Full-text search across 66401 articles. Combine with topic and date filters; results sorted by relevance.
HoloViz released Lumen 1.4.0, an open-source framework for conversational data exploration in which an LLM proposes SQL, transformations and charts that stay visible, editable and re-runnable. The release lets `lumen-ai …
Moonshot AI released Kimi K3.1 Pro, an open-weight model in its 3T class with a one-million-token context window and a Highspeed version at 2 to 3 times the speed, live in Kimi Code and the Kimi API with full weights due…
Developer voska released bambu, an open-source command-line tool that slices, sends and monitors Bambu Lab 3D prints over a local network without a cloud account or Bambu Connect. The tool drives the Bambu Studio CLI wit…
A new benchmark called octobench, built by the team behind octomind's developer:reverse-spec tooling, scored the same open model at 24/25 with one coding harness versus 19/25 with another across 25 tasks harvested from r…
Two remote code execution vulnerabilities in PHP, tracked as CVE-2026-109618 and CVE-2026-109619 and marketed as "PHP2Shell," were found separately by security teams at Calif, as part of OpenAI's "Patch the Planet" progr…
Databricks cut its AI coding token costs by as much as 90% and held token spend per head roughly flat while total usage kept rising, after an initial policy of unlimited access to the most expensive models produced an ex…
Anthropic released Claude API version 2026-10-21 on 21 Oct, its first new API version since 2023-06-01, alongside new major SDKs including Python 2.0.0 and TypeScript 1.0.0. Requests on the new version that send temperat…
Wiz Research disclosed multiple critical memory-safety vulnerabilities in core PHP, grouped under two CVEs with a maximum CVSS score of 9.2, that allow unauthenticated remote code execution and sensitive memory disclosur…
The October 2026 SlopCodeBench update found that enabling XHigh reasoning effort on Claude Opus 5.5 costs 4.22x more while producing fewer correct results than Medium, with Opus 5.5 falling from 28 to 25 strict checkpoin…
Zscaler ThreatLabz reported in October 2026 that it traced an AI-assisted hacking campaign combining a customized ARTEX penetration-testing deployment with Claude Code, Codex and Grok workspaces, extending well beyond th…
GLM 5.3 Flash, Zhipu AI's lighter-weight model in the GLM-5 family, ran at roughly 50 to 60 tokens per second on a decade-old dual RTX 3060 Dell Precision T5810 workstation in a hands-on local test conducted through Proj…
Recursive harness-policy loops such as Google's Harness² and the Self-HPO framework break the performance ceiling of frozen-model agents, delivering 5-10% absolute gains on coding, reasoning and tool-use benchmarks, acco…
A developer published a FinOps post-mortem arguing that AI tooling architecture, not model intelligence, is now the primary cost bottleneck, citing context-window pollution and token evaporation from routing all tasks th…
A security assessment of LiteLLM version 1.77.7 uncovered multiple high-severity vulnerabilities, tracked as CVE-2026-35029 and CVE-2026-59823, that were resolved in LiteLLM version 1.83.9. The most significant flaw was …
Redis launched Redis Docs MCP, a public, unauthenticated MCP server at redis.io/mcp that exposes three tools — search, fetch, and ask — over the Redis documentation, built by four engineers over one quarter from April to…
Microsoft demonstrated a hybrid intelligence routing system in Windows and GitHub Copilot that sent roughly 1.6 million tokens to an on-device coding model while routing only about 10,000 tokens through the cloud, a rati…
A Nature paper introduces 'byteification', a two-stage conversion procedure that retrofits existing subword-tokenized large language models into byte-level models with minimal extra training. The resulting byte-level mod…
Andrej Karpathy published a GitHub gist titled "LLM Wiki" on April 4, 2026, describing a pattern in which an LLM builds and maintains a markdown wiki from raw sources while the human only curates sources and directs the …
Reflection announced Beam, a text-only 501B-total / 23B-active mixture-of-experts model for coding, agentic and scientific work, trained from scratch on 23.8T pretraining tokens with full weights under Apache 2.0 due thi…
Qualixar released SuperLocalMemory (SLM), a local-first long-term memory tool for MCP clients including Claude Code, Codex and Cursor, installable via `npm install -g superlocalmemory` on Node 18+ and Python 3.12+, or `p…