You know what costs more than a rocket launch?

The meetings about the rocket launch.

NASA’s major projects have racked up $12 billion in cumulative cost overruns and 28 years of schedule delays. The James Webb Space Telescope - gorgeous piece of engineering, absolute masterpiece - came in 175% over budget and 7 years late. Boeing’s Starliner burned through $5 billion while being, charitably, “behind schedule.”

Cannot keep my tongue not to mention the Ariane 6 of ESA etc..

And here’s the thing, most of those overruns aren’t engineering failures.

They’re coordination failures. Coordination.

The thermal engineer didn’t talk to the power engineer. The change request sat in someone’s inbox for three weeks. The review meeting got rescheduled because the propulsion lead was at a conference in Bremen.

So I did what any reasonable person would do.

I replaced the entire team with AI agents and put it on GitHub.

24 Agents, Zero Coffee Budget

Let me explain.

[Space Mission Agent Team](https://github.com/Gorfsharpem/space-mission-agent-team) is a multi-agent system that mirrors the exact organizational structure of a real satellite project. Not a simplified version. Not a demo. The whole thing.

Project Manager. Mission Director. EPS Engineer. ADCS Engineer. Thermal. Structures. Propulsion. TT&C. OBSW. Safety. QA/PA. Ground Segment. Launch Campaign Manager. Bid Manager. Proposal Manager. R&D Manager.

Twenty-four agents. Each with domain-specific knowledge. Each producing real engineering deliverables - link budgets, power budgets, Tsiolkovsky calculations, hazard registers, FMEA extracts. Not summaries. Not bullet points. Actual tables with actual numbers.

They communicate through a pub/sub message bus. Any agent can query any other agent by name. There’s a dependency graph that tracks who affects whom. And they follow ECSS standards; the same European space standards that govern every ESA mission.

Yes, the European space standards. All 139 of them or 141 or more. The ones that are so prescriptive about roles, deliverables, and review gates that they’re basically a specification for an agent system. Turns out the ECSS accidentally wrote the world’s longest AI prompt. :)

The Part Where They Fight

Here’s where it gets interesting.

Most multi-agent demos show agents collaborating beautifully. Agent A hands off to Agent B, everyone agrees. Very LinkedIn. Very inspirational.

That’s not how engineering works.

In real satellite projects, when someone proposes a change, every affected subsystem pushes back. The structures engineer says “that adds 2kg I don’t have margin for.” The thermal engineer says “that blocks my radiator view factor.” The QA lead says “show me the test plan or this isn’t happening.”

So I built that. On purpose.

When any agent raises a change, the system resolves all affected agents through the dependency graph. Each affected agent performs a critical review - and they do NOT agree by default. Every review must cover impact, objections, interface conflicts, alternatives, and a verdict. The verdict options are: ACCEPT, ACCEPT WITH CONDITIONS, or REJECT (and if you reject, you’d better propose an alternative).

The Mission Director handles the conflicts. QA/PA validates traceability. Only then does the baseline update.

Adversarial review isn’t a bug. It’s the entire point. Because rubber-stamping is how you get $12 billion in overruns.

“But Does It Actually Work?”

You can start a full mission from a one-line business objective. Or deep-dive into a single subsystem. Or propagate a change from any agent and watch the ripple effects. Or run a parallel milestone review where all agents produce domain-specific deliverables simultaneously — at SRR, PDR, or CDR and Deltas etc...

“python start.py --from User --mission “MY-SAT-1” --description “LEO earth observation, 500km SSO, 5-year lifetime”

One command. The system spins up business analysis, flows into mission analysis, parallelizes subsystem design, runs QA gates, plans ground segment, runs launch campaign planning, and produces a master mission design document.

There’s also a web dashboard if you’re not a terminal person.

(No judgment. Okay, slight judgment.)

Why Open Source This?

The space industry is famously paranoid about intellectual property. ITAR restrictions. NDAs stacked on NDAs. Export control classifications that make sharing a PowerPoint feel like espionage.

And look - I get it. Specific mission data, proprietary sensor designs, classified payloads - that stuff should be locked down.

But the process of designing a satellite? The fact that you need a power budget, a link budget, a mass budget, and a thermal analysis? That the structures engineer needs to talk to the propulsion engineer about center of mass?

That’s not a secret.

That’s literally published in ECSS-E-ST-10. It’s a standard. It’s public.

The knowledge of how satellite projects are organized has been open for decades. I just made it executable.

The Uncomfortable Realization

Here’s the human nature part, and I’ll keep it brief because it’s a bit uncomfortable.

People spend entire careers building institutional knowledge about how satellite missions work. That knowledge lives in standards documents, review templates, and the heads of senior engineers who’ve been through fifteen mission cycles.

Maybe 80% of that knowledge is procedural. Coordination. Documentation. Making sure the right people reviewed the right thing at the right time. The genuinely creative engineering- the novel solutions, the elegant trade-offs, the “what if we flip the spacecraft upside down” moments, that’s maybe 20% of the work.

The 80% is important. Critical, even. But it’s also encodable.

2026 is officially being called “the year of multi-agent systems.”

