000 words of authentic human conversation across roughly 9,000 exchanges, more than 300 shared images, and 21 daily sessions per pair. Each conversation runs around 14x longer than the human-collected benchmarks the industry has relied on until now.
The takeaway:
For the first time, real human conversation could be measured directly against AI-generated conversation. The results explain exactly why chatbots feel fake.
Sahara, led by chief executive Sean Ren and co-founder Tyler Zhou, a former Binance Labs investment director, is building a blockchain where people contribute data, models and compute and get paid on-chain. It raised a $6 million seed in March 2024 and $43 million that August from backers including Binance Labs, Pantera and Polychain; the SAHARA token trades on Ethereum and BNB Chain.
The project is one of several putting data ownership on a blockchain, alongside chains such as Vana. Sahara has also released open-source "agentic" protocols built on x402, the agent-payments standard DeAI News has tracked spreading across the agent economy, from Coinbase's bazaar of pay-per-call AI services on down.
(HOW AI WAS USED IN THE PRODUCTION OF THIS PIECE: This story was drafted on Claude Desktop using Distro's blog-rewrite skill, with background research on Sahara AI's funding, founders and technology gathered via web search and Sahara's own site and docs. I then filed the story as a draft to DistroVerse using our Distro Publisher MCP server, and I edited the piece on our DistroVerse native (non-AI) story editor. All facts were checked against the original blog post and cited public sources before publication.)