Anthropic's Internal Claude Usage Reportedly Tops Employee Pay at Retail Prices

·1 views

A Business Insider report says Anthropic's internal use of its AI models, valued at standard retail token prices, now exceeds what the company pays its employees. The figure was described as a striking indicator of how heavily staff rely on AI tools in their daily work. The article was shared on Hacker News, where it drew modest attention and discussion. Only the headline and link were available to this summary, so specific figures and methodology could not be independently verified. The item highlights the growing scale of AI consumption within leading AI companies.

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)

Vibe-Coding a Game in 2026: What AI Builds and Where It Breaks

A blog post on vlstudio.app examines the practice of building a video game largely through AI-assisted "vibe coding" in 2026. The article reportedly describes which parts of game development AI tools handle well and which areas tend to fail. It was shared on Hacker News, where it had gained 5 points and no comments at the time of capture. The source text provided is limited to the headline and link metadata, so the specific findings are not detailed here.

0 comments Read more at Hacker News

ProgrammingHacker News ·

Subvocal Startup Builds Glasses That Decode Silently Mouthed Words Into Text

Two founders from Georgia Tech, Aarush and Arnav, have developed smart glasses that convert silently mouthed words into text on a computer. The idea grew from frustration at being unable to use voice tools in libraries, offices, or public spaces. After a 2025 hackathon prototype using facial electrodes proved impractical, the team switched to radio frequency sensing, with antennas embedded in the glasses frames. Their foundation models were trained on data gathered from paid participants wearing the glasses, which improved performance across different users. The company, The Subvocal Company, is part of Y Combinator's F26 batch and raised pre-seed funding from Afore Capital.

0 comments Read more at Hacker News

ProgrammingHacker News ·

Momentic Launches AI Agents That Test Web and Mobile Apps by Exploring Them

Momentic, a Y Combinator W24 startup, has launched QA agents that autonomously plan, explore, and bug-bash web, iOS, and Android apps to surface defects. Its founders, both former engineers from companies like Robinhood and Retool, built the tool to replace the brittle, hand-maintained scripts typical of Playwright and Selenium. The company's custom agent harness, called Mo, uses a self-organizing graph of parallel sub-agents, each running its own browser or device. In a 60-minute benchmark across four open-source apps, Mo found 269 bugs with 84% precision, far more than Codex's 86 bugs at 86% precision, though at a higher cost per bug.

0 comments Read more at Hacker News

ProgrammingDEV Community ·

Salesforce Screen Flow Data Tables can now display Apex collections without a custom object

Starting with Winter '26 (API 65.0 and later), the Screen Flow Data Table accepts a collection of an Apex-defined type as its Source Collection. Developers designate a field as the Unique Identifier, which lets the table treat each item like a record that users can select and the flow can reuse. Display and row selection are supported now, but inline editing is not yet available. Previously, developers had to create mirror custom objects, insert and delete temporary records, use Transform elements, or build custom Lightning Web Components to achieve this. The article outlines the setup steps, including @AuraEnabled fields, an invocable wrapper class, and key selection guidance.

0 comments Read more at DEV Community