Forty percent of enterprise apps are expected to have embedded agents by year’s end.

We are sure that the space industry will get there too - probably slower than everyone predicts (it’s space, we love our heritage designs) but faster than the old guard is comfortable with.

In the meantime, the repo’s on GitHub. MIT licensed.

Agents who never miss a review, never go to conferences in Bremen, and never let a change request sit in someone’s inbox.

They do, however, argue with each other constantly.

Just like a real engineering team ;)

Let me tell you about my Easter.

I paid a travel agent to plan a trip with my son. Simple enough, right? A few flights, a hotel, maybe a car rental. That’s literally their job. That’s what the invoice said. That’s what I paid for.

Instead, I lost days.

Days comparing flights myself and visa and other documentation and for every single details she texts me and eventually forgets what I have said since days ahead.

Then, days cross-referencing hotel availability. Hours figuring out transfer logistics, check-in times, whether the rental place is even open on Easter Monday. Days doing the exact job I paid someone else to do - while my actual work piled up, while my son asked “are we going yet?”, while I sat at the kitchen table at 2 AM with six browser tabs open wondering why I’m functioning as an unpaid travel agency.

The agent? Sent me one email: “Hi! Just checking in on the itinerary? :)”

Days of my time. Time I should’ve spent doing my job. Which means someone waiting on me now has to wait longer. And someone waiting on them waits even longer.

That’s not a bad holiday. That’s a $8.9 trillion economic phenomenon.

The Chain Reaction

Here’s what happens when someone can’t do their job: you end up doing it for them.

The contractor does shoddy work, so you spend your weekend fixing it. IT support can’t solve a basic issue, so the engineer fixes her own computer. The government clerk can’t process paperwork, so you visit the office six times, each time bringing a document they forgot to mention.

Every time an incompetent person touches a task, they don’t just fail at it - they export that failure to everyone downstream. Your productivity drops. Then the productivity of everyone waiting on you drops. It cascades.

Researchers have a name for this: the competence tax. The more capable you are, the more you end up subsidizing everyone else. You attend meetings to “provide input” on things that aren’t your job. You mentor colleagues who should already know this. You redo work that was done wrong the first time. You become the barrel that everyone lines up behind - until you burn out.

And here’s the cruel part: nobody notices. Because competence is invisible. Nobody writes a review saying “the transaction was processed correctly and on time.” But incompetence? We’ve been marinating in it so long we think it’s just... how things work.

One Bad Apple: The 30-40% Rule

You might think one incompetent person is just one person’s problem. The research says otherwise.

Will Felps studied what happens when you put just one underperformer in a team. Not a team of bad people - one bad apple among good ones. The result? Team performance dropped by 30-40%.

But it gets worse. The incompetence was contagious. Team members started arguing more, giving up sooner, and - this is the part that should terrify you - adopting the bad apple’s behavior. The incompetent person didn’t rise to the team’s level. The team sank to theirs.

Now multiply that across every team, every department, every organization in the economy. Gallup puts the global cost at $8.9 trillion in lost productivity annually. That’s 9% of global GDP. Gone. Not because of recessions or wars or natural disasters. Because of people who can’t do their jobs and the people who have to pick up the slack.

Why Nobody Gets Fired

Companies know who the dead weight is. Everyone knows. The team knows. The manager knows. HR knows. And nobody does anything about it.

Why? Because the math is perverse.

Firing someone means paperwork, legal risk, lean staffing gaps, and zero credit for the manager who pulls the trigger. Meanwhile, the competent employees will just... absorb the extra work. Because that’s what competent people do. They don’t complain. They just quietly carry twice the load.

Managers are also victims of their own cognitive bias. When they depend on someone - even an underperformer - they blame external factors for poor output instead of the person. “Oh, the project was complex.” “The timeline was tight.” Never: “Dave simply can’t do this.”

The system doesn’t tolerate incompetence out of kindness. It tolerates it because competence is exploitable and self-correcting. The good people will fix it. Until they don’t - and then they leave. Which costs 50-200% of their salary to replace.

The cycle feeds itself.

Now Let’s Talk About Countries (Everything can be translated onto geography)

Here’s where this stops being an office problem and becomes a civilizational one.

In Denmark, if a bank clerk is incompetent, it barely matters. The digital system catches errors. The process is automated. The institution compensates for the individual. You might not even interact with a human.

In a developing country, that same incompetent clerk is the system. There’s no automation catching mistakes. No process backing them up. No accountability mechanism flagging errors. When they fail, you become your own bank, your own lawyer, your own HR department.

An entrepreneur in a low-income country spends 50% of the country’s per-capita income just to start a business. In a high-income country? 4.2%. Same paperwork, same concept - wildly different institutional competence behind it.

Vietnam and Bangladesh have nearly identical GDPs. Vietnam has 2.5 million garment workers exporting $48 billion. Bangladesh has 4 million garment workers exporting $42.83 billion. Fewer people, more output. The difference isn’t the workers - it’s the system around them.

Developed countries didn’t eliminate incompetence. Let me be clear about that. They just built systems that quarantine it. When one person fails in Stockholm, the institution absorbs the shock. When one person fails in a weak-institution country, the failure cascades - and everyone downstream pays the competence tax.

