$ man subscribe
The weekly Build With newsletter
One email a week from Build With. The week's best AI builder tool drops, ranking changes, and new tools worth trying.
Free. Unsubscribe anytime.
$ subscribe --weekly
One email a week: new tools worth your time and ranking changes. No spam.
$ subscribe --why
Why subscribe
- ›Tool drops. The week's best new AI builder tools, each one tied to a source.
- ›Ranking changes. When a list on the site moves, the letter says what changed.
- ›Worth trying. New tools that earned a look, not every launch in the feed.
It uses the same lists as the site: 63 ranked tools across 11 categories. Rankings are editorial picks, not paid placements.
$ tail ./new --since 7d
Last week's drops
40 verified releases in the feed from 2026-11-15 to 2026-11-21. The letter picks from drops like these.
2026-11-21 #
Cline 4.1.27 ↗With no saved model, the Cline provider now starts on the first free model instead of Claude Sonnet 5.5. read_files results are no longer cut to 8k characters (now up to 2,000 lines or about 48k characters), LM Studio gets an API key field, MCP tools declared read-only get a badge, the model catalog adds 94 providers from models.dev, and reasoning-effort fixes stop GPT-6 on Azure and custom gpt-5.6/gpt-6 endpoints from rejecting tool calls.
2026-11-21 #
OpenVorld ↗Free, local, MIT-licensed harness for AI agents shown as a pixel-art village: each project is a "vorld" with named agents, roles and houses, and each agent can run on Claude Code, Codex or any OpenAI-compatible API (OpenRouter, Ollama). Agents act only on goals you give them and ask before touching a folder, command or website, with per-agent permissions ("writs") you can grant or revoke. No account needed. Launched on Product Hunt on Nov 21.
2026-11-20 #
Currents for Claude and Codex ↗Currents, which records every Playwright test run in CI, now has a plugin in the Claude and ChatGPT/Codex directories. Sign in and ask your agent why CI failed: it gets the failure, the test's history across runs and branches, and the trace, plus skills that tell it to fix the cause rather than add retries or timeouts. Works on every Claude plan; Currents has a 14-day free trial.
2026-11-20 #
tensorViz ↗Free, open-source (MIT) tool that runs one real forward pass of your local PyTorch model and turns it into an interactive graph of every module and tensor shape. Claude Code, Codex and Cursor can search the graph and read measured shapes, so they fix model code from evidence instead of guessing. A public Model Explorer has graphs of 560+ models from recent papers. Preview is CPU-only.
2026-11-20 #
Kiro for VS Code ↗Kiro's agent now ships as a native VS Code extension, generally available on macOS, Windows and Linux via the VS Code Marketplace or Open VSX. It brings vibe and spec sessions, steering, MCP, workflow agents, memory and "approve for me" into a VS Code side-bar hub, reads the same .kiro folder as Kiro IDE (specs move both ways), and shows per-turn diffs you can accept or reject file by file. Needs VS Code 1.147+ and is included in every Kiro plan with no billing change.
$ ls ./issues
Past issues
Every sent issue, as it went out, is on the archive.
- #006 Replit's agent swarm, plus Vercel's eve hits 1.0 Nov 19, 2026
- #005 Fable got cheaper, Gemini 4 Carbon lands Nov 12, 2026
- #004 Claude Ultra drops its 5-hour cap, Copilot fixes your PRs Nov 5, 2026
- #003 Cursor drives your desktop, plus a Git forge for agents Oct 29, 2026
- #002 A 1,000 tok/s diffusion LLM and crash-proof agent runs Oct 22, 2026