Jason Ginsberg

I’m proud to be joining SpaceX and xAI with @milichab It has become clear that software is changing fundamentally. More and more, people can shape the tools they use directly, and the ceiling of what can be built keeps rising. What makes xAI special is the scale of its ambition: to build from first principles all the way out to the stars. I’m especially grateful to work on products that expand human agency and freedom. That mission is deeply personal to me. My family came to the United States fleeing communism, and the belief that freedom should be part of the next generation of the internet has driven me every day since Andrew and I started Skiff. Now, we get to work on intelligence, understanding, and freedom on a universal scale.

@itsclarkholden @elonmusk @xai Looking into this. Can you DM the account/email?

I literally can't login to Grok or Grok Build. I just keep getting put here after logging in. Someone help? @elonmusk @xai

@mehulmpt extremely helpful! we'll work on these today

I've used Grok 4.5 inside the grok CLI for a few days. Here's some feedback from my real work usage to improve the CLI experience. 🧵

Grok 4.5 is now available as an orchestrator model in Computer for Consumer Pro and Max subscribers. We evaluated it against five other orchestrator configurations on WANDR. It scored higher than every other configuration at roughly half the cost of Opus 4.8.

@yuki_eliot thank you! followed

Grok 4.5 is INSANE for web dev. I built this scrolling space explorer in just a couple hours with Grok Build. Link below, try it out!

@LuxVeritasAeter thank you! extremely helpful

Sent via X Chat. Well, sent you several session IDs. I've done a mix of Ralph Loops (which did not give the uplift I wanted) and walking it from session to session with memory docs and session hand off files to avoid Context Rot at the >40% point. Hand holding it through iterations and being hyper granular in critiques, while also leveraging 'use super powers' has seemed to get me much further than anything else I've tried. Still is a far bit away from what I was shooting for. I have a parallel project that I am doing using Fable 5, and it got a lot of the game art and feel right in first pass. Not perfect by any means, I am still tweaking to taste, but it was fully navigable and playable environment.

some special features in Grok Build if you're new /dashboard: shows you every agent running in your TUI. no need to tab jump. can click and respond. /imagine: create videos + images w/ an agent. useful for three.js textures or complex video edits. X search tool: great for research. try asking for the best recent AI papers. Grok 4.5 is strongest on tough coding problems in real, messy codebases. push it hard and send us feedback.

@LuxVeritasAeter can you send a session ID? we're actively working on improving this

@JasonBud I find it struggling on game design, map layouts, procedural art etc. I may be missing something. Any particular skill or flow that helps feed 4.5 to handle those better?

@xl0garithm yeah its available to the agent as a tool, so Grok 4.5 will use imagine automatically for you when useful for a given task

@JasonBud Can the imagine commands be incorporated into the coding workflow directly? Like prompting it to iterate over something like threejs image assets and embed them directly into the dev flow?

@jpschroeder @Tips_Singer any features you're missing in Grok Build that would make it work better for you?

@Tips_Singer same. codex (cli) is my favorite so far, but i do prefer grok build's aesthetics. Super curious what your opinion is after using it a bit

Grok 4.5 is really good. I already prefer it to Opus (<Fable). I also really like Grok Build, delightful take on a CLI agent (please add `esc` = steer). Only downside so far is it burns sub usage pretty fast.

you can also keep up with the latest at x.ai/build/changelog, /changelog in Grok Build, or by tagging me, @milichab, @skcd42, @aksheyd, @theskory, @AmanGotchu, @krgeorge, @wwzeng1 and anyone else on the team

@nanliuuu appreciate the feedback. if you have specific session ID you can share, we'll take a look.

Currently trying Grok 4.5, but it doesn’t seem to revert previous changes sometimes after I told it to do differently. Codes do get a bit messy if you don’t look at your code. Prob not ideal if token efficiency partly comes from this as this leads to a bit of unpleasant user experience.

@intheworldofai this looks great. what were the corrections you needed to get it perfect? single HTML file intentional?

Grok 4.5 just built a full macOS-style web OS in one shot and actually crushed it. Single HTML file with embedded CSS + JS. From one prompt it delivered: > macOS menu bar + dock > Draggable and resizable app windows > Functional file explorer > Working terminal + browser mockup > SVG desktop icons + notifications > Settings app with theme switching > Fully functional sound player (real Web Audio synthesis) > Playable games inside the desktop Almost no corrections needed. The working sound system and interactive elements inside the windows is where most models usually fall apart, but this one handled it cleanly. Grok 4.5’s frontend and UI generation is actually on another level right now. @SpaceXAI

Game on!

Congrats to @SpaceXAI on Grok 4.5 — trained on NVIDIA GB300 NVL72 systems and purpose-built for coding, agentic tasks, and knowledge work. This is what happens when world-class AI infrastructure meets world-class AI research.

@aipulseda1ly can you share the session (or request) ID? will take a look at why it froze

Grok 4.5 built this NYC scene in under 3 minutes. Thats the headline. Its a clear step up from the old Grok, improvements across the board. For three.js its not the best, Fable and Opus still render noticeably better and it settled on a frozen frame. But three.js is a visual test. For backend coding and fast iteration im sure its strong, that just still needs to be tested. I even ran goal mode for 30 minutes to push the scene further and got basically the same result. The speed is real. Where it actually shines is the next thing to find out.

try it out for free on your hardest coding problems. we’ll incorporate your feedback into the model and future iterations.

Grok 4.5 in Grok Build also stands out for its efficiency. Grok 4.5 in Grok Build cost $2.49 per task while Fable 5 in Claude Code cost $11.80 and GPT-5.5 in Codex $5.07. This is driven by relatively low token pricing and the model using far fewer tokens than comparable models (1.9M average tokens used per task), significantly less than Fable 5 in Claude Code (7.2M) and GPT-5.5 in Codex (6.2M)

please give Grok 4.5 a try. it's an incredibly useful daily driver.