The Brain Drain Isn’t the Disease - It’s the Symptom

And here’s what happens when competent people get tired of carrying the load in countries where the systems can’t protect them.

They leave.

Sub-Saharan Africa loses 70,000 skilled professionals every year. Zimbabwe has lost 68% of its college graduates. Ghana loses 500 nurses per month to Western countries. The most competent people, the ones who could theoretically fix the institutions are the first to go.

We call this “brain drain” like it’s the problem. It’s not. It’s the immune response. The real disease is the system that makes staying feel like a punishment. You’re a qualified engineer, and you spend half your time navigating bureaucracy that a functioning institution would handle in minutes. You’re a doctor, and you can’t get supplies because procurement is run by someone’s nephew.

Speaking of which nepotism is the incompetence engine that keeps on giving. 80% of people in studies acknowledge it hurts the economy. But it persists because the incompetent people doing the hiring... hire more incompetent people :). Their relatives. Who hire their relatives. The cycle is self-reinforcing. Merit never enters the building.

What This Really Says About Us

The uncomfortable truth.

When government paperwork takes six visits, we don’t blame the clerk, (Like I mentioned above in corporate environment) we blame “the system.” We’ve externalized individual failure into institutional identity.

The social contract of mediocrity protects everyone. If we started actually demanding competence across the board, the whole house of cards gets uncomfortable, because nobody wants to be the first one measured against that standard.

So we keep paying the tax. And the economy bleeds $8.9 trillion a year while we all pretend this is just how things work.

Some countries built systems to hide the bleeding. Others can’t afford the bandage.

Either way, you’re paying for someone else’s incompetence right now. You just call it “a normal day” ;)

The narrative around the “agent economy” treats it as an emerging opportunity - a technology trend that forward-thinking space companies might choose to adopt. That framing misses the point entirely.

The space sector went from 3,371 active satellites in 2020 to approximately 14,000 by 2026. ESA projects 100,000 by 2030.

The agent economy did not arrive in space as an innovation. It arrived as a structural necessity. The industry scaled its orbital assets far beyond its capacity to operate them with human-centric ground systems.

This analysis examines where that pressure is already reshaping operations, where the market is heading, and what the competitive consequences will be for operators, manufacturers, and service providers.

From Consumer AI to Mission-Critical Autonomy

A significant shift is underway in the origin of AI systems being deployed in space operations.

In December 2025, NASA’s Perseverance rover completed its first AI-planned drive on Mars using Anthropic’s Claude — a general-purpose large language model, not a bespoke aerospace system. The rover executed autonomous drives of 689 and 807 feet, with JPL validating over 500,000 telemetry variables through a digital twin prior to command transmission. The underlying intelligence, however, was commercial-grade AI.

This pattern is repeating across the sector. AIKO, a Turin-based startup, became the first European company to demonstrate deep learning running in orbit. Their software enables satellites to process imagery, assess collection priorities, and replan missions autonomously — without ground intervention. The spacecraft makes its own operational decisions in real time.

The strategic takeaway: the AI capability gap between commercial and aerospace-grade systems is narrowing rapidly. Mission-critical autonomy is no longer the exclusive domain of purpose-built defense systems. General-purpose AI agents are crossing into orbital operations, and that crossing is already behind us.

Market Sizing and Performance Benchmarks

The quantitative case for agent-driven space operations is substantial:

Market scale: The AI-in-space-operations market reached $2.36 billion in 2025, projected to grow to $15.05 billion by 2034 at a 22.91% CAGR. The broader AI-in-space-exploration market is on track toward $110 billion by 2035.

Operational performance: Cognitive Space’s CNTIENT.Optimize AI outperformed traditional heuristic-based satellite scheduling algorithms by 4x in head-to-head comparison, delivering 87% time savings per operator per week in parallel testing with a commercial constellation operator.

Robotics and servicing: Stanford researchers demonstrated machine learning-driven planning for ISS robots achieving 50–60% faster task completion than traditional methods. Northrop Grumman is developing “cognitive” spacecraft using NVIDIA’s Omniverse and Isaac Lab — simulation platforms originally designed for gaming and robotics — to train autonomous docking and in-orbit servicing capabilities.

Defense investment: The U.S. Space Force budget is projected at $40 billion for FY2026, a 40% year-over-year increase, driven significantly by demand for resilient, AI-enabled space architectures.

Enterprise adoption trajectory: Gartner forecasts that 40% of enterprise applications will incorporate task-specific AI agents by end of 2026, up from less than 5% in 2025. The space sector is not leading this adoption curve — it is being pulled along by it.

The Migration of Intelligence from Ground to Orbit

The common framing of “AI replacing ground controllers” is imprecise. What is occurring is more fundamental: operational intelligence is migrating from terrestrial ground stations into the spacecraft themselves.

Sixty years of investment in mission control infrastructure — Houston, Darmstadt, Bangalore — built the foundation of space operations around centralized human oversight. That architectural assumption is now being inverted. The satellite is becoming its own ground control.

