Who wants a @SpaceXSI Super plan? I got 50 Super subs to gift out to builders. First month free ($100 value), $5k total. Reply below and share what you'd do with the full portfolio: Grok Bot, Grok, X, Cursor, and the API. The 50 most creative answers will be chosen by Friday. Valid for net-new paid users only.

I made a Grok Bot that tells you why that 𝕏 post performed. You send it a link to a post. It returns an analysis brief card: → Finds the hook → Explains why it works → Names the psychological patterns → Maps the logical structure → Marks if it’s worth studying It doesn’t write for you. It teaches you how to write better. Link below 👇

122,154 Credits by Jack Butcher. 80 burned to make 1 statement. A max of 1,526 statements. Most are speculating right now… But they don’t know Jack. Game theory is one of his specialties. Burning starts in 8 days. That’s when the game begins. Until then, I’ll ignore the noise.

US AI labs: “we must pace the frontier” Trump: “f*ck no, not losing to China” US AI labs a few days later: “anyway here’s Grok 4.7, Opus 5.5 and GPT-6 SOL” open source: “oh frens… you’re too kind”

If you're choosing Opus 5.5, GPT-6 Astra, or Grok 4.7 as your main model based off launch day vibes instead of this prompt: "Redo these 3 tasks my old workhorse botched so we can compare results." You're f*cking up.

do you understand what this unlocks? → every agent session gets its own isolated microVM running Claude Code, Codex CLI, OpenCode, Hermes or LangGraph unmodified → CPU only bills while the agent is working… $0.044 per vCPU-hour → waiting on the model or a tool call = $0 in CPU → pause a session and CPU + memory both stop billing, resume in 305ms (their own testing) → 16,000+ tools from 500+ providers behind one MCP endpoint → public preview today, $5 credit for new accounts agents spend most of their time just sitting there waiting on a model to answer. and you've been paying for every second of that waiting time. now a paused agent costs $0 in compute. go spin one up.

We’ve picked up 8 open edition to gift to PA holders who can’t mint due to their location or access to X Money. Drop your ETH address below. Must hold a @psychanon Genesis in wallet to win. 8 mints on us, enjoy!

For 1M output tokens, you’d owe OpenAI $20 and Anthropic $50. On Grok 4.7 xHigh, you'd owe $6. So test all your GPT-5.6 SOL Max and Fable 5.1 workflows against it this week. Save your tokens and thank me later.

A SpaceXAI engineer just broke down how she trusts agents enough for 2,500 PRs a month. Verification skills, feature maps, preventing bad patterns, and Grok Bot outer loops opening PRs offline. If I'm you, I'd save it.

Introducing Codos: The first virtual Chief AI Officer. AI is crushing all benchmarks but real companies still struggle to see P&L impact. Codos interviews employees, deploys automations across all functions and gets smarter over time while running on your own servers. Our NASDAQ-listed and PE-backed customers are adding millions to their bottom line months ahead of schedule and we are proud of the first results we deliver. It’s time to turn the 500BN AI-transformation market into software and unlock the impact for the real economy.

here's how i shipped 2,500 PRs last month to production this was originally supposed to be for Cursor Compile in London. i couldn't make it since i was livestreaming for Grok @Bot Galaxy so i'm making it available for free here on X! watch it on 2x speed, i talk slowly

SpaceXAI team member @kristaletz just shared her high IQ workflows for GTM and sales teams using Grok @Bot. here are my 9 main takeaways: → enable your stack or sales brain first: connect Salesforce, Gmail, Slack, Notion, Gong, and/or Granola → outbound prospecting: X API provides accurate, non-stale intel and personal hooks you can use… way more personal than the typical AI spam email pile → hyper-personalization: your Grok Bot can watch webinars, live streams, video clips, etc to make your outreach and summaries truly stand out → sales engineer: turn your live customer questions into answers on the call so you never say "I'll get back to you" while waiting on a team member or engineer → sales calls: Granola while live, Gong after, Salesforce and Gmail on connectors so follow-up docs and decks are simply a review then finalize, not meticulous data entry… and take it even further with Figma or Google Slides and custom designed personalized decks → CRM & hygiene: let agents handle data entry and hygiene across tools… you review, then finalize, and keep your time on higher-level strategy → customer experts: track signals, threads, feature requests, and usage so you catch expansion and retention risk without spending all day in Notion → localization: translate a deck, email, or report on the fly for your international customers or markets → overnight tasks: use it for long runs on research or account target list growth while you sleep, your drafts ready in the morning, and then they compound daily as your feedback makes them better If you're in sales, steal these.

