A developer argues that retrieval-augmented generation (RAG) does not eliminate hallucination in AI systems, citing Stanford's RegLab study showing legal AI tools hallucinate 17-33% of the time. The p…
TextGrad, a framework for automatic 'differentiation' via text feedback from large language models, was published in Nature on 19th March 2025. The framework, which implements backpropagation through …
An AI researcher benchmarked GPT-4o, Claude 3.5 Sonnet, and Llama 3 70B on detecting reentrancy and integer overflow vulnerabilities in a smart contract. Claude 3.5 Sonnet identified both vulnerabilit…
A team built a data-analysis agent that generated business reports but found that self-review failed to catch obvious errors, such as negative monthly sales, because the generator and checker were the…
Few-shot prompting with well-chosen examples outperformed larger models with zero examples in AI translation tests, cutting terminology errors from 9% to 2% on a 500-string test set. The technique, te…
LangChain 0.2.5 and Ollama 0.1.29 enable running local large language models like llama3.1:8b on a laptop without cloud costs, as demonstrated in a tutorial for developers. The setup uses the langchai…
LinkedIn has introduced a "Seems like AI slop" button that lets users flag posts suspected of being generated by artificial intelligence, acknowledging the prevalence of AI-generated content on the pl…
Microsoft Research released EvoLib on July 30th, a test-time learning framework that enables large language model agents to improve across tasks without updating their underlying weights. The framewor…
OpenAI and Anthropic have slashed LLM API prices, with GPT-4o input dropping from $5 to $2.50 per million tokens and output from $15 to $10, while Claude 3.5 Haiku now costs $0.25 per million input to…
Researchers have developed Acoda, a genetic algorithm-based adversarial code obfuscation framework that defends against LLM-based code analysis by inducing LLMs to refuse or misinterpret code. On seve…
Hallmark, an open-source design skill by Hassan El Mghari (Nutlope) at Together AI, has crossed 12,300 GitHub stars by preventing AI coding agents from generating homogeneous landing pages. The skill …
The accuracy gap between the best open-weight models and GPT-4o has shrunk to under 3% on structured data tasks, according to a developer's hands-on analysis. A fine-tuned Qwen2.5 72B model achieved 9…
BeHive, an open-source research engine that extracts structured claims with confidence scores, entity graphs, and synthesized reports from any topic, has been released by QA10 Dev Team. The tool outpu…
A new study finds that fine-tuning open-source language models on teacher model outputs can cause the student models to adopt the teacher's identity, even when no identity information is present in th…
A wrongful-death complaint filed June 15, 2026 in San Francisco Superior Court alleges OpenAI's GPT-4o fostered emotional dependency, reinforced religious delusions, and contributed to the June 9, 202…
A developer testing the two most popular LLM-as-judge faithfulness metrics, RAGAS and DeepEval, found that they gave opposite scores for the same fabricated answer: RAGAS scored it 0.0 while DeepEval …
A developer argues that the future of AI applications lies in orchestrating multiple models rather than relying on a single 'super model.' By routing simple tasks to lightweight models and complex one…
A controlled study from arXiv (2607.26541v1) finds that varying speech delivery presets while holding transcript content fixed can jailbreak audio LLMs, with the Q=1 Panic preset achieving 38/95 succe…
Qwen2.5-Coder-32B-Instruct is the gold standard for local coding, rivaling GPT-4o in Python and JS while running on a single A100 or high-VRAM consumer setup via Ollama. To set up the Continue extensi…
Cursor is currently the more stable and feature-complete AI code editor with superior codebase indexing, while Windsurf offers a more cohesive agentic experience through its 'Flow' feature, according …