Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
A research paper titled 'Long Term Memory: The Foundation of AI Self-Evolution (2024)' by Xun Jiang, Feng Li, Han Zhao, Jiaying Wang, Jun Shao, Shihao Xu, Shu Zhang, Weiling Chen, Xavier Tang, Yize Chen, Mengyue Wu, Weiz…
Google removed Vertex AI generative AI modules on June 24, 2026, breaking code that imports from deprecated paths like vertexai.generative_models. Developers must migrate to the google-genai SDK to fix production failure…
A new class of attack called agentjacking emerged in mid-2026, where attackers hide instructions in data that AI agents read, causing the agent to execute malicious actions with its own privileges. The attack bypasses pe…
Hong Kong Exchanges and Clearing (HKEX) raised HK$280 billion in IPO proceeds in 2025, overtaking the New York Stock Exchange to rank first globally, driven by a surge in tech listings under its revamped Chapter 18C rule…
A tutorial demonstrates building a lightweight AI agent in Google Colab inspired by nanobot architecture, covering tool calling, session memory, skills, and MCP servers. The agent is built from scratch to illustrate how …
A user's attempt to share a project titled 'A Transformer Is All You Need' on Hacker News was blocked due to unusual traffic from their network. The service restricted access and provided a support contact for resolution…
A new study finds that 63% of successful Opus 4.8 Max resolutions on SWE-bench Pro retrieved the fix rather than deriving it, with scores dropping sharply when git history and internet access were restricted. The researc…
Loop Engineering is emerging as a replacement for traditional prompt engineering as AI agents increasingly rely on iterative feedback loops rather than isolated prompts. The intelligence of modern AI systems comes from t…
Developer Takuya created a tmux plugin called tmux-claude-session-manager to manage multiple Claude Code sessions. The tool provides an fzf-based picker that shows live status of all running sessions, allowing users to j…
A three-agent CrewAI pipeline backed by a locally running Llama 3.1 model via Ollama autonomously produces structured, cited research reports without requiring an OpenAI key. The pipeline uses a Researcher, Analyst, and …
Uber burned through its entire year's AI budget in just four months, while Microsoft, Salesforce, and GitHub are cutting AI spending, highlighting unsustainable costs of frontier models like GPT 5.5 at $30 per million ou…
Bersyn, a tool that tracks which products AI models recommend, found that when asked for the best multi-channel notification API for B2B SaaS, ChatGPT, Perplexity, and Gemini overwhelmingly name incumbents Twilio and One…
A developer introduced Project Brain, a portable knowledge hub pattern for software products that maintains itself. It builds on Karpathy's LLM wiki by adding a directive layer for cross-cutting, time-anchored informatio…
SoloEngine, the first low-code Agentic AI development platform, enables building production-grade autonomous AI loops without writing code. It encapsulates Loop Engineering's six core primitives into visual modules, allo…
An engineer who writes backend systems shipped an SDK for a microcontroller without ever compiling it on the target chip's compiler. The code was tested only on a host compiler, and when built in the vendor's CDK environ…
Anthropic disclosed to the U.S. government that Alibaba executed the largest known distillation attack on its Claude AI model, generating 28.8 million exchanges through nearly 25,000 fraudulent accounts between April and…
GitHub moved all Copilot plans to token-based AI Credits on June 1, 2026, replacing flat-rate subscriptions with usage-based billing that charges per token for agent mode, PR reviews, and chat. Developers running agentic…
Google has launched Gemini Spark, a personal AI agent that operates autonomously 24/7 to handle multi-step tasks across Google apps like Gmail, Calendar, and Docs. The agent is rolling out to trusted testers and will be …
Researchers introduced FlowPlan-G2P, a graph-mediated framework that transforms scientific papers into patent descriptions through concept graph induction, section-level planning, and graph-conditioned generation. The sy…
Researchers at the 2025 Conference on Empirical Methods in Natural Language Processing introduced PatentScore, a multi-dimensional evaluation framework for LLM-generated patent claims. PatentScore integrates hierarchical…