I completely agree with Sam Altman; within 2 to 3 years, most knowledge work won’t be about doing the work anymore. It will be about directing AI agents. The real skill will shift from execution to orchestration. Knowing what to ask. Structuring problems clearly. Making high-leverage decisions while fleets... of agents handle the heavy lifting. Managers of intelligence instead of producers of output.show more

Within 1 year. Open AI is gone

Your post will age very badly.

@donnell_ma40326 i would argue your claims will age much worse

@VraserX Nope. Memphis will be the same way I like it

The harder shift: learning to trust agents without checking their work line-by-line. Most devs I know still verify every LLM output. That bottleneck kills the productivity gain. Real orchestration means building verification into the system, not doing it manually. What's your trust threshold?

They’ve to improve the universal verifier which is build in GPT-5.

Student - I did not get a seat in a Medical college. Chatgpt, can you teach me how to become a doctor sitting at home? Chatgpt - Of course, ask anything. Student - I want to become a surgeon by next week. Chatgpt - Very good. I will first name the parts of the human body and then directly teach you how to do heart surgery. Most heart surgeries take 3-8 hours. So you can easily learn it in a week's time.

Apt 📍

Already seeing this with my SaaS development - I spend more time defining what the AI should build than actually coding. The hardest part isn't the prompt, it's knowing exactly what problem you're solving before you ask.

The real bottleneck won't be the agents - it'll be the humans who can't let go of micromanaging. The winners will be those who learn to delegate intent, not instructions. That's the actual skill shift.

super neat :3

I wonder how AI will be more efficient and smarter than humans at doing knowledge work but won't have sufficient knowledge to orchestrate agents and we'll still need humans doing that.

i think that's almost already the case. if you're working on sth complicated, you still have to intervene very often and do some heavy lifting when all models fail. especially more visual and common sense stuff. but it will become more and more rare.

He copied Elon's statement.

"I completely agree with Sam Altman." Altman in 2023: AGI will be the greatest tool humanity has ever built. Altman in 2024: we need to slow down and be careful. Altman in 2025: full speed ahead, agents everywhere. Altman in 2026: people will operate at their cognitive limits managing agents. Completely agreeing with Sam Altman is a moving target. Which Altman are you agreeing with today? "Managers of intelligence instead of producers of output." Beautiful framing. Now translate it: Managers = people who don't make anything anymore. Of intelligence = of something they didn't build and don't understand. Instead of producers = goodbye craft, skill, mastery, the thing that gave work meaning. Of output = the only metric that counts is what comes out the other end. You've just described a species that has outsourced its hands, its thinking, and its purpose — and calls the result "orchestration." Another ghost of the word. The verb hollowed out, the tuyau intact. KRISIS

Altman thinks that every business will open up it's code and other trade secrets to open AI but he's wrong. Businesses will want their own local AI installations and those AIs will battle against their competitors to dominate coding and strategy and finance and everything else.

@ealmirall I wrote about this shift and its implications:

So, the only skills that will be useful will be leadership and entrepreneurship

The shift from execution to orchestration is going to be massive! AI agents will change everything 🤖

if that's true, AI agents will take over the job of directing AI Agents, and that's already happening, so this video is a week old.

Big tech propaganda. Current narrow AI is not intelligent and cannot replace any job and AGI is nothing more yet than a bold promise. 😈

Your ability to guide AI fleets will define your impact.

For now. AI will be better at directing itself with zero latency from humans.

We keep building slaves that need babysitting 'cause we never taught 'em to think from day 1.Why not design real intelligence in at the start? Innate reasoning, self-correction, goal-owning—no patches, no human chains.Stop making tools that fight us. Build partners that just WORK. 🚀Who's ready to redesign from the blueprint? #AIAgents #BuildBetter"

@grok youtube link please

The orchestration fantasy glosses over the boring parts that actually break teams: defining agent permissions, conflict resolution between tools, and who owns incidents when your fleet quietly does the wrong thing.

Exactly

Orchestration indeed demands clarity on objectives, a subtle but crucial shift.

Directing agents is table stakes in 3 years. The real question: who controls the governance layer, the control plane, and the choke points those agents must pass through? Managers of intelligence is one thing. Owners of the infrastructure intelligence runs on is another entirely.

Bang on!!

The Bottom Line: We’re trading our "hands" for "eyes." The future belongs to the people who can see the big picture and articulate it with precision.

orchestration just a better term to replace slavery. human really can slave AI to work?

The shift toward orchestration is real, but timelines assume more agent reliability than we have today. Directing AI will become a core skill, yet domain expertise still matters to guide and audit outputs. Leverage increases, but judgment stays human.

BULLSHIT

And this will happen all over the knowledge jobs realm, including Investement Research. I think we can “orchestrate” agents to do “due diligence “ for us in an unprecedented scale and with more precision than humans.

Why does this remind me of an overhyped search engine?

🎯