Coding

DuckDB crunches gigabytes of CSV and Parquet with no database server, no import step, and no waiting around. You aim a SELECT straight at a file on disk and it answers. On 41 million rows of raw NYC taxi data, I clocked a full group-by aggregation in 20ms and a two-table join in another 20ms, read straight off Parquet with nothing loaded, copied, or indexed first. That is a multi-gigabyte analytical query returning before you lift your finger off Enter key! Every number in this post comes from a benchmark you can run yourself; the scripts and raw results live in a GitHub repo .

uv workspaces give Python a Cargo-style monorepo setup. You get one lockfile, one virtual environment, and auto-resolved inter-package dependencies. Cold installs finish in seconds, not minutes. Pair uv with Ruff for linting and formatting, and the pair replaces Poetry, Black, isort, flake8, and pip-tools in one shot. The rest of this post covers workspace setup, inter-package deps, Ruff config, CI, publishing, and the traps that snag teams moving off older tools.

Google shut down Gemini CLI on June 18, 2026, pushing free, Pro, and Ultra users onto the closed-source, Go-based Antigravity CLI . You can keep the open-source Gemini CLI running with a paid API key, but its separate quota pool is gone. Enterprise and Code Assist Standard licenses still work unchanged.

What happened to Gemini CLI?

On June 18, 2026, Gemini CLI and the Gemini Code Assist IDE extensions stopped serving requests for free, Google AI Pro, and Ultra personal accounts. The Google Developers Blog announcement confirmed the cutoff and named the replacement: Antigravity CLI, run with the command agy.

The Charmbracelet Bubble Tea framework lets you build live terminal dashboards in Go using the Model-Update-View pattern from Elm. Pair it with Lip Gloss for styling and Bubbles for ready-made widgets. You get live panels, key navigation, and flex layouts. It all ships as one binary with zero runtime dependencies.

Terminal dashboards fill a niche that classic CLIs and web apps both miss. Think of a monitor that runs over SSH on a headless box. Think of a database explorer that starts in milliseconds with no browser. Think of a log viewer your ops team can reach with no auth layer to set up. These are the use cases where TUI dashboards shine. Bubble Tea now sits at v2 with over 41,000 GitHub stars and more than 18,000 apps built on it. It has become the go-to framework for this kind of work in Go.

OpenCode, Claude Code, and Cursor solve the same job three different ways. On one production-codebase test, Claude Code finished 45% faster while OpenCode wrote 29% more tests, and Cursor is the IDE-native option neither benchmark page even mentions. The real winner depends on the model you run and the budget you keep.

What is the difference between OpenCode, Claude Code, and Cursor?

These three tools split along two lines: who picks your model, and where the agent lives. Claude Code is the managed option. It works out of the box. The catch is that it ties you to Anthropic models like Sonnet, Haiku, and Opus. It runs in your terminal and mostly “just works” with no setup.

The best AI coding agent in 2026 comes down to two numbers most reviews skip. The first is real cost per completed task. The second is how locked in you are to one vendor’s models. Get those two right and the rest is preference. Get them wrong and you either overpay every month or hand a single vendor control of your roadmap. This compares seven agents on exactly those axes: Claude Code, Codex CLI, Gemini CLI, Cursor, OpenCode, Pi, and GitHub Copilot.