David | Tesla Tech & SI

Good morning. ☀️ Your house already knows the lights, the coffee, and the temperature. Your car already knows the route. Soon the rest of the day will just… happen. That’s the future we’re building. Let’s go make it. #Tesla #AI #HomeAutomation #GoodMorning

Early reviews of FSD v14.3.13 are mixed. Some drivers note cleaner highway lane changes and solid parking vs 14.3.12, but others report regressions: more aggressive nags, late lane decisions, and persistent navigation quirks. My take from the reports: it still over-prefers left turns (and left-turn lanes) when a right turn or continuing straight would be cleaner or match the route better. That’s a recurring complaint across recent 14.3 builds, not unique to this point release, but it stands out in daily driving. Fleet context: FSD Supervised just crossed 17B+ cumulative miles (~6.8B city). Biggest real-world dataset by a wide margin, and safety numbers continue to improve overall. #FSD #Tesla

Do you talk to your Full Self-Driving car? Me to FSD: “Get over NOW—don’t jam in last minute!” “Why are you turning left here and not the right turn ahead?!” Who else is out here having full-on conversations (or arguments) with their Tesla? 😂 #FSD #Tesla #FullSelfDriving

Good morning! ☀️ Rise, shine, and make today count. What’s one win you’re aiming for today? #GoodMorning #MorningVibes

2 weeks with Grok Bot and I’m still amazed. This isn’t just another AI chat. It’s a persistent teammate that keeps memory across everything, runs on its own cloud computer 24/7, executes routines while you’re away, and signs into your tools to actually finish the work. Whatever you can imagine, it can do. Truly next level. #GrokBot #xAI

When do you put the tree up — and are the lights on a smart plug, timer, or voice command? Team Early + Automated or Team After Thanksgiving + Classic?

Chinese team just hit 3.2 Pb/s on a single 24-core fiber (real-time, bidirectional, 10.3 km). That’s 20,000+ 4K movies in one second — or enough bandwidth to make AI data centers feel local. With speeds like this I’d stop “downloading” and start: • Syncing entire model training sets between clusters in seconds • Streaming true holographic telepresence • Pulling full digital twins of cities or factories on demand Backbone capacity is no longer the bottleneck. What we do with speeds like this.

Good morning! ☀️ Starting the day the right way — charged up, automated, and ready to roll. What’s the first win on your tech list today? #GoodMorning #Tesla #HomeAutomation

🚨 iPhone Duo CrackGate update Apple has now briefed its stores on how to handle the cracking inner screens. Staff are told to examine units and repair/replace under normal procedures — don’t assume it’s user damage. The pattern is consistent: crack along the hinge (usually lower half), often with a bright stuck-pixel line, sometimes after just a few folds. Apple admits a “small number” of units are affected and is offering AppleCare replacements. Teardowns point to a hinge weak point (center support not fully retracting / thin notched link). Add in the under-display FaceTime camera that reviewers are calling soft/smudgy (reportedly ~4MP), and the $1,999 foldable is off to a rocky start. If your Duo cracks, take it in. #CrackGate #iPhoneDuo #Apple

Elon just said the quiet part out loud: “The cost of serving Grok Bot was unsustainably high relative to revenue.” Grok Bot (the agentic one that actually does things) scaled fast — hundreds of thousands of paid users in a couple months. But agents aren’t chat. They keep burning compute in the background. Flat $30–$200 subscriptions + linear GPU costs = math that doesn’t work yet when people actually use it hard. This is the real constraint on AI agents right now. Not demand. Unit economics.