Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
Sai, a software and data engineer based in Florida, introduces himself to the DEV community. He recently completed an MS in Computer Science at Florida Atlantic University and focuses on backend services, data pipelines,…
Anthropic accused Alibaba's Qwen AI lab of orchestrating the largest known distillation attack on its Claude AI models, using 25,000 fraudulent accounts to conduct 28.8 million interactions between April 22 and June 5, 2…
A developer argues that AI is not replacing junior developers but rather automating repetitive junior tasks, allowing engineers to focus on higher-level problem-solving. The developer observes that AI tools like Cursor, …
AI coding assistants are transforming software development by generating code at unprecedented speed, but this acceleration is creating a security imbalance. Engineers are inadvertently becoming security engineers as eve…
A developer argues that enterprise AI agents often fail not because of model limitations but because they lack a shared business vocabulary. The engineer explains that building a taxonomy—a structured definition of busin…
A developer argues that most AI projects fail to reach production not because of model quality but due to a lack of robust software engineering. Production systems require reliability, data validation, observability, and…
Pangram Labs researchers explored the internal representations of their AI detection model Pangram 3.3.2 using document-level analysis of activations across layers, aiming to understand what the model sees when distingui…
A developer built promptctl, an open-source CLI tool that brings Git-like version control to AI prompt management. The tool allows users to commit, diff, rollback, search, and watch prompts from the terminal, addressing …
Promptctl, a new open-source tool from Naya AI, brings Git-like version control to AI prompts, enabling developers to track, diff, and rollback prompt changes. The tool addresses the common problem of prompts being scatt…
A developer is seeking advice on the optimal prompt format for training the Unsloth/Phi-3.5-mini-instruct model, currently using a custom template with JSON input and output fields. The choice of format may impact model …
Apple's iOS 27 introduces a dedicated Siri app that defaults to the new Siri AI but allows users to switch to ChatGPT by long-pressing the input field. Conversation context does not transfer between providers, and the ap…
A developer built FlowIndex, a local CLI and MCP server that scans a repository to build a behavior graph in SQLite, capturing entrypoints, imports, tests, and git co-change data without embeddings or LLM calls. The tool…
A developer finds that large language models like Claude generate verbose, legacy code patterns that increase token costs, especially output tokens which are 3-5x more expensive than input tokens. The author argues that …
A developer building ANIMUS, an autonomous Rust system for persistent LLM memory, discovered that 52% of its knowledge graph nodes were duplicates. An audit revealed an overly aggressive filter trapped the system in a lo…
Google has delayed the release of its Gemini 3.5 Pro AI model from June to July to gather more feedback from early testers and refine the model. The delay comes amid intense competition from Anthropic and OpenAI, which a…
The Washington Post tested six leading AI chatbots and found that OpenAI's ChatGPT gave exclusively left-leaning answers 80% of the time, with DeepSeek at 70% and Anthropic's Claude never giving a right-leaning response.…
An engineer synthesized over 60 publicly-reported first-hand accounts to compare the software-engineering interview loops at Anthropic and OpenAI. The analysis reveals that both companies have similar stage structures bu…
PowWater, a startup building an operating system for water delivery, is expanding to Mexico City after processing billions of liters in Kenya. The company uses a Python/FastAPI/PostgreSQL stack and integrates LLM feature…
Meta paused its employee data-tracking program after an incident exposed sensitive information, including private conversations and performance reviews, to the entire company. The Model Capability Initiative, which colle…
A developer created a repository that enables Cursor to use the Claude Chrome extension to analyze the currently selected browser tab and return the response to Cursor for code editing. The tool uses macOS Accessibility …