The Canadian Space Agency is funding 75% of a yearlong mission — in partnership with Mission Control (the company) and Spire Global — to demonstrate continuous deep learning execution aboard an orbiting satellite. Governments are not merely permitting autonomous orbital operations. They are actively financing the transition.

This migration is being accelerated by a parallel hardware shift. Airbus (OneWeb), Thales (Inspire), and MDA (Aurora) are deploying fully software-defined satellites capable of reconfiguring beam patterns, frequencies, and service capabilities post-launch. These platforms are designed for dynamic, autonomous management.

However, a critical gap remains: the AI systems required to manage fleets of reconfigurable satellites at scale are still maturing. The industry has deployed reprogrammable spacecraft before fully developing the autonomous systems to operate them — creating both an urgent market need and a significant integration challenge.

Competitive Implications and the Consolidation Risk

There is a widely held assumption that AI agents will democratize space operations by reducing per-satellite costs. This assumption warrants scrutiny.

The capital investment required to develop, validate, and deploy autonomous operations stacks — encompassing onboard AI, digital twins, simulation environments, and ground-to-orbit integration — is substantial. Mega-constellation operators such as SpaceX and Amazon possess the scale, data, and capital to build these systems internally. Smaller operators increasingly face a choice: develop proprietary autonomy capabilities at significant cost, or procure agent-as-a-service from the very companies they compete against.

The cost per satellite decreases, but the barrier to competitive entry increases. Autonomy is emerging as a structural moat, not a market leveler.

Compounding this dynamic is the trust and regulatory bottleneck. The technology for autonomous satellite operations has been demonstrated — AIKO in orbit, Cognitive Space outperforming human planners by 4x. But regulatory frameworks, insurance underwriting models, and defense procurement protocols remain anchored to the assumption of human-in-the-loop oversight. Autonomous spacecraft will, in practice, still require human approval gates for the foreseeable future.

The workforce implications are also material. A 2026 Mercer survey found that 40% of employees now report concern about AI-driven job displacement, up from 28% in 2024. In the space operations sector — where the operator-to-asset ratio is already collapsing — that concern reflects an accurate reading of the trajectory.

Strategic Outlook

Three dynamics will define the next phase of the agent economy’s impact on the space sector:

1. Autonomy as competitive infrastructure. Operators who treat AI agents as an optimization tool will fall behind those who treat them as core operating architecture. The question is no longer whether to deploy autonomous systems, but how quickly an organization can integrate them into mission-critical workflows without unacceptable risk.

2. The trust gap as the binding constraint. Technical capability is outpacing institutional readiness. The organizations that solve the trust problem — through transparent AI decision-making, robust digital twin validation, and proactive regulatory engagement — will define the operational standards for the next generation of space systems.

3. Consolidation through autonomy. The capital intensity of building autonomous operations stacks will drive consolidation. Smaller operators will either be acquired, become dependent on third-party autonomy platforms, or find defensible niches where human expertise remains irreplaceable. The window for independent competitive positioning is narrowing.

The space industry has consistently built systems that outpace its ability to manage them — from orbital congestion to spectrum coordination to debris mitigation. The agent economy is the latest instance of this pattern, and potentially the most consequential. The sector does not have the luxury of gradual adoption. The assets are already in orbit. The operators are already outnumbered.

The strategic question is not whether AI agents will reshape space operations. It is whether individual organizations will be positioned to lead that transition — or be restructured by it.

Fakhri Babayev | Equinox Consulting — Mission Design, Engineering, Program Management

You’re not going to read this article.

I mean, statistically, you won’t. You’ll skim it. Your eyes will do that F-pattern thing - read the first line, scan down the left side, maybe catch a bold word or two, and then decide if it’s worth your 10 bits per second.

Don’t feel bad. That’s what brains do now.

Here’s where we are: 52% of new online articles are AI-generated.

Not edited by AI. Not assisted by AI. Written entirely by machines, published, indexed, and sitting in your search results right now.

That was 5% before ChatGPT launched.

But wait, it gets better. Europol estimates up to 90% of online content could be synthetic by the end of 2026. Bot traffic already surpassed human traffic in 2024 - for the first time in the history of the internet, more than half the activity online isn’t coming from people.

And the cost of producing an AI blog post is less than one cent. Whereas, human writer charges $10 to $100 for the same piece.

Your Brain Didn’t Get the Memo

Caltech published a study in Neuron (December 2024) that measured the actual processing speed of human thought. Not sensory intake - we take in about a billion bits per second through our eyes, ears, skin. No, the actual thinking speed.

Ten bits per second.

That’s it. That’s what you’ve got. Ten bits per second to process a world that’s generating 402 million terabytes of data every single day. The average person now makes 4,909 digital interactions daily — up from 298 in 2010.

Your working memory holds four items. Not seven (that was debunked). Four.

You were never designed for this. Nobody was.

The Decline Was Already Happening

Let me be clear — AI didn’t start this fire. It just dumped jet fuel on it.

Reading was collapsing globally before ChatGPT existed. 49% of 15-year-olds worldwide now read only when required - up from 36% a decade earlier. The OECD’s PISA scores show steady reading performance declines across most industrialized countries over the last ten years.

