Articles
Industry news, technical articles, and product introductions
📚 Claude Tutorials Hub
40+ step-by-step Claude guides — prompt engineering, Claude Code, API, agents. Browse by topic →
Loading...
Industry news, technical articles, and product introductions
📚 Claude Tutorials Hub
40+ step-by-step Claude guides — prompt engineering, Claude Code, API, agents. Browse by topic →
Loading...
A supply chain attack on keyv and 10 related npm packages infected 434 packages with 2B+ monthly installs. The worm steals npm tokens, GitHub PATs, AWS credentials, Kubernetes secrets, and more. Plus: why the Airtable acquisition is the same risk in business clothes. Five things to do right now.
Here's the short version: **Pro is worth it for agent workloads. Flash has a hidden token-eating habit you won't notice until a long task eats your context.**
Same machine, same user, same global config. Two Claude Code windows running the same CLI — one worked, one didn't. The obvious suspect was the permission classifier, which really does block commands. But the cause sat a layer down: the daemon's plist was installed and never loaded, so every gateway-bound subcommand died while purely local ones printed fine. That half-working shape is what sells the permission theory. Full trace, including the hypothesis I got wrong by misreading my own logs.
70B on a single 4GB GPU, or small models at scale on the edge? AirLLM and Cloudflare show two roads to cheaper inference. A no-hype comparison of what each trades, and where each fits.
How to write a correct mcp.json for Claude Code, Cursor, VS Code and Claude Desktop: file locations, root keys, stdio vs HTTP transport, the #1 command/args mistake, and how to validate configs without committing API keys.
How much VRAM do I need to run a local LLM? This guide breaks down the four factors that decide it — weights, quantization, KV cache, and overhead — with a rule of thumb (0.6 GB per billion params at Q4_K_M), real examples (DeepSeek R1, Kimi K3), and the exact math from our free LLM VRAM calculator.
Three releases in one week—Reasonix, ds4, and Kimi K3 on AMD MI355X—all pointing in the same direction: LLMs are moving from cloud APIs to persistent local processes. Here's why it matters and what to do about it today.
GitHub shipped gh-stack, making stacked pull requests an official workflow. No more third-party tools, no more long-branch hell. Quickstart + gotchas inside.
Three lines of nginx config for real client IPs took down an entire VLESS node. This post breaks down PROXY protocol asymmetry, over-trusted real_ip ranges, and stream-vs-http module confusion — with working config templates and verification steps.
A week in, the verdict on Claude Opus 5 is unusually split. CodeRabbit's 96-real-bug benchmark found the highest precision it has ever measured — alongside lower recall and 4x the nitpicks. Claire Vo coined 'neurotic AF' and 'Claudeslop' to complain about it, then ranked it first in her own blind test. Read together: precision up, recall down, verbosity, timidity, and refusing to touch someone's branch are five readings of one knob. Plus a practical guide to choosing failure modes.
Anthropic's unreleased Claude Mythos model spent 60 hours and roughly $100,000 of compute halving the best known analysis of post-quantum candidate HAWK. The next day, HAWK's team withdrew from NIST standardization. Inside the 48 hours: why attacking the 256-bit version killed the whole scheme, the model that refused to try until encouraged for three days, the sober expert assessments — and the real story: discovery took 60 hours, verification took a month.
Two milestones in one day. The largest MCP spec revision ever shipped: the initialize handshake and session IDs are gone, the core goes stateless, and Roots, Sampling, Logging, and HTTP+SSE enter a 12-month deprecation countdown. The same day, gateway startup Runlayer sued Rippling, alleging a year-long 'customer trial' that ended in a cloned product — MCP's first commercial lawsuit. What breaks, what each kind of developer should do, and why these two events read as one story.