고정된 트윗
Vikram M
23.7K posts
Vikram M
@Vvikramai
Techy but not interested in coding
In The world of Ai 가입일 Temmuz 2024
566 팔로잉2K 팔로워
Vikram M 리트윗함
7400 followers in 83 days 🥳
need 150 followers a day
to hit 10k by day 100
"only" got 87 today
English
Satya Nadella was asked directly: in two years, will Microsoft have more engineers or fewer ?
He didn't answer with a headcount.
He answered with a job description that doesn't exist yet.
In the 1980s, if someone had predicted 3.5 billion people would spend their days typing, the world would have laughed.
Nobody needs 3.5 billion typists.
Except that's exactly what happened and every one of them had a wage, a title, and a career built around it.
Now here's where it gets interesting.
The software developer of the future isn't writing code.
They're managing 100 agents, 1,000 agents and doing something Nadella's team just named for the first time.
"One of the new things that we are learning is what I'll call cognitive coverage."
His point: when your entire codebase is written by agents, the human job becomes comprehending what was built. Auditing it.
Understanding the decisions the agent made and why. That is not a task AI can replace because the AI is the thing being understood.
So do the math on what that means. The workflow changed.
The artifact changed.
The input output format of software development changed.
And the job changed with it not away, but upward.
"That's the job of a software developer.
In order to do that you've got to go to school.
You've got to learn computer science and have cognitive coverage."
Nadella is not saying jobs are safe.
He's saying the jobs that survive are the ones AI cannot verify.
And the unverifiable part of human work the meeting observations, the judgment calls, the things that leave no trace is exactly what no model can be trained on.
I wonder why nobody in San Francisco is talking about that.
English
@Vvikramai 10% GDP needs the workflows to change, and that runs on procurement time, not training time
English
Vikram M 리트윗함
Satya Nadella gave himself a number for AGI a year ago: 10% GDP growth.
We're nowhere close. He knows it. And what he said next is the most underrated take in AI right now.
The bottleneck isn't model capability. It's economics.
His exact framing the marginal cost of productivity improvement has to match the marginal cost of the token.
That's the actual equation for 10% growth.
Not "AI got smarter." Not "more people adopted it."
The cost of running the model has to be priced against the value it actually creates.
Then he said the quiet part out loud: everyone "token maxing" right now burning compute because it feels productive isn't moving that needle.
If the business doesn't benefit, the spend doesn't count.
He admitted he does it too. Calls himself a token maxer.
Says it's addictive you love the tool, then you have to step back and ask what you're actually trying to build.
His fix isn't restraint.
It's matching the tool to the problem.
He name checked Copilot's auto mode for this exact reason stop throwing frontier models at non-frontier problems.
And he's not theorizing.
He built an agent that watches every Microsoft 365 discussion tied to his repo and auto updates the plan in real time.
No token waste. Pure leverage.
Most companies are token maxing right now and calling it AI strategy.
Are you matching token cost to actual value, or are you just maxing ?
English
@Kkk728514079191 Satya’s edge isn’t AI.
It’s knowing that technology adoption is usually constrained by economics, not capability.
English
@Vvikramai He is a really smart guy. No reason he can't turn msft around
English
@SProfiler1 The constraint isn’t intelligence anymore.
It’s whether AI can create $10 of value for every $1 of compute consumed.
That’s the metric that will decide who wins this cycle.
English
@Vvikramai If token cost outpaces incremental ROI, scaling is pure waste. How do you benchmark that break‑even point in practice?
English
Vikram M 리트윗함
The US government just forced Anthropic to pull its two most powerful AI models offline.
No public explanation.
No specific threat. Just a letter on a Friday afternoon citing "national security concerns" and a deadline.
Anthropic's response: kill the models entirely.
Because they couldn't figure out how to verify who was a foreign national including many of their own employees.
Here's where it gets interesting.
The whole thing started because Amazon researchers found a way to bypass Fable 5's guardrails.
Andy Jassy walked that finding straight to the White House. And it spiraled from there fast.
But cybersecurity experts aren't buying the national security framing.
They signed an open letter demanding Trump reverse the order.
Their argument pulling these models doesn't make America safer.
It pulls advanced cybersecurity tools away from the very people defending US networks.
Anthropic itself pointed out the jailbreaks found in Fable 5 exist in several other models too.
So why just Anthropic ?
Because Anthropic has never gotten along with this administration.
There's an active lawsuit.
The government already labeled them a supply chain risk.
And now any excuse however thin becomes a hammer.
Then there's the contradiction nobody wants to say out loud.
A week before Fable launched, Anthropic was warning the world that AI was getting dangerously powerful and everyone needed to slow down.
Then they dropped their most capable model ever and called it too dangerous for public release but available to paying users.
You can't call your own product a threat to civilization and then act surprised when the government takes you at your word.
The strangest twist: Claude downloads spiked after the last government clash with Anthropic.
Turns out, "the model so powerful they banned it" is a compelling pitch.
In AI in 2026, a government crackdown might be the best marketing you never asked for.
Who do you think actually benefits from this Anthropic, OpenAI, or neither ?
English