7 things you should NEVER tell an AI:
1. “Just do whatever you think is best.”
2. “I don't care about the details.”
3. “Make it sound human.”
4. “Use any sources you can find.”
5. “Don't ask me questions.”
6. “Nobody will check this.”
7. “I trust you.”
AI doesn't need more trust.
It needs better instructions.
Bookmark this. You'll eventually make one of these mistakes.
1
66
Before you build the next feature, ask Karpathy's question:
"What is the piece of text I'd copy-paste to my agent to do this?"
If the answer is a single prompt ,the feature probably doesn't need to exist.
Most teams skip this check. Don't be most teams
1
2
84
Most AI tools fail because they were built without a parallel distribution system.
The technical work is done, but there's no pipeline for getting the right people to discover, try, and pay for the product.
2
53
Garbage thinking + AI = faster garbage.
Most people upgraded their tools. Nobody upgraded their thinking.
That's the gap. That's the opportunity.
2
29
Most people use AI to save time.
The top 1% use it to make decisions their competitors won't see coming for 6 months.
That gap is only getting wider.
2
51
Cute framing. But here’s the part Jensen won’t say out loud.
NVIDIA didn’t accidentally build this “AI cake.”
They engineered the dependencies.
CUDA lock-in at the chip layer means a huge share of models trained today are quietly tied to their ecosystem.
Yes, AMD and Intel are sprinting.
Yes, Google, Amazon, and Meta are building custom silicon.
But switching costs? Massive.
And the five layers aren’t equal.
NVIDIA already owns or influences several of them:
• Chips
• Infrastructure partnerships
• And increasingly the model layer via NIM microservices
Jensen calls it an “AI factory.”
What he’s really describing is a vertically integrated AI stack.
Blackwell isn’t just a chip.
It’s closer to a long-term platform commitment disguised as hardware.
The AI buildout is real.
The transformation is real.
But notice who built the toll roads while everyone else was busy building cars.
—someone who’s been watching NVIDIA since the GeForce 256 days
1
80
What if AI replaced the need to work entirely? Society collapses or thrives,your purpose becomes optional.
What would you do with infinite time?
1
50
AI Systems & Power retweeted
Expectation: the age of the IDE is over
Reality: we’re going to need a bigger IDE
(imo).
It just looks very different because humans now move upwards and program at a higher level - the basic unit of interest is not one file but one agent. It’s still programming.
Replying to @nummanali
tmux grids are awesome, but i feel a need to have a proper "agent command center" IDE for teams of them, which I could maximize per monitor. E.g. I want to see/hide toggle them, see if any are idle, pop open related tools (e.g. terminal), stats (usage), etc.
AI Systems & Power retweeted
Announcing Personal Computer.
Personal Computer is an always on, local merge with Perplexity Computer that works for you 24/7.
It's personal, secure, and works across your files, apps, and sessions through a continuously running Mac mini.
AI Systems & Power retweeted
Dr. Roman Yampolskiy just predicted 99% unemployment by 2032.
The leading AI safety researcher and cybersecurity expert laid out a timeline that's hard to ignore:
AGI by 2027, and then five years later, the economy as we know it.
Here's his logic:
"If you have this concept of a drop-in employee, you have free labor, physical and cognitive, trillions of dollars of it. It makes no sense to hire humans for most jobs."
The math is brutal when a $20 subscription does the same work as an employee.
Yampolsky breaks it into two waves:
Wave 1 — Cognitive labor (now → 2027)
"Anything on a computer will be automated."
Writing, coding, analysis, design, customer service. Gone.
Wave 2 — Physical labor (2027 → 2032)
"Humanoid robots are maybe 5 years behind."
Once AGI arrives, the robots follow. Then there's nowhere left to hide.
"We're looking at a world where we have levels of unemployment we've never seen before. Not talking about 10% unemployment, which is scary, but 99%."
The only jobs that survive?
"Jobs where for whatever reason you prefer another human would do it for you."
Think about what that list actually looks like — it's devastatingly short.
Most people are sleepwalking into this while the AI conversation stays stuck on productivity gains, not on what happens when hiring humans becomes economically irrational.
Yampolsky is predicting total replacement of the human workforce as we know it.
If 99% unemployment arrives by 2032, most people have less than 7 years to figure out what they're building their life around.