South Korea is the canary in the coal mine. In 1994, 87% of Korean adults read at least one book a year. By 2023, the nummber is 43%.

Only 32.7% of children aged 8-18 enjoy reading in their free time. That’s the lowest point in two decades. And only 41% of children under 4 are being read to regularly - down from 64% in 2012.

Oxford’s 2024 Word of the Year was “Brain rot. You are getting there, right? :)

The Deep Reading Circuit Is Dying

Maryanne Wolf at UCLA describes something she calls the “deep reading circuit” - a network in the brain that develops through sustained, focused reading. It’s what gives you empathy when reading fiction. Critical thinking when reading arguments. Imagination when reading descriptions.

Brief-burst processing - the scrolling, skimming, snippet-consuming way we interact with text now - “curtails the development of the contemplative dimension of the brain.”

Translation: skimming isn’t a different way of reading. It’s the absence of reading.

We know this. We’ve known this. We just keep scrolling anyway.

The Absurdity

Here’s my favorite part.

Google’s Deep Research tool scanned 147 sources to answer a single prompt. The tool designed to help you deal with too much information... created a 15-page document you now have to read. The solution to information overload is more information.

Natalie Wexler wrote an article about the dangers of using LLMs to summarize text. Then she used an LLM to summarize her own article for promotion. the result was, indeed, ironic.

The World Learned to Read Just in Time to Stop

What really haunts me is this:

UNESCO reports that global youth literacy hit 93% in 2024. The highest it’s ever been. More humans can read right now than at any point in history.

And they’re choosing not to. :)

Not because they can’t. Because the economics of attention have fundamentally shifted.

God, I love that notion.

The economics of attention have fundamentally shifted.

When content is infinite and free and mostly machine-generated, reading stops being a pathway to knowledge and starts being a filtering problem. And the filters we’re building (AI summarizers, AI curators, AI research agents) just add more content to the pile.

Nobody is proposing how reading should evolve. They’re all proposing better filters. That’s like inventing a bigger bucket instead of fixing the leak.

What Comes Next

The real question isn’t “how do we get people to read more?” That ship has sailed. The content volume went exponential. Human cognition is fixed at 10 bits per second. Those lines are never crossing again.

The real question is, what replaces reading as the primary method of knowledge acquisition?

“Written by a human” is already becoming a luxury good.

We may be the last generation that reads and writes in the traditional sense. Not because literacy dies - but because the definition of literacy changes into something we don’t have a name for yet.

Something has to emerge. Not better filters. Not faster reading. Something fundamentally different.

I just hope whatever it is, it still makes you feel something ;)

And everyone is nodding along like this is profound wisdom :)

Can we pause for a second and ask the obvious question?

Why on earth should we listen to these guys?

The Credibility Problem

Let’s check the receipts.

Elon Musk has a bachelor’s in physics and economics from UPenn. He enrolled in a PhD in applied physics at Stanford and quit after two days. Two. Days. Then he went and built software companies. PayPal? Software. Tesla’s edge? Software. SpaceX’s cost revolution? Largely software-driven manufacturing.

Jensen Huang? Bachelor’s and master’s in electrical engineering. Not physics. Not mathematics. Electrical engineering. And then he built a software-and-chips empire.

So the guy who quit physics after 48 hours and the guy who never studied physics are telling you that physics is the future.

This is like a fast-food CEO telling you to eat organic. Technically not wrong. But maybe not the person you should be taking dietary advice from.

The Survivorship Bias Nobody Mentions

Here’s what drives me crazy about billionaire advice. We only hear from the winners.

For every Musk who dropped out and built SpaceX, there are thousands who dropped out and built nothing. Research consistently shows that college dropouts earn less and are 70% more likely to be unemployed than graduates. But we don’t interview those people on podcasts.

When a billionaire says “study physics,” what they’re really saying is: “physics is useful for the kind of problems , “work on”. Robotics. Autonomous vehicles. Rocket engineering. Space. That’s their world. It’s not everyone’s world.

They’re not giving career advice. They’re placing a recruiting ad.

But Does the Research Actually Support “Back to Basics”?

Here’s where it gets uncomfortable for the “physics is king” crowd.

Physics graduates currently have a 7.8% unemployment rate - the second worst out of 70+ popular degrees analyzed. Only about 28% of physical science majors end up working in STEM fields at all. Meanwhile, computer science graduates start at a median of $80,000, climbing to $115,000 mid-career, with far clearer job paths.

So the data doesn’t exactly scream “everyone go study physics.”

However

There’s a nuance the raw numbers miss. The World Economic Forum projects that 59% of workers will need reskilling by 2030. AI will displace 92 million jobs and create 170 million new ones. The question isn’t what degree gets you hired tomorrow — it’s what thinking keeps you employable for the next forty years.

And that’s where the fundamentals argument gets interesting.

The Thing AI Can and Cannot Do

In 2025, AI scored gold at the International Mathematical Olympiad. Thirty-five out of forty-two points. Incredible. But five human contestants got perfect scores. The AI didn’t.

More revealing: OpenAI’s own researcher admitted the Physics Olympiad is “definitely harder” for AI because it has an experimental section. The AI can reason on paper. It cannot run an experiment. It cannot feel what happens when friction meets momentum.

