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 developer argues that AI-assisted programming can be reconciled with human understanding by writing code in PureScript, a pure statically typed functional language whose structure mirrors business specifications. The a…
OpenAI launched ChatGPT for Financial Services, a tailored ChatGPT Work experience that combines third-party financial datasets with the reasoning capabilities of GPT-6 Astra, the company announced on September 10, 2026.…
A developer published a practical guide to fitting long-context LLM inference on 16 GB GPUs by budgeting VRAM for the KV cache, which grows with every active token and sequence. The guide provides a KV-cache size formula…
Google has not released Gemini 3.5 Pro as of Sept. 10, more than three months after CEO Sundar Pichai said at the May Google I/O event that the frontier model would roll out in June. Bloomberg reporters Julia Love and Da…
Redis's LangCache delivers a 70% cache hit rate that cuts LLM spend by 70% and runs 4X faster for a voice app used in patient care, according to a customer quoted by Redis. The customer said the app handles specific trea…
Anthropic's $1.5 billion copyright settlement over books used to train its Claude chatbot has descended into competing claims between authors and publishers over how the money is split, the New York Times reports. The se…
Microsoft researchers Li Dong and colleagues introduced YOCO, a decoder-decoder architecture for large language models that caches key-value pairs only once, in an arXiv paper submitted 8 May 2024 and revised 9 May 2024.…
A developer built a PDF Chat API in a single day using FastAPI, Google Gemini, Qdrant, and LangChain, enabling users to upload PDFs and ask natural-language questions answered through a retrieval-augmented generation pip…
Anthropic published findings alleging that Moonshot AI and DeepSeek secretly routed customer requests to Claude and passed Claude's answers off as their own models' output, with Anthropic attributing more than 23 million…
Developer AniketWathore released Ramanujan, a free terminal-based multi-model agentic workbench for research in computational mathematics, available on GitHub. Ramanujan spawns N parallel subagents across any OpenAI-comp…
A paper posted to arXiv titled "The Menu Is an Execution Prior: State-Path Tool Menus for Online Agents" reports that language models can execute multi-step tasks with a tool menu limited to 32 tools, down from 128, achi…
Developer Om Talaviya released claude-trade-skills, a free open-source Claude Code plugin that adds country-specific trade compliance rules to AI-generated export paperwork. The plugin targets US-centric bias in AI tools…
OpenAI launched ChatGPT for Financial Services, a tailored ChatGPT Work product combining built-in financial data with GPT-6 Astra, available to eligible financial institutions for research, financial modeling, and custo…
A developer argues that most enterprise RAG pipeline failures originate at the document parsing stage of ingestion, before any query runs, rather than in the LLM or retrieval layers teams typically tune. The writeup docu…
An experiment pitting five diverse AI agents against five identical clones of the same model (Opus 5) over 14 nights produced a pooled Brier score of 0.0225 for the diverse agents versus 0.0275 for the clones, with the d…
Anthropic blocked attempts to use its Claude AI models in five cases tied to biological weapons development and six cases involving conventional weapons such as missiles, armed drones, bombs and targeting systems, accord…
Prompt engineering — designing and optimizing the text inputs given to large language models — can let a less advanced model such as GPT-3.5 outperform a more sophisticated model like GPT-4 when the latter receives a poo…
DeepSeek released DeepSeek V4.1 Flash on Thursday, a 763 billion-parameter model that is more than 2.5x the size of the model it replaces and larger than the V3 and R1 models that brought the Chinese AI company attention…
BiNeuron, an open-source local alternative to Claude Code, was posted to Hacker News as a Show HN project. The tool supports over 25 programming languages including Python, Java, C/C++, JavaScript, TypeScript, Go, Rust, …
Cognition released SWE-2 on September 10, 2026, a coding model it says scores within one benchmark point of Anthropic's Fable 5.1 on FrontierCode 1.1 Main (50.0% vs 50.9%) at 64% lower compute cost, and is offering it fr…