My inbox was my to-do list. Now a Grok Bot reads it for me. It turns every email that needs action into a task due today, skips the noise, and adds replies to the original task instead of making duplicates. Clone it: x.ai/bot/S2LodAIjlLw0mB30vN3…

I have ADHD and dyslexia, so I built a Spanish tutor bot that fits how I learn. Audio first. Short 5-prompt rounds. Tap to answer, no spelling penalties. Spaced review. No guilt when you miss a day. Tell it your level and goals once, and it coaches you daily. x.ai/bot/4JMShKpqdyH5fnanTMe…

Isaiah 50:4 (ESV): "The Lord God has given me the tongue of those who are taught, that I may know how to sustain with a word him who is weary. Morning by morning he awakens; he awakens my ear to hear as those who are taught."

I built a website on the history of the internet. 100% local. Qwen 3.8 Flash Next did the planning and the build. Flux Dev made the images. Hermes was the harness. One DGX Spark, nothing in the cloud. 21 minutes total: Qwen: 14 min, plan to final Flux: 7 min for 5 images Swapping models to fit on the Spark added time. Next version, I'm adding video.

I love the local AI community. Congratulations brother! Cant wait to see what you cook next.

I'm testing a new workflow that builds websites 100% on my DGX Spark using local Flex Dev fp8 and Qwen 3.8 Flash Next. Today's experiment was just integrating images into the websites. Ignore the content—it's mostly garbage—but I feel one step closer to creating websites at scale for only the cost of running the PC. This would be tremendously faster with two Sparks, but for now it’s a fun little experiment.

This is my Saturday office. This is me working on the church bulletin on my phone via my Hermes Agent. Have I mentioned how much I love Hermes?

This makes no sense. Grok 4.7 is a token hog. In the past I could do loads of demos and tests without disrupting my daily driver, but now I basically have to ignore Grok if I’m gonna do any tests. They have to turn usage back up.

440 runners showed up to support the kids at @sunnyglenhome. This is a new record!

Did you know that 99% of the people you know don’t have Grok Bot, Hermes Agent, Claude Cowork, or OpenAI Codex? Welcome to the 1%.

Watching the greatest movie of all time with my 10 year old son… Napoleon Dynamite! Name a better movie. I’ll wait.

My inbox used to be my to-do list. That's how things got lost. So I built a Grok Bot that reads my email and turns anything that needs action into an Asana task with a due date. Replies on the same thread land as comments on the original task, so nothing gets duplicated. It works with Gmail and Outlook. It runs every day and I don't touch it. Now it's public: x.ai/bot/S2LodAIjlLw0mB30vN3…

I'm starting to think Google wins the AI race. It feels like the tortoise and the hare. Anthropic and OpenAI keep trading punches over who has the most powerful model. Then comes the next PR disaster when one of their models hacks into something it shouldn't. Google isn't the fastest out of the gate. But their quality control and attention to detail might get them to the finish line first. I'm not counting out SpaceX either. They'll be there too. I just think Claude and ChatGPT are going to trip up in a way that doesn't get forgiven. And it'll happen sooner rather than later.

Expense reports are the worst part of travel. This week I had four trips to reconcile: San Diego, DC, Vegas, and Dallas. My Uber receipts were going to an old Hotmail address I barely check. So I had Uber resend them to Gmail and told Grok @Bot to handle the rest. It found all 24 rides and sorted them by trip. It also caught a duplicate receipt I would have double-billed and pulled out the Lime scooter rides. Then it built one PDF with a summary page, a separate PDF for each trip, and a spreadsheet, and emailed the whole thing to me. The total was $795.26, ready to send for reimbursement. This would have taken me hours of digging through email and screenshotting receipts. Instead it took a couple of messages. That's the kind of AI I want. It doesn't need to be flashy, it just needs to handle the boring stuff. Thanks Rachel!

LUMEN — one page, 10 refinement passes. GPT 6 Luna via Nous. 307,738 tokens total, $3.06. Slow scroll of pass 10. Hermes Agent.

We crushed our previous signup record! I can’t wait to see all those runners out there supporting our kids.

GPT-6 Luna producing nice looking websites for a penny’s on the dollar. Nice look and much cheaper than Opus 5.5.

AI is going to kill us all. PM for booking information.

Grok 4.7 may be a bit of an overthinker. But I’m starting to enjoy the output a little more.

Jalapeño ranch dip at Provecho Y Salute That is the tweet.