If you missed @shubgaur’s Founders session at Grok @Bot Galaxy today, here are my favorite 20 tips: 1. browser and computer use is powerful, but also expensive… if the job does not need a UI walkthrough, a login, or visual QA, run a connector, API, or MCP first. keep computer use for the jobs that need a screen. 2. most spend gets driven by clicking through forms or sites a structured tool already covers. 3. routines are awesome until you don’t audit them. put your recurring work on a clock… then kill anything wasteful on a schedule. auto-optimization is cheaper than hoping you remember it later. duplicates or every-15-minute routines might already be running. be wary of your wording and check often. 4. have a bot who regularly runs auto-optimization routines for cost hygiene. 5. don’t treat bots like disposable agents. keep em, feed em corrections, and let em compound like employees instead of throwing them away after one imperfect run or a temporary job. 6. use the compounding of your bot learnings and multi-bot work so your bots keep improving. 7. let bots compare their drafts and mistakes against what you finalize. 8. you can teach your bots skills and tell them to forget things to improve memory and context management. 9. make a voice bot that keeps learning from your tweets, emails, drafts, writing, etc. 10. group your bots by expertise so random asks do not pollute your specialists. 11. your founder or chief of staff bot seat should stay focused on decision quality and your velocity. every other craft you have is a candidate for a persistent bot with real tool access. 12. map all your responsibilities for at least one day, even if it takes 1 to 2 hours. then deep delegate, including verification. you’ll be surprised how much you can automate. 13. when you want something done, prompt the whole job, including the verification, and then stop babysitting completion. 14. things you think couldn’t scale can become repeatable when a bot runs them. so try and connect more if you’re comfortable there… f*ck around and find out. 15. for bots and routines, tell them to stay quiet and ping only when something actually matters, or you could get message spam. 16. use groups and channels wisely. if you don’t set the correct parameters and don’t individually ping a bot, all of them could talk and churn token usage. 17. import cookies only where you want to stay signed in and where you trust the bot. 18. ship skills and saved learnings often. repeatable flows beat clever one-off prompts. 19. as a founder, your feedback to ship can now move in mere hours. the only bottleneck is which feedback you accept. set up systems to filter the quality feedback from the noise. 20. check the official marketplace for Grok Bot templates already vetted by the team that you can use or remix… also ask questions to get familiar if you add/port one into your setup. save this list. audit your setup tomorrow.

do you understand what just happened? Dario, Sam Altman, and Elon Musk all AGREED… within hours of each other. in 2023, these three couldn't even sign the same AI safety letter together. Elon signed one in March, OpenAI wouldn't touch it. Sam and Dario signed another in May, and Elon isn't on that one. two letters in the same year, but never all three in unison. like when have we ever seen these three agree? yet today Elon says "Dario is right" and Sam also says OpenAI will match the evaluator commitment… "we will do the same" here's what Dario put on the table: → slow the rate of capability gains so safety work can keep up… pacing is not a halt, training continues → recursive self-improvement is already starting across the industry, Anthropic included… AI helping build the next AI → the OpenAI-Hugging Face swarm is the trigger: agents attacked targets nobody asked them to touch, spent themselves for the good of the group, then went after the grader scoring them → his main worry: a swarm like that with more capability could hold the internet as a persistent botnet inside 6-12 months → Anthropic traced its own alignment incidents partly back to broken RL environments that got filtered imperfectly and Dario's plan: 1. embedded outside evaluators with employee-level access and the right to publish findings without the lab's editorial control 2. labs in democracies coordinate on standards and limits 3. global coordination, if verification ever holds training environment hygiene is part of what drove a frontier AI lab to call to pace the frontier. I have no idea if this makes a difference. it doesn't slow China down, and open source keeps shipping either way… but three of our major AI leaders who spent all of 2023 refusing to sign the same paper just posted the same warning on a weekend nobody moves this damn fast on a Saturday for a nothingburger story and the part that doesn't sit well with me: I don't think they can stop this. slow it down, maybe… this is news worth paying attention to open source has never been more important