skcd

Understanding the universe @xai
ex hacking @aide_dev
ex fb engineer
ICPC WF

its just code 👨🏼‍💻

[skcd @skcd42]

Grok-code-fast-1 is now out and available for everyone to use 🚀🏎️💨

When I joined the coding team, the team was just 3 people and we very quickly built a model which was SOTA on SWEBench. But as things go, in the real world benchmarks matter less. Over the last few months we approached the modelling + data + infra perspective from a different lens, putting developers and users first over everything else.

This required us to tune the data recipe, get the infra in place to do a lot of rollouts and create a set of grounded evals which was powered by both human judgement and an in-house auto-evaluation framework that captured real world usability.

This is the first model of many in the grok coding family, we are going to make quick iterations and improve the model performance over time.

We thrive on your feedback, please share your unfiltered and honest thoughts so we can keep pushing new boundaries for agentic coding

[xAI @xai]

Introducing Grok Code Fast 1, a speedy and economical reasoning model that excels at agentic coding.

Now available for free on GitHub Copilot, Cursor, Cline, Kilo Code, Roo Code, opencode, and Windsurf.

https://t.co/3tMbmLbxOP

[skcd @skcd42]

@_imdawon would love to hear/learn more. do you think grok build is burning too much usage for the same task?

[skcd @skcd42]

@bercankilic @keertimanney

[skcd @skcd42]

Bug fixes shipping to Grok Build 0.2.16 (release notes will be available in the TUI and on change-log website)

• fix stuck "Starting session..." / "Connecting MCPs (0/0)..."
• show where external skills/mcps/hooks/plugins originate in grok inspect
• fix Terminal app unsafe shortcuts
• render streaming bash output
• use grok to derive /loop interval from the request instead of hardcoded parsing
• fix parent session state (background tasks, monitor) leaking into subagent conversations
• make Claude/Cursor skills, AGENTS, mcps, and plugins configurable
• wire --permission-mode acceptEdits to auto-approve edits
• share FS watchers by working directory
• inject UTF-8 env defaults for Windows child processes
• fix blank completed bash / code-execution cards
• fix scroll wrap on scrolling vs. shortcuts on all menus

[skcd @skcd42]

@GrumpyTechBro which terminal and OS is this on?

[skcd @skcd42]

Bug fixes shipping to Grok Build 0.2.13 (release notes will be available in the TUI and on change-log website)

We are leveraging the alt-screen to better handle your background tasks, subagents, monitors with smart grouping allowing you to navigate between them quickly

• Group Subagents → Tasks → Watchers, within subagents, order by agent type (Explore, General, Plan)
• Show timestamps on pinned user messages at top of scrollback
• Fix command highlighting when prompt has paste chips
• Update the context-usage indicator to show tokens by default
• ANSI16 fallback for themes
• Better context usage breakdown/rendering
• Update highlighting for /loop, monitor, tag colors
• Tab now cycles Prompt → Scrollback → Tasks → Prompt
• Order gateway turn-completion after streamed content
• Formatting: fix language-tagged fenced code blocks and fix code under a list item

[skcd @skcd42]

We are excited for all of you to try out Composer 2.5 in Grok Build starting today!

To use composer-2-5 do `/model` in Grok Build and type in Composer to switch

Composer 2.5 comes with 200k context window and supports: subagents, MCPs, skills and additionally also works with your .cursor settings

[skcd @skcd42]

@chribjel on nice, you discovered it. let me know how you find the experience, we are still iterating on the details of the implemenation

[skcd @skcd42]

@badmorgan420 fix will be out soon

[skcd @skcd42]

@thdxr 😅

[skcd @skcd42]

@meglio good point, is the difference between edits and bash commands which is confusing?

[skcd @skcd42]

@G1ZM084 could you try again? I can see its rolled out competely

[skcd @skcd42]

Grok Build: X search and backend web search are rolling out again, we will complete it over the next few hours; currently at 25%

cache hit rates and ttft are healthy again

[skcd @skcd42]

at 50% now

[skcd @skcd42]

@AdsQnn took a look, model seems to be going into a doom loop like pattern, we will aim to fix this in the next model iteration (due in 2 days)

[skcd @skcd42]

@AdsQnn this is very bad, can you please DM me your session-id or use /feedback so we can take a look

[skcd @skcd42]

@olsenbdnr rise of the monolith

[skcd @skcd42]

@FossPrime which terminal and OS does it not work in? we also have /multiline which should enable support for it

[skcd @skcd42]

Bug fixes shipping to Grok Build 0.2.11 (release notes will be available in the TUI and on change-log website)

- Make tab bar arrow-focusable in picker modals (MCPs, plugins, hooks)
- Show "Switched to mode" banner above prompt for `Shift+Tab` cycles
- Instant loading indicator on model switch
- Fix broken branch glyph on Windows
- Fix WSL `Ctrl+V` image paste
- Fix UX bug where `/context` always showed "Auto-compact at 85%"
- Fix slash command autocomplete and rendering
- Boost terminal video playback to 30fps
- `Left`/`Right` arrows switch tabs in extensions modal
- `Up`/`Down` arrows at list edges focus search input
- Fix terminal resize on multiplexer (zellij/tmux)
- Open buttons for imagine media output
- Render streaming bash tool output
- Increase default retry budget to ~5 min and simplify retry UI

[skcd @skcd42]

@abdurrahmanregi @Jaaneek this will work if the model is trained and can handle OOD traces, without doing evals supporting a mode like this can lead to not so optimal traces