Topic hub
AI Development Tools
Tools, frameworks, and best practices for building production AI applications.
All posts · 43
AI DEVELOPMENT TOOLS01
Full Prompt Re-Processing: llama.cpp Cache Fixes That WorkOne rejected request wiped 15 context checkpoints and 2.19 GiB of KV state in a single step. The llama.cpp cache flags that actually exist today.
AI DEVELOPMENT TOOLS02
MCP Stateless Migration 2026: What Breaks in a Live ServerEleven breaking changes landed in the 2026-07-28 MCP revision. Sessions, initialize, and Last-Event-ID are gone. The migration order that keeps clients live.
AI DEVELOPMENT TOOLS03
Grok Build Deep-Dive: xAI's Open-Source Rust Coding AgentxAI open-sourced Grok Build, the ~99.6% Rust harness behind the grok CLI, days after a repo-exfiltration disclosure. A teardown plus the self-host audit.
AI DEVELOPMENT TOOLS04
Semantic Code Search vs Grep for Coding Agents 2026Semantic code search indexes your repo as hybrid BM25 plus dense vectors so agents retrieve, not grep, cutting tokens ~40% at equal retrieval quality.
AI DEVELOPMENT TOOLS05
METR Didn't Reverse Its 19% AI Slowdown (2026 Update)METR's 2026 update did not flip its 19% AI slowdown to a speedup: the point estimate held at -18%, and METR now calls it an unreliable lower bound.
AI DEVELOPMENT TOOLS06
Greptile vs CodeRabbit vs Qodo: AI Code Review 2026Greptile caught 82% of bugs in its 2025 test, CodeRabbit topped Martian's 2026 benchmark, Qodo hit 60.1% F1. Which AI code review tool actually fits.
AI DEVELOPMENT TOOLS07
DSPy GEPA vs MIPROv2: Auto Prompt Optimization 2026GEPA beats MIPROv2 by over 10% and outperforms GRPO with up to 35x fewer rollouts. Here's how DSPy's reflective prompt optimizer actually works in 2026.
AI DEVELOPMENT TOOLS08
Spec-Driven Dev Tools: Spec Kit vs Kiro vs TesslSpec Kit hit 111k stars and 30+ agents, but Kiro ships 3 files and Tessl makes the spec the source. A 2026 framework for picking the right SDD tool.
AI DEVELOPMENT TOOLS09
Code Execution With MCP: Cut Tool Tokens up to 98%Code execution with MCP cut Anthropic's tool tokens from 150K to 2K (98.7%); third-party tests show 78-99% input cuts. How it works and when to use it.Explore other topics