Babies — literal babies who can’t walk yet — show surprise when objects violate basic physics. They intuitively understand that things fall down, not sideways. That’s not knowledge. That’s embodied understanding built from living in a physical world. No language model has that. No matter how many physics textbooks it ingested.

So here’s the real answer to “AI knows physics too, why does it matter?”

AI knows physics the way a GPS knows your city. It can route you anywhere, but it doesn’t know that the shortcut through the park is actually faster on Tuesdays because the school traffic clears by 8:15. That knowledge comes from being there. From having a body. From experiencing the system (that’s the Bluetooth, mesh and stuff), not just modeling it.

What They’re Actually Right About (Annoyingly)

Here’s the thing I hate to admit: Musk and Huang are right about the direction, even if they’re wrong about the specifics.

The “learn to code” era is the same as the “learn Latin” era, the “learn accounting” era, and the “learn Excel” era. Every generation picks a hard skill, turns it into a religion, and then acts shocked when it gets partially automated.

In the 1980s, fourth-generation programming languages were supposed to kill coding. No-code tools have been “replacing developers” for a decade. Ian Bremmer said on television that “learn to code” is now worse advice than getting a face tattoo. Meanwhile, Andrew Ng — an actual AI researcher, not a billionaire pundit — pushed back, comparing this to the shift from punch cards to keyboards.

The coders adapted. They always do.

The real insight isn’t “study physics instead of coding” . It’s that any single skill has a shelf life, and the people who survive every cycle are the ones who understand systems deeply enough to ride the transitions.

Physics happens to be good at teaching that. So is math. So is — wait for it — coding, when done with curiosity rather than just tutorial-following.

The Uncomfortable Truth

We don’t need to “go back” to fundamentals. That framing is wrong. It assumes we left.

We didn’t leave. We just let LinkedIn influencers convince us that a 12-week bootcamp was a substitute for learning how to think. It wasn’t. It was a vocational shortcut. Useful? Absolutely. A replacement for deep understanding? Come on.

The real question isn’t physics vs. coding. It’s this: Are you learning a tool, or are you learning how to learn?

Because AI has all the knowledge. Every equation. Every algorithm. Every physics textbook ever written. What it doesn’t have is judgment - the ability to look at a problem and know which question is even worth asking.

Should you listen to Musk and Huang? Sure. But not because they’re rich. Listen to the pattern: every tool gets automated, every skill gets commoditized, and the only people who stay ahead are the ones who learned the game beneath the game.

That’s not physics. That’s not math. That’s not code.

That’s curiosity. And no billionaire has a monopoly on that ;)

Fakhrim B. @ Equinox Consulting | Airlock Newsletter

If you’re a project manager reading this in 2026, you’ve probably had the dream. You know the one. You show up to work, and there’s a chatbot sitting in your chair, running the standup, and nobody even noticed you were gone.

Thanks for reading Fahrim’s Substack! Subscribe for free to receive new posts and support my work.

Terrifying, right?

Well, I have good news and bad news. The bad news: that chatbot is already here. The good news: it’s terrible at the parts of your job that actually matter.

Let me explain ;)

First, Let’s Be Honest About What Project Managers Actually Do

Here’s a dirty secret the entire profession doesn’t love admitting: about 60% of traditional project management is glorified administration.

Updating timelines. Chasing status updates. Scheduling meetings about meetings. Writing reports that nobody reads. Moving cards on a Kanban board like it’s a spiritual practice.

And guess what AI is really, really good at?

Yeah. All of that.

AI can track tasks, flag risks, generate status reports, allocate resources, and predict delays faster than any human. It doesn’t get tired. It doesn’t forget. And it definitely doesn’t need a coffee break after updating Jira for the 47th time today.

So if your entire value as a PM is “I keep the spreadsheet updated” — we need to talk.

The Great PM Identity Crisis of 2026

Here’s what’s actually happening in the industry right now, and nobody’s framing it correctly.

AI isn’t replacing project managers. It’s exposing which project managers were never really managing anything in the first place.

Ouch.

(I told you this newsletter doesn’t sugarcoat things.)

The PMs who were basically human Gantt charts? They’re in trouble. The PMs who were actually leading — navigating office politics, making judgment calls under uncertainty, getting a dysfunctional team to somehow deliver on time — they’re about to become the most valuable people in the room.

Because here’s what AI absolutely cannot do: convince Dave from engineering to stop being passive-aggressive about the deadline he agreed to three sprints ago.

That’s a human problem. And human problems require human solutions ;)

What AI Actually Changes (And It’s Not What LinkedIn Thinks)

If you scroll LinkedIn right now — which I don’t recommend for your mental health — you’ll find approximately 9,000 posts about “AI-powered project management” that all say the same thing. Predictive analytics! Automated workflows! Real-time dashboards!

Cool. Sure. All true.

But that’s the boring part.

The interesting part is what happens to decision-making when AI gives you perfect information. Because historically, project management has been about making the best decisions you can with incomplete, messy, often contradictory data.

