Artificial Analysis Releases Version 5.0 of Its Intelligence Index Methodology

·2 views

Artificial Analysis has published version 5.0 of its Intelligence Index, a framework for benchmarking AI model intelligence. The methodology is detailed on the company's website, which outlines how models are evaluated. The source material available here is limited to the announcement link and a Hacker News discussion thread with modest engagement. Specific changes from earlier versions and detailed scoring criteria are not described in the available text.

Read the full story at Hacker News

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

USDD Gas Audit Finds 15-25% Savings Potential, Low Direct Security Risk

An independent gas-efficiency audit of the USDD stablecoin, which has roughly $1.17 billion in total value locked across Ethereum and several Layer 2 networks, was completed on 28 October 2026. The review identified 36 inefficiencies spanning storage layout, loops, external calls, and transfer logic, with the largest costs arising in bridge calls and whitelist iteration. Optimizing the most-used functions such as mint, burn, and rebalance could cut gas costs by an estimated 15-25%. The auditor found no critical security flaws directly caused by these inefficiencies, though unbounded loops could enable denial-of-service via gas griefing. The overall risk score was assessed at 3 out of 10, indicating low-to-medium economic risk.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Blik360 Marks First Year with Paying Customers and Multilingual Expansion

The developer behind Blik360, an open-source 360 performance review tool, reflects on its first year, which began as a prototype built to support Dreyfus-model employee reviews. The tool's defining feature is strong reviewer anonymity, designed so that even application administrators cannot easily link answers to individuals. Despite releasing the software as open source, the project has attracted paying customers for its hosted service, along with contributors who added translations and companies running trials. Feedback from these users prompted a redesign of the interface, including drag-and-drop questionnaire ordering, improved filters, and a reworked menu. The developer is now seeking input on future features such as PDF reports, AI-assisted report queries, and additional questionnaire frameworks, and asks readers what makes a 360 review worth pursuing.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer's 'Séance' App Lets Users Chat With Simulated Dead Software

A developer built a small local web app called It Looks Like You're Trying to Hold a Séance for a Hacktoberfest challenge. Users create a summoning card describing a piece of defunct software, and a locally run open-weight model via Ollama plays that spirit in character. The spirit is restricted to the information the user supplies, so it admits when it does not know something rather than inventing answers. The app also includes Clippy, who appears uninvited and gradually decays as the conversation goes on. Everything runs on the user's own machine with no cloud service or account required.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Meta Sets Per-Country WhatsApp API Message Pricing for 2026

Meta bills businesses using the Cloud API per delivered message, with no charge for messages that fail to deliver. The cost depends on the message category, the recipient's country, and, from 1 October 2026, the number of service replies each business phone number sends in a month. Rates are set by country or regional rows in a published rate card, with utility and authentication messages also qualifying for volume-based discounts. Meta only changes rates on quarterly dates, so a budget calculated from list prices serves as a stable upper bound for a full quarter. A developer has published a Python method that estimates monthly bills from an open CSV copy of the rate card.

0 comments Read more at DEV Community