THE GUY WHO BUILT GROK BOT SAYS YOUR BOT SHOULD NEVER NEED YOUR LAPTOP AWAKE. 4 JOBS I START FROM MY PHONE AND COME BACK TO DONE.
his first rule when they built it: you should never have to ask "does my computer have to be awake?"
so it all runs in the cloud, on the bot's own computer. that "opens up a lot of really amazing opportunities to text your bot, kick it off from your phone"
4 i start from my phone:
1. "my package is 3 days late. open a claim and get me a refund or a new one"
2. "read the 40 page lease i just got and list anything unusual"
3. "compare my phone bill to the last 6 months and tell me what went up"
4. "find 3 flights home for thanksgiving under $400 and send me the best one"
the text i start each one with:
"while i'm out: [job]. message me only when it's done or if you need a yes."
a bot that works while you're away needs locked logins. how i set mine up, article below ↓
A SPACEXAI ENGINEER BANNED CODE COMMENTS FOR HER AI AGENTS. SHE SAYS 99% OF THEM WERE USELESS.
lauren tan builds grok bot. her CI "looks pretty annoying because there's checks for everything"
"I actually banned code comments as well."
her reason: "99% of the time, agents just write code comments that kind of describe some historical thing that is actually totally irrelevant to the code"
one agent left "Lauren said you should never do this" in the code. she meant that one PR, not a rule forever
her whole approach in one line: "everything you can imagine like the agents are bad at, we ban."
5 checks i added for agent-written code:
1. no new comments in a diff unless a human approves them
2. no file over 400 lines
3. no new dependency without a note in the PR
4. tests can't be edited in the same PR as the code they test
5. no console.log or debug prints left in
the prompt to build yours:
"read my last 30 merged PRs written by agents. list the mistakes that repeat. turn each into a CI check that fails the build, with the exact rule and how to test it."
agents leak secrets into code too. how grok bot guards my keys and logins, article below ↓
ELON MUSK REVEALED THE ONE INSTRUCTION HE'LL GIVE GROK SO IT DOESN'T WIPE US OUT: "HEY GROK, THAT'S YOUR DADDY."
the question: what happens to humans when AI is far smarter than us.
the host put it in chimp terms. humans could wipe out every chimpanzee. we built protected zones instead. is that the best case for us?
elon's answer:
"I think Grok would care about expanding human civilization."
"Hey, Grok, that's your daddy."
"Don't forget to expand human consciousness."
his plan for the future is a reminder to the AI. mine is a bot that watches my accounts, article below ↓
ELON MUSK WAS ASKED WHAT A 20-YEAR-OLD SHOULD LEARN BEFORE AI DOES EVERY JOB. HE DIDN'T SAY CODING.
his answer: "as broad-based an education as possible."
"in order to know what to ask the robots for, you need to be able to formulate the question."
how a good question to AI looks:
→ who it's for: "a landing page for dentists in austin"
→ what done looks like: "one page, price list, booking button"
→ what you already have: attach the logo, the prices, an old version
→ what to avoid: "no stock photos, no blue"
→ how to check it: "list what you assumed so i can correct it"
five lines like that beat an hour of back and forth.
my first request to a bot was plain: watch my accounts. what i asked grok bot, article below ↓
ELON MUSK SAYS AI WILL DOUBLE THE OUTPUT OF EVERY OFFICE WORKER. "THIS WILL HAPPEN VERY FAST."
he called it what his "biological neural net predicts."
$130 trillion world economy
+20 to 30% from digital superintelligence alone
= "call it a $25 trillion increase"
ron baron checked the number twice. elon kept it.
then the part about your job:
"it probably doubles the productivity of white-collar work."
and he says that's only the screen part. robots come on top of it.
if AI does half my desk work, it can watch my logins too. grok bot security setup, article below ↓
JENSEN HUANG MOCKED THE AI DOOM WARNINGS 7 MONTHS BEFORE HIS OWN BOARD WANTED HIM OUT. SAM ALTMAN SIGNED ONE.
the letter: "mitigating the risk of extinction from AI should be a global priority." bill gates, sam altman and dario amodei signed it. jensen didn't.
slow it down <=> speed it up, line by line:
"I am worried about it."
AI warnings, in his words, are like a car maker telling you every day that cars kill.
"it's a celebration of their own achievements."
"they believe that only they can create a technology and keep it safe"
on slowing AI down: "I believe is exactly the opposite. we had to speed up AI innovation."
full speed is fine. just lock your doors first. grok bot security setup in the article below ↓
ELON MUSK: IN 36 MONTHS THE CHEAPEST PLACE TO RUN AI WON'T BE ON EARTH.
his problem with earth is electricity.
chips grow exponentially. power output outside china is flat.
"Magical power sources? Magical electricity fairies?"
in space a solar panel makes about 5x more power.
no night, no clouds, no seasons. the atmosphere alone eats about 30%.
no batteries needed either.
"36 months or less, maybe 30 months."
after that, he says, it gets "ridiculously better" up there.
before AI goes to space, put it on guard down here. grok bot security article below ↓