Now imagine you actually know that the project is 73% likely to miss its deadline. That Developer A is the bottleneck. That the client’s “must-have” feature is used by 4% of users.

AI gives you the data. But you still have to walk into that room, look the VP in the eye, and say: “We need to cut scope or push the deadline. Pick one.”

No algorithm is doing that for you.

Not because it can’t generate the recommendation. But because nobody wants to hear bad news from a chatbot. They want to hear it from someone who understands the politics, the stakes, and the fact that the VP’s bonus is tied to this launch date.

That’s management. The real kind.

The Tools Are Insane Now (And Getting Insaner)

Okay, let me give credit where it’s due. The AI tools hitting the PM space right now are genuinely impressive.

We’re talking about AI that can sit in your meetings, transcribe everything, assign action items, and follow up with people who haven’t done their tasks. Automatically. Without the awkward “hey, just circling back on this” email that we all pretend isn’t passive-aggressive.

There’s AI that reads your project history and predicts which phases will go over budget before they start. Based on patterns from your actual past projects, not some generic template.

There’s AI that generates risk assessments by scanning Slack conversations for red flags. Like when your lead developer starts saying “it should be fine” a lot. (Spoiler: it’s never fine.)

This is genuinely useful stuff. And if you’re a PM who isn’t using these tools, you’re basically choosing to do arithmetic by hand while a calculator sits on your desk.

Stop being that person.

But Here’s Where It Gets Weird

The more AI handles the operational side of project management, the more the job starts to look like... therapy.

I’m only half joking.

Think about it. When the AI handles scheduling, tracking, reporting, and resource allocation, what’s left? People. Motivation. Conflict resolution. Alignment. Trust-building. Navigating egos. Reading the room.

The future PM is less “schedule master” and more “organizational psychologist who happens to know what a sprint velocity is.”

And that’s actually a more important job. It’s also a much harder one. Because you can learn Jira in a weekend. You cannot learn emotional intelligence in a weekend.

(Lord knows some of your coworkers have been trying for decades.)

The Uncomfortable Question Nobody’s Asking

Here it is: if AI can manage the project, what does the team look like?

Because AI isn’t just changing how projects are managed. It’s changing who’s on the project.

When AI can write code, generate designs, draft copy, and test products, suddenly your 12-person team might become a 4-person team with 8 AI agents. And managing a team of AI agents is a fundamentally different skill than managing humans.

You don’t motivate an AI. You don’t give it feedback in a way that doesn’t crush its self-esteem. You don’t invite it to the team offsite.

(Although honestly, some of my past coworkers had about as much emotional range as a language model, so maybe the transition won’t be that dramatic.)

The PM of 2030 might spend Monday managing humans and Tuesday managing AI workflows. Same project. Completely different skill sets. That’s wild when you think about it.

What You Should Actually Do About All This

Look, I’m not going to give you the standard “upskill or die” LinkedIn advice. That’s lazy and everyone says it.

Instead, here’s my actual take:

Stop protecting your admin work. If AI can do it, let AI do it. Don’t cling to busywork because it makes you feel needed. That’s insecurity disguised as diligence.

Get uncomfortable with ambiguity. The future of PM is less about having the right answer and more about asking the right question when nobody else will. Practice that.

Learn to speak “human” better. Communication, negotiation, influence, empathy — these aren’t soft skills anymore. They’re the entire job. They always were, actually. We just dressed it up in Gantt charts to feel more legitimate.

Understand AI deeply enough to direct it. You don’t need to code. But you need to know what AI can and can’t do so you can orchestrate it properly. The PM becomes the conductor. The AI plays the instruments.

(Terrible metaphor? Maybe. But you get it.)

The Bottom Line

The future of project management in the AI age is not about AI replacing project managers.

It’s about AI stripping away all the stuff that was never really “management” in the first place — the tracking, the reporting, the scheduling — and leaving behind the raw, messy, deeply human core of the job.

Leadership. Judgment. Empathy. Politics. Persuasion.

The stuff you can’t automate because it requires understanding why Dave is being passive-aggressive, why the client really wants that feature, and why the CEO’s “quick question” is never actually quick.

AI makes the easy parts of project management obsolete.

It makes the hard parts indispensable.

And if that scares you — good. It means you’re paying attention.

Now go update your Jira board one last time. For old times’ sake ;)

Space technology can contribute to climate change solutions in several ways:

Monitoring and Research: The utilization of space technology in the context of climate change entails a sophisticated approach to observation and study. Satellites, equipped with advanced instruments, play a crucial role in continuously monitoring alterations within the Earth’s atmosphere and climate patterns. These observations are not just periodic but occur in real-time, providing an unparalleled level of precision and immediacy.

The data gathered from space-based instruments becomes a robust foundation for scientific inquiry. By analyzing this information, researchers can discern intricate details about the evolving climate scenario, including variations in temperature, atmospheric composition, and weather patterns. This comprehensive dataset serves as a vital tool for gaining insights into the intricacies of climate change, allowing scientists to quantify its impact and identify potential trends.

Importantly, the real-time nature of the data stream enables swift responses to emergent environmental changes. Timely information is crucial for understanding the evolving nature of climate-related events, such as extreme weather conditions or shifts in ecological patterns.

