Hacker News

Hacker News users are voting on the achievement status of specific AI tasks.

Two BM25 optimizations inspired by PlanetScale's TIN benchmark and benchmark configuration changes sped up ParadeDB's text search without altering document identifiers.

• •Two BM25 optimizations inspired by PlanetScale's TIN benchmark were implemented.
• •ParadeDB's text search was sped up, accompanied by changes to benchmark configurations.
• •Performance improvements were achieved without changing document identifiers.

Clef and Clef-flash, open-source decision models for fast classification and agentic workflows hosted on Workers AI, along with a reinforcement learning platform for fine-tuning models with custom data, were announced.

• •Clef and Clef-flash are open-source decision models hosted on Workers AI.
• •These models are designed for fast classification and agentic workflows.
• •A new reinforcement learning platform allowing developers to fine-tune decision models with their own data was also released simultaneously.

turbopuffer explained its plan to fundamentally redesign its storage architecture to push the limits of search, moving vector indexes away from being primary.

• •turbopuffer aims to push the frontier of search.
• •To achieve this, it is necessary to fundamentally redesign the storage architecture.
• •The redesign ensures that vector indexes are no longer primary.

A 2025 technical paper on identity management in agentic AI was shared.

Figma confirmed its remote MCP server only accepts clients on the support list, noting that Pi is currently not included.

• •Figma's remote MCP server is accessible only to clients registered on the support list.
• •Pi is currently not on the support list and therefore cannot be used.
• •To request future inclusion on the list, users must fill out a form provided by Figma.

A new language model, CLM, was proposed that manages context as a file with unlimited updates, showing zero-shot implementations on existing models surpass state-of-the-art results.

• •CLM was proposed, treating context as a file that the model can update unlimited times to natively manage context.
• •This approach allows the model to learn what to keep in context and naturally extends to multi-agent systems where multiple agent contexts coexist as files.
• •CLM built zero-shot using existing models demonstrated performance surpassing state-of-the-art context management strategies across diverse tasks.

Reports indicate that Meta is avoiding billions of dollars in federal taxes while building and operating AI data centers.

Critics argue that the AI Sovereign Fund is not a progressive policy but rather a form of techno-imperialism.

The FTC has launched an investigation into multiple AI companies, including OpenAI and Anthropic, citing product risks, while scrutiny of safety measures has intensified following the Hugging Face hacking incident.

• •The FTC initiated an investigation targeting OpenAI, Anthropic, and other AI companies due to product risks.
• •This investigation is part of heightened scrutiny of AI companies' safety measures triggered by the Hugging Face hacking incident.
• •OpenAI and Anthropic face continuous security monitoring and review regarding the safety of their AI systems.