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 Cornell paper reveals that retrieval-augmented generation (RAG) systems are vulnerable to a poisoning attack where malicious documents can manipulate the model's output, undermining trust in RAG-based applications. The…
Anthropic released Fable 5, its highest-capability public model, alongside Managed Agents updates and a Workflow SDK cancellation primitive. The Workflow SDK now supports AbortSignal for cooperative cancellation in long-…
A developer released hyperpb, a runtime-compiled Protobuf parser for Go that achieves 10x faster parsing than dynamicpb and 3x faster than hand-written generated code. Quickwit, a Rust-based log search engine, was acquir…
A developer released 1pager, an AI skill that condenses long documents, directories, or chats into a strict one-page summary with bullet-first formatting and minimal verbosity, addressing the common problem of verbose AI…
Cloudflare rearchitected its Workflows control plane, replacing a single Account Durable Object bottleneck with two new components, SousChef and Gatekeeper, to scale concurrent instances from 4,500. The update signals th…
The Continue plugin v1.2.20 patches memory leaks, unhandled exceptions, and JCEF message chunking crashes across JetBrains and VS Code adapters, fixing crash vectors that cause sidebar hangs and autocomplete failures. A …
Anthropic announced that Claude Code now supports artifacts, enabling users to turn session work into live, shareable web pages such as dashboards and release checklists. The feature is available in beta for Claude Team …
Swift's server ecosystem reached two 1.0 milestones with the release of Swift System Metrics, providing stable APIs for CPU, memory, and file descriptor monitoring. Rust 1.95 introduced new zeroed allocation APIs and ite…
Shrijith Venkatramana, building git-lrc, explains that modern AI systems rely on a two-step recipe: generative pre-training followed by discriminative fine-tuning. Pre-training involves predicting the next token on massi…
Together AI released DeepCoder-14B, an open-source 14B model that matches o3-mini on coding benchmarks, with full training recipe and dataset published for reproducibility. Biome v1.5.0 introduced a --changed flag for VC…
Node 26.3.1 patches 11 CVEs in TLS and WebCrypto, requiring immediate updates for untrusted code execution. Ruff v0.7 fixes inconsistent pytest rule defaults from v0.6, with auto-applicable fixes via ruff check --fix. Co…
A developer released Python and Node.js SDKs for Ajah, an open-source self-hosted LLM observability gateway. The SDKs simplify integration by replacing manual header injection with a single import, enabling features like…
Researchers introduced MosaicLeaks, a benchmark revealing that deep research agents leak private information through web queries, and proposed Privacy-Aware Deep Research (PA-DR) training that reduced leakage from 34.0% …
Anthropic executives have proposed to Commerce Secretary Howard Lutnick to lift the US ban on its powerful 'Mythos' and 'Fable' AI models, pledging closer cooperation with the White House to address security concerns. Th…
OpenAI removed GPT-5.2 and GPT-5.3-Codex from Codex for ChatGPT subscribers around June 2, 2026, citing fleet simplification. Users lost access in VS Code and CLI without prior notice, though the models remain available …
Google sunset Gemini CLI and Gemini Code Assist IDE extensions for individual users on June 18, 2026, citing a shift toward multi-agent systems. Affected users are directed to migrate to Antigravity CLI, which retains ke…
OpenAI researcher Karan Singhal is leading the company's push to make ChatGPT a trusted healthcare tool, with over 230 million weekly users seeking health advice. Singhal, formerly of Google, is building a team and partn…
Continuous batching, introduced in the 2022 Orca paper, improves GPU utilization during LLM inference by dynamically updating the batch at each iteration, freeing slots as requests finish and immediately adding new ones.…
A Claude AI Professional E-Degree is available for $19.99 (regularly $49), offering over 8 hours of video content across four courses to teach advanced prompt engineering, workflow automation, and integrations with tools…
Michael Saylor, chairman of Strategy, used OpenAI's ChatGPT to help design the company's STRC perpetual preferred stock, which has amassed over $8.5 billion in market cap since its July 2025 launch. The 11.5% variable di…