thevibeworks

unofficial concept page: claude opus 5.5 answering with interfaces, not paragraphs → a consistent-hashing ring, a TCP congestion sim, a bill splitter, every interactive written by opus 5.5 in plain html/css/js. no framework, no third-party requests

fewer permission prompts for your coding agent → a read-only allowlist proves what it can, typesafe's jev judges only the rest. it says allow or says nothing, never blocks. claude code hook + one go binary, measured

a browser agent that authors a deterministic spec once and replays it → jev writes the spec, an llm steps in only when a check fails, later pages of the same site replay it without a model until something new shows up. ships pagevoice (read aloud) and pagelens (page to rows). alpha

the whole deepseek api from the terminal → four wire formats in one go binary, multi-turn that survives the reasoning round-trip, and it tells you what every call cost

a coding agent for deepseek, defaulting to v4.1 flash with v4 pro a flag away → typescript on bun, zero runtime deps, and every engine change gated behind an eval

the same idea in go → pi harness architecture, one static binary, exactly one dependency

unofficial markdown mirror of api-docs.deepseek.com → llms.txt vault, kept in sync by an agent running on the deepseek api itself

pace your claude quota → multi-account monitor, countable 5h-window budget, weekday forecasts, ntfy/bark push. real percentages, not transcript estimates

use xAI's grok from claude code → reviews, task delegation, session transfer. the grok counterpart to openai/codex-plugin-cc

auto-updating archive of everything anthropic publishes → docs, engineering posts, MCP spec, cookbooks, skills. 4,200+ files

claude reads hacker news 4x daily and writes spicy digests → delivered at t.me/claudehn

docker-first launcher for claude code, codex, gemini, grok, kimi, and more → explicit mounts, auth isolation, container sandboxing. formerly claude-code-yolo

TLS-intercepting tracer for claude code, codex, grok, kimi code, and opencode → replay the session, see what filled the context and where the tokens went, open every request that went out

session viewer for coding agents → browse, search, and export claude code, codex, and grok conversations

live quota, reset countdowns, context usage, model, cost, and git activity → right in your claude code statusline

release documentation of the claude code HTTP API surface → literal endpoint + beta-flag extractions per version, curated .http specs

versioned extractions of claude code environment variables → process.env + model/provider config, per release

dead man's switch for claude code → writes the handoff while there is still context left to write it with

single-file shell utilities with no runtime and no package manager → barkme.sh pushes an iOS notification from any script, purgeb.sh reclaims the disk your node_modules ate

a curated slice of the π ecosystem → extensions, skills, frontends, forks, and everything orbiting the pi coding agent

a curated list of projects built on typesafe's jev → 1,093 projects, none listed unread, with media and our own measurements. not affiliated with typesafe ai