Ornith-1.0, a June 2026 open-weight coding model, claims to learn by building its own harness during training, and its 9B model matches or beats Gemma4-31B on SWE-Bench Verified and Terminal-Bench 2.1…
On March 24, 2026, malicious versions of the LiteLLM library (1.82.7 and 1.82.8) were hosted on the Python Package Index (PyPI) for approximately 40 minutes by threat actor Team PCP, exfiltrating 78,3…
Boards are now pressing startups on AI vendor concentration risk, asking founders to prove their business isn't dependent on a single model provider. OpenAI's August 2025 GPT-5 rollout briefly cut off…
HarnessRouter, a new open-source tool, provides a canonical API to switch between agent harnesses like Claude Code, Codex, and Hermes, using the Unified Harness Protocol (UHP) to standardize request/r…
A new guide from an unnamed author details how to run Qwen locally on Windows and access it remotely from any device using Ollama, LiteLLM Proxy, and Tailscale. The setup was tested with a Qwen 3.5 9B…
Zalando SE, a European e-commerce company, reported that its LiteLLM-based API proxy, deployed in January 2024, now serves 2,000 monthly active users with just six small pods, and its custom CLI tool,…
A dev.to write-up describes a setup where Claude Code is routed through a LiteLLM proxy to cheaper DeepSeek models via OpenRouter, a pattern that trades away data governance and supply-chain security.…
A developer has built a dual-backend setup for Claude Code, routing cheaper tasks through a self-hosted LiteLLM proxy to DeepSeek models while keeping a full Anthropic subscription for trusted work. T…
In benchmarks comparing proxy overhead, agentgateway 1.4.0 achieved 35,502 QPS with 0.863 ms P50 latency, while LiteLLM 1.98.0's Rust mode achieved 984 QPS with 32.139 ms P50 latency, making agentgate…
Anthropic's Claude Code can now run on-premises with AMD Instinct GPUs, serving GLM 5.2 at full quality via SGLang and LiteLLM, eliminating cloud API dependency and per-token costs. The setup uses an …
AI coding tools now generate 42% of committed code, projected to reach 65% by 2027, yet the security pass rate has stalled at 55% since 2023, according to Veracode's July 2026 GenAI Code Security Repo…
The TeamPCP campaign in March 2026 used a stolen token to cascade across five ecosystems in eight days, publishing backdoored LiteLLM versions 1.82.7 and 1.82.8 on PyPI that swept LLM API keys, cloud …
DSPy 3.3.0's MIPROv2 optimizer can automatically compile a support-ticket classifier prompt that beats a hand-tuned baseline, according to a tutorial by Mariana Souza. The approach uses a DSPy signatu…
A developer warns that the software supply chain for AI coding agents is vulnerable, citing a backdoored LiteLLM package on PyPI that was downloaded 47,000 times in three hours. The post details three…
A single misconfigured AI agent loop can burn through a month's cloud budget in an afternoon, according to production engineering guidance, and token budgets and rate limits are distinct mechanisms th…
RAGless, a retrieval-only question-answering system, eliminates LLM API costs at runtime by using vector retrieval with Gemini embeddings and a local Qdrant database, achieving zero hallucinations and…
Crusoe is now a native model provider in Pydantic AI, enabling developers to run agents on Crusoe Managed Inference with a single string like 'crusoe:zai/GLM-5.2'. The integration, contributed upstrea…
OpenRouter, an API proxy service, unifies access to 300+ LLM models from providers including OpenAI, Anthropic, and Meta with a single API key, standardized responses, and automatic fallback. It addre…
Microsoft's August 2026 Patch Tuesday fixed 421 CVEs, including two critical vulnerabilities affecting Visual Studio Code: CVE-2026-70336 enables remote code execution in VS Code, and CVE-2026-70335 a…
A developer investigating discrepancies between tracked LLM spend and provider invoices found that the two numbers almost never match, with causes documented in the tools' own docs. OpenAI and Anthrop…