New technology shows up every week, and some of it becomes an edge for those who move first. Dexpert Radar follows what changes in AI, software, apps, data and cloud, and shows how to use each novelty to sell more, cut costs and innovate.
A GitHub cofounder wants to redesign Git for the agent era, five Git commands assess a new codebase and Anthropic expands capacity with Google and Broadcom.
Google releases the open Gemma 4 models, Mintlify replaces RAG with a virtual filesystem and Android developer verification starts with Brazil.
Google presents TurboQuant to compress vectors and model caches, Reco saves $500K a year rewriting JSONata in Go with AI, and Swift 6.3 is out.
The makers of uv and Ruff join the Codex team, Java 26 brings HTTP/3 and performance gains and Mistral launches Forge for custom models.
JavaScript's new date API reaches its final stage, DuckDB runs big data benchmarks on an entry-level MacBook and Cloudflare crawls whole sites with one call.
GPT-5.4 operates computers with a 1 million token context, a CLI brings Google Workspace APIs together and a comparison of 44 VM types shows the best value per dollar.
vinext, a Next.js rebuild made with AI for $1,100 in tokens, Google's new image model and PgDog to scale PostgreSQL without changing the application.
Sonnet 4.6 improves AI computer use, the llama.cpp team joins Hugging Face and Go 1.26 ships a rewritten go fix to modernize code.
Google updates Deep Think for science and engineering, OpenAI releases a coding model with over 1,000 tokens per second and a former GitHub CEO founds Entire.
Claude Opus 4.6 with a 1 million token context, Xcode with Claude Agent and Codex built in, and why keeping data in Postgres makes agents' work easier.