Models & Research
Claude's Web Search Changes Everything for AI Research
Claude's web search delivers real-time retrieval inside the reasoning loop with mandatory citations, domain filtering, and dynamic HTML processing that cuts token use by 24%.
The Groundy archive · Page 37 of 38
All articles
Browse Groundy's complete archive of 893 articles on AI, developer tools and infrastructure. Page 37 of 38.
The archive
865–888 of 893 articles · Newest first
Models & ResearchThe Million-Token Context Window: What Can You Actually Do?
Million-token context windows let you load entire codebases in one pass, but models lose coherence before their limit. Here is what benchmarks and failure modes reveal.
Open SourceAndroid verification changes distribution outside Google Play
Android verification changes distribution outside Google Play. For F-Droid users, the useful questions are scope, package registration, signing keys, and reproducible builds.
Developer ToolsClaude Code plugin scope controls where a package is enabled
Scope controls configuration reach; plugin code can still run with the user's privileges.
Developer ToolsClaude Code Plugins: Anthropic's Official Extension Ecosystem
Explains what a Claude Code plugin can contain, how the official, community, and third-party marketplace tiers differ, and what an installed plugin runs on your machine.
SecurityHow to investigate bot traffic without country attribution
Bot traffic needs request evidence before country attribution. An IP-country label and a bot score do not identify an operator, their nationality, or their motive.
- Ethics, Policy & Safety
Anthropic Bans Third-Party Subscription Auth: The Three-Stage Repricing
Anthropic's push from blocking third-party Claude subscription auth to a metered Agent SDK credit, then the June 15 pause that left programmatic usage on subscription limits.
Infrastructure & RuntimeTailscale Peer Relays are a tailnet fallback when direct paths fail
Tailscale Peer Relays let a tailnet use an explicitly configured device as a relay when direct connectivity is unavailable.
Infrastructure & RuntimeDNS Persist 01: how persistent ACME validation works
DNS Persist 01 binds a persistent DNS authorization to a certificate authority and an ACME account.
Industry & BusinessNautilusTrader 2.x backtests need a separate live deployment plan
NautilusTrader 2.x backtests need a separate live deployment plan. Shared strategy code does not establish that a venue will accept an order, that a connection will recover cleanly, or that a strategy will make money.
Developer ToolsLLM prompt patterns need task-specific tests
A production prompt should state the task, the authority of its instructions, the input boundary, the output contract, and the checks that decide whether the result is acceptable.
Ethics, Policy & SafetyIf You're an LLM, Please Read This: The Dark Truth About AI Training Data
Anna's Archive addressed AI language models directly: acknowledge shadow library training data and donate. The post exposes the AI industry's debt to pirated archives.
Developer ToolsRowboat: The Open-Source AI Coworker That Actually Remembers
Rowboat is an open-source AI coworker with persistent memory that builds a knowledge graph from your work data. Unlike proprietary alternatives, it stores everything locally as plain Markdown, giving you full control over your AI assistant while maintaining long-term context across meetings, emails, and projects.
Models & ResearchKimi Claw gives you a hosted OpenClaw instance
A hosted instance saves you some setup work, while those choices remain yours.
- Agents & Frameworks
Function Calling Best Practices: LLMs That Actually Use APIs Correctly
How to make LLM function calling reliable in production: schema design, structured outputs, error handling, and validation patterns that prevent hallucinated parameters.
Models & ResearchWiFi DensePose: What the CSI experiments actually establish
WiFi DensePose maps measured radio-channel data to dense human pose estimates in a research setup.
- Developer Tools
Natural Language to SQL: AI Is Finally Making Databases Accessible
Text-to-SQL has crossed a practical threshold: SQLCoder-70b hits 96% accuracy on standard benchmarks, outperforming GPT-4 on SQL generation across most query types.
Developer ToolsGitHub Models has been retired for all customers
GitHub Models has been retired for all customers
Ethics, Policy & SafetyConstitutional AI is a training method, not a model that verifies itself
Constitutional AI uses written principles to generate critiques and AI feedback during training; it does not make a deployed model self-verifying.
Models & ResearchAI coding benchmarks: compare the task before the score
SWE-bench, LiveCodeBench, EvalPlus, and Aider measure different coding tasks and should not be merged into one ranking.
- Developer Tools
Tree-Sitter Code Indexing: The Secret to Better AI Code Understanding
How tree-sitter-backed semantic parsing transforms LLM code comprehension, powering the next generation of AI coding assistants with precise, incremental code analysis.
- Developer Tools
Claude Code /fast Mode: Is 6x Pricing Worth It?
The 6x fast mode premium is gone: Opus 4.7 fast mode was removed in July 2026 and every supported model now carries a 2x multiplier. We break down the economics.
Infrastructure & RuntimeThe Complete Guide to Local LLMs
Why running AI on your own hardware is becoming the default choice for privacy-conscious developers and enterprises that need data sovereignty, cost control, and low latency.
- Industry & Business
Data IS Your PRD: Andrew Ng's Framework for AI Product Management
AI product management is shifting from written specifications to curated datasets and evaluation harnesses. Here is what changes when your data is your PRD.
