LIVE

Kapso Launches Iterative AI System to Automate Enterprise Data Operations

🚀 Startups & Funding 1 min read 👁 7

A new developer tool named Kapso applies continuous experimentation to help enterprise teams handle complex data engineering and predictive modeling tasks.

🚀Startups & Funding

Key points

  • Kapso automates enterprise data work using iterative experimentation and sandboxed testing.
  • The platform maintains a git-backed lesson bank that ranks and updates hypotheses based on past trial evidence.
  • It scored top results on benchmarks including Stanford's RelBench and OpenAI's MLE-bench.

Data management in large organizations traditionally requires extensive human planning, meeting-heavy coordination, and predefined business taxonomies that can stall projects for months. Addressing these bottlenecks, creators introduced a new AI-native system designed to automate iterative data investigations through repeated experimentation rather than upfront documentation analysis.

The system operates by taking a target objective and success metrics, then executing structured trials within sandboxed environments. It stores historical findings in a git-backed lesson repository, evaluating each past outcome with a credence score and applying these insights to future ideation cycles. An independent verification module reviews experiment results to prevent bias, while the underlying architecture remains agnostic to specific large language models and coding agents.

According to the project's developers, the platform achieved notable results across several established benchmarks. It secured the top position on Stanford's RelBench for relational database predictive modeling, outperformed prior open-source configurations on OpenAI's MLE-bench, and demonstrated competitive performance during the International Olympiad in AI 2026.

What to watch next: Track how enterprise engineering teams adopt sandbox-driven, iterative learning frameworks compared to traditional retrieval-augmented generation tools.

Source Hacker News Front Page — news.ycombinator.com/item?id=50471281 ↗

Contact the newsroom

Got a tip, a proposal or a partnership question? Write to us and we will reply shortly.