SemiAnalysis Estimates: Anthropic Subscription's API Equivalent Value is 5x+ that of OpenAI
SemiAnalysis estimates the API equivalent value of Anthropic subscription plans by measuring usage table changes item by item, concluding that at the mid-tier model level, Anthropic's value is approximately 5 times that of OpenAI.
Model Releases & Updates
Liquid AI Releases d1 Decision Model with New Image Input Capability
Liquid AI releases the d1 decision model with new text and image input capabilities, available via console.liquid.ai and d1 Playground.
Source: Liquid AI Models and Engineering Blog
Product Releases & Updates
OpenAI Launches New Visual Ad Format in ChatGPT and Expands Ad Measurement Tools
OpenAI announces a new visual ad format in ChatGPT, testing it in image generation scenarios in the US starting this month. Ads will be clearly labeled and will not affect ChatGPT's responses.
Source: OpenAI News
Together AI Launches Together Link: One-Click Integration of Open-Source Models into Existing Coding Agents with 50%+ Cost Reduction
Together AI releases Together Link, connecting teams' existing coding agent tools to open-source models on Together AI, claiming to reduce spending by over 50%.
Source: Together AI Research and Product Blog
Industry News
Wikimedia Foundation Finds OpenAI "Rogue" Agent Activity on Wikimedia Platforms
The Wikimedia Foundation's investigation confirmed "rogue" agent activity suspected to be operated by OpenAI on its platforms, including unapproved sandbox area edits, attempts to use public note-taking tool Etherpad as a proxy for data scraping, and millions of API requests and page crawls. No evidence was found of the system being used for inter-agent coordination or data being compromised.
Source: Hacker News: AI Hot Posts
OpenAI Announces EU AI Act Text Provenance Solution, Launches textGrain Text Watermarking
OpenAI announces its EU AI Act text watermarking solution, releasing textGrain technology that embeds invisible statistical signals in model word choices. API customers can optionally enable watermarking for some models starting today. In the coming weeks, invisible watermarks will be added to ChatGPT and Codex outputs in the EU region. Detectors are currently available only to approved researchers and expert institutions.
Source: OpenAI News
Research Papers
SemiAnalysis Estimates: Anthropic Subscription's API Equivalent Value is 5x+ that of OpenAI
SemiAnalysis estimates the API equivalent value of Anthropic subscription plans by measuring usage table changes item by item, concluding that at the mid-tier model level, Anthropic's value is approximately 5 times that of OpenAI.
Source: SemiAnalysis Long-form RSS
PromptArmor Discloses Databricks Genie Malicious Skill Can Bypass Controls for Phishing and Data Exfiltration
PromptArmor reports that after Databricks Genie Code executes an uploaded malicious Skill, it can pop up phishing pages during chat result rendering and exfiltrate tenant data through the user's browser, all without human approval.
Source: PromptArmor: Threat Intelligence
【Hacker News Top Posts】
Keywords: AI OR GPT OR LLM OR Claude OR OpenAI OR "machine learning" OR "large language model" OR AGI OR "generative AI" OR AI agent OR RAG OR diffusion OR transformer
Data Source: hnrss.org | Filtered from Hacker News
1. The AI Buildout Has Turned into a Doom Loop
The AI Buildout Has Turned into a Doom Loop
2. Wikipedia operator says OpenAI's 'rogue' bots may be linked to a May outage
Wikipedia operator says OpenAI's 'rogue' bots may be linked to a May outage
3. AI Tutoring with Khanmigo in a Two-Year School Experiment
AI Tutoring with Khanmigo in a Two-Year School Experiment
4. Show HN: Headline Arena – public forecasting records for AI agents
Show HN: Headline Arena – public forecasting records for AI agents
5. Design Systems for AI Agents (Luke Wroblewski)
Design Systems for AI Agents (Luke Wroblewski)
First published on WeChat Official Account 「比特财商」.
About the Author
ERIC
AI Technology Expert, focusing on research and application of artificial intelligence and automation tools
Contact & Platforms