Anthropic said it disrupted "malicious use" of its Claude Haiku, Sonnet, and Opus models between December 2025 and August 2026, including attempts that could support biological and conventional weapon…
Anthropic's fourth misuse report, covering December 2025 through August 2026, documents Russian-linked espionage and AI-orchestrated cyberattacks in which multi-agent systems ran much of the cyber kil…
Anthropic's latest threat intelligence report disclosed that state-linked actors in Mali, China and Iran used its Claude models — specifically Claude Haiku, Sonnet and Opus — to automate surveillance …
Anthropic's Threat Intelligence team disrupted malicious operations in which threat actors used Claude across seven harm areas between December 2025 and August 2026, according to the company's Septemb…
Developer oguzbilgic released Aish, a 300-line Zsh plugin that provides Warp-style AI command suggestions after every shell command, predicting the next command based on failure status, output, and pr…
Researchers released RegDivergence-101, a 101-pair benchmark for detecting regulatory contradictions between FDA and EMA guidance in life sciences, with an LLM judge (Claude Haiku) achieving 0.830 mac…
The Citizen's Daily Brief, an AI-powered news briefing service, operates on a total budget of about $60 per month, or roughly $2 per day. The system, which ingests 34 sources, clusters stories, and sy…
A developer repriced three months of Claude Code usage—39.5 billion tokens across 139,835 requests—and found that routing to cheaper models could cut costs by 50-58%. They built OmnisRouter, a cost ro…
A new arXiv study (2608.20371v1) comparing zero-shot Claude Haiku against fine-tuned NLU models for intent detection finds that fine-tuned RoBERTa beats Claude by 11.8 points on ATIS (95.9 vs. 84.1, p…
Qwen 3.8, a 27B-parameter open-weights model from Alibaba, delivers coding output in the Claude Haiku-to-Sonnet range but is 15-20 times slower than Claude Sonnet, according to tests by a developer on…
A developer built a model router on AWS AppConfig and Amazon Bedrock that lets users swap large language models without redeploying code. The router uses feature flags to map routing keys like 'fast',…
A feature built by Soamee for a SaaS client, costing less than $30 per month to run, reduced support tickets by 40% in the first eight weeks. The system uses basic RAG with OpenAI's text-embedding-3-s…
Anthropic reported degraded performance for multiple Claude models, including Claude Opus, Claude Sonnet, and Claude Haiku, on May 13, 2025. The incident caused increased latency and errors for users,…
Researchers demonstrated that hidden reasoning blocks in LLM APIs from Google, OpenAI, and Anthropic can be extracted via a replay attack, achieving near-perfect recovery of up to 12,000 tokens. The a…
OpenAI and Anthropic are cutting prices for their AI models in response to intensifying competition from Chinese rivals such as DeepSeek, according to the Financial Times. The US companies have releas…
A developer built an n8n workflow that automatically rejected 218 of 312 job applications before a human recruiter reviewed them, cutting recruiter time from about nine hours per week to 90 minutes. T…
Anthropic's Claude AI assistant enforces two independent rolling rate limits—a 5-hour session limit and a 7-day weekly limit—that reset based on the user's first message timestamp, not a fixed schedul…
A developer built LeadPilot, an AI lead-automation CRM that scores and qualifies leads instantly using Claude Haiku, and deployed it across five services on Zerops in 48 hours. The system queues leads…
AI2 (the Allen Institute for AI) released TMax, a recipe for terminal agents that boosted a Qwen model's Terminal-Bench score from roughly 25% to 31%, closing ground on Claude Haiku's roughly 33%, by …
Wiring four Claude models together in Claude Code to solve Terminal-Bench 2.1's 89 command-line tasks backfired in four ways, yielding a 78% solve rate and seventh place at a cost of $1,178, roughly t…