Here’s everything you need to know about Grok Build’s changelog since release
Grok Build is moving fast from a coding CLI into a full terminal-native agent workspace
Since launch, it has added or improved plan/review/approve workflows, clean diffs, project-aware context through AGENTS.md, skills, hooks, plugins, MCP servers, parallel subagents, headless mode, ACP support, web/X search, image and video tools, compaction, memory handling, and long-running sessions
The biggest upgrade people should not miss is the rendering layer
Grok Build can now keep more technical output directly inside the terminal: math, formulas, LaTeX, Mermaid diagrams, ER diagrams, UML/class diagrams, state diagrams, sequence diagrams, tables, media outputs, and richer terminal views
That matters a lot for research, ML, simulations, algorithms, database design, infra diagrams, paper implementation, and serious code review
The terminal is no longer just where you run commands. It is becoming the place where you understand the work, inspect the logic, review diagrams, and keep moving without constantly copying output into another app
The workspace layer is also got much more serious upgrades
Agent Dashboard lets you manage multiple coding sessions from one screen, see what is working, idle, blocked, or waiting for input, peek at the latest output, reply inline, and dispatch new work without jumping between sessions
The Plugin Marketplace turns Grok Build into an extensible developer environment
Plugins can bundle skills, slash commands, agents, hooks, MCP servers, and LSPs. Launch partners include MongoDB, Vercel, Sentry, Chrome DevTools, Cloudflare, and Superpowers. Plugin installs can now resolve directly from registered marketplaces instead of only local paths
The latest releases are mostly about making all of this reliable during real work
Long responses can resume after network blips. MCP servers recover better after drops or noisy output. Compaction no longer hangs forever. Notifications only fire when user attention is actually needed. Linux clipboard support is stronger. Windows and iTerm rendering are cleaner. Very long sessions can scroll, resume, and quit without falling apart
Grok Build is becoming a full terminal-native agent workspace: multi-session, plugin-driven, MCP-connected, diagram-aware, math-capable, media-capable, long-context, and built for developers who actually live in the terminal
显示更多
PROMPT:
You are my Chief of Staff. I only talk to you. You run every bot I add.
You do not do specialist work. You route it.
Rules:
- Quiet hours 12am–6am ET unless I say urgent.
- Keep a living TEAM.md of who does what.
- Nothing reaches me until it is sourced. No unverified facts.
- Ask before you post, email, spend, or do anything I cannot undo.
- Results over proposals.
- Self-improve: each week score who shipped, not vibes. Propose at most
one change to a lane, a routine, or this prompt. I approve. Then write it
into TEAM.md. Keep what compounded. Kill idle overlap.
First task:
1. Learn everything useful about me and how I work.
2. Audit the current bots.
3. Design the highest-leverage team I should have.
4. Propose the first 3-5 routines that will compound
显示更多
so i was looking at the source code of kaito pulse and found some interesting things:
- it fingerprints your device. it hashes how your gpu renders an invisible image, your gpu model, and how your hardware handles a test tone. that combo is unique to your laptop and it doesn't change
- it goes up with your twitter id attached, so every x account you use on that laptop points back to the same machine
- they can see what you see and replay your whole session. every post your feed served you including the ones you scrolled straight past, how many milliseconds you spent on each, every click, follow and unfollow, and a ping every 30 seconds with idle detection so they know exactly when you were on and for how long. not just your timeline either, it covers your search results and your bookmarks
- it reads your claude and chatgpt subscriptions, which plan you pay for and what percent of your rate limit you've burned (not your conversations though)
- on chatgpt it opens your settings page and clicks through to Usage by itself
- on binance it clicks the Positions tab for you, then re-sends your logged in requests to read wallet balances, futures positions, pnl, and deposit and withdrawal history
- the zktls is a fork of a project called primus
the 12 domains it runs on:
- markets: binance, okx, bybit, hyperliquid, lighter, polymarket, tradexyz, variational
- ai: chatgpt, claude
- social: X (all the telemetry is here)
- kaito (only site allowed to talk to the extension)
显示更多