The implications of space-derived data extend beyond the realms of research. Policymakers can draw upon this wealth of information to formulate evidence-based decisions aimed at mitigating the effects of climate change. The strategic use of space technology in monitoring and research thus not only enhances our understanding of the environmental challenges we face but also empowers us to proactively address them through informed and effective measures.

Carbon Sequestration: Space technology can also play a role in carbon sequestration, which involves capturing and storing carbon dioxide, a major greenhouse gas, in the Earth’s atmosphere. This can help to mitigate the effects of climate change by reducing the amount of greenhouse gases in the atmosphere.

Space-Based Solar Power: Space-based solar power is another promising technology that can contribute to climate change solutions. By generating electricity in space and then transmitting it to the Earth, space-based solar power can significantly reduce the amount of greenhouse gases produced by traditional power generation methods. This can help to reduce carbon emissions and mitigate climate change.

The concept of space-based solar power holds the potential for significant advancements in addressing climate change. As technology progresses, innovations in materials science and automation may make the construction and maintenance of solar power stations in space more feasible. This could lead to the development of global energy transmission networks, enabling efficient transmission of solar energy from space to Earth. The successful implementation of space-based solar power has the potential to reduce the world’s reliance on fossil fuels, contributing to a substantial decrease in greenhouse gas emissions. This shift could enhance global energy security by providing a constant and reliable source of clean and renewable energy, unaffected by weather conditions.

The development of space-based solar power may require international collaboration, fostering collective efforts to address climate change on a global scale. Additionally, this technological advancement could create new economic opportunities, driving growth in industries related to space exploration, satellite manufacturing, and energy transmission technologies.

In essence, the extrapolation of space-based solar power suggests a future where cutting-edge technologies and international collaboration contribute to a more sustainable and resilient global energy landscape.

Space can also contribute to climate change solutions through asteroid mining. Mining asteroids for resources like water, minerals, and other valuable materials can help to reduce the demand for Earth resources, which can in turn help to mitigate the impacts of climate change.

Space technology can help to monitor space weather, which can have significant impacts on Earth’s climate. By monitoring space weather, scientists can predict and prepare for events like solar flares and geomagnetic storms, which can have severe impacts on Earth’s climate.

In conclusion, space technology holds great potential for addressing climate change. From monitoring and research to carbon sequestration and space-based solar power, these technologies can play a crucial role in mitigating the effects of climate change and helping to create a sustainable future.

References:

Climate Change and Space Weather
Space weather refers to changes in the space environment and the effects that those changes have on mankind's…www.sws.bom.gov.au

Space Weather Impacts
24-Hour Observed Maximums Current Space Weather Conditions HF Radio: Weak or minor degradation of HF radio…www.swpc.noaa.gov

The idea for placing the space stations or the satellites into the orbit has been around since long time and main figure behind this specific idea was nobody else than worldly known Science fiction author- Arthur C. Clark.

Arthur C. Clarke

Clarke followed up on this private paper with an article published in October 1945 in Wireless World titled, “Extra-Terrestrial Relays: Can Rocket Stations Give World-wide Radio Coverage?” The paper discusses how rocket technology, such as that used in German V-2s during the war, could be turned to peaceful ends by launching artificial satellites into orbit. All you needed, Clarke argued, was a rocket capable of pushing a payload past an orbital-insertion velocity of 8 km/second [5 miles/second].

Because of basic orbital mechanics, the farther out you could get a satellite, the slower its orbit around the Earth would be. At one point, about 42,000 km [about 26,100 miles] from the center of the Earth, the satellite’s orbit would be exactly 24 hours, the same as the Earth’s rotation. Clarke wrote, in Wireless World:

Evaluation of the “more stable” orbit lends to the idea of “A body in such an orbit, if its plane coincided with that of the earth’s equator, would revolve with the earth and would thus be stationary above the same spot on the planet. It would remain fixed in the sky of a whole hemisphere and unlike all other heavenly bodies would neither rise nor set.

And this made the so-called Geostationary orbit satellite truth to come into reality.

According to Arthur C. Clarke launching 3 satellites into such orbit could have a power to cover the whole Earth with the footprint.

17 years after Clarke’s notion the first telecommunication satellite Telstar launched.

In 1965, Intelsat began launching the first satellite system based on geostationary satellites, and there are more than 300 such satellites in Clarke orbits today. The future of communications evolved much as Clarke had foreseen it.

In order to perfectly match Earth’s rotation, the speed of GEO satellites should be about 3 km per second at an altitude of approx 36000 km.

GEO is used by satellites that need to stay constantly above one particular place over Earth, such as telecommunication satellites.

credit: Intelsat

This way, an antenna on Earth can be fixed to always stay pointed towards that satellite without moving. It can also be used by weather monitoring satellites, because they can continually observe specific areas to see how weather trends emerge.

However, launching the satellites into GEO orbit was the first part of the goal.

To maintain the orbital location of those satellites and broadband, it needed additional activities to be completed in a daily basis which we will call Spacecraft Operations from now on.

I will try to talk about the operations aspects of satellites in my following entries.