Q: Did cancel culture accidentally revolutionise the defence industry?

@anduriltech has a wild origin story - 

From a homeschooled anime fan living in a trailer to a $30 billion defence powerhouse reshaping American warfare.

All kicked off by public shaming?

Let's dig in. ⬇️

This story will cover: 
1. A short history of the US defence industry 
2. Palmer Luckey's origin story
3. The principles Andruil used to grow a $31 bn valuation  
4. Andruil's robotics portfolio

It's an open secret that the US defence industry has been decaying for decades.

The problem started in 1961 when Defence Secretary McNamara imported central planning into the Pentagon. 

The system was unintentionally modelled on communist economic planning: 
- Centralised requirements setting, 
- Multi-year resource allocation, 
-  Bureaucratic approval processes that prioritised control over innovation.

As historian Charles Ries wrote, military planning became "almost socialist in its metaphysics."

Ironic: They were using Soviet methods to fight Soviet adversaries.

1993's "Last Supper" made it worse. 

Defence Secretary Perry forced consolidation from dozens of contractors down to the Big 5: Lockheed, Boeing, Raytheon, Northrop, and General Dynamics. 

They now control 80% of defence revenue with zero incentive to innovate.

The F-35 Joint Strike Fighter tells the story:
• $1.7 trillion program
• 8 years over schedule  
• Still doesn't work properly

Under cost-plus contracts, Lockheed literally makes MORE money by being late and over-budget.

Meanwhile, Silicon Valley completely abandoned defence after Vietnam.

By 2018, 4,000+ Google employees signed a letter: "Google should not be in the business of war." Google cancelled Project Maven (helping the US military) 

Silicon Valley was built on defence but now, America's technologists were refusing to defend America.

Enter Palmer Freeman Luckey, born 1992 in Long Beach, California.

Homeschooled kid obsessed with Yu-Gi-Oh's Seto Kaiba - an orphaned genius who inherits a weapons manufacturing company and builds VR games. 

"I'm quite concerned that I'm doing what I was programmed to do when I was 8 years old," Palmer later reflected.

Age 17: His parents kick him out. He's living in a gutted 19-foot camper trailer in their driveway.

No bathroom (uses the laundromat). 

No kitchen (Jack in the Box loyalty card). 

Twin mattress on boxes, six-screen computer setup, and a soldering iron. $300 in his bank account.

His life's goal: building VR 

And to be fair to him, he achieves this goal. 

Launches Oculus, Kickstarter instead: raises $2.4 million from a $250,000 goal.

The GOAT, @ID_AA_Carmack - creator of Doom  - discovers Palmer through forum posts and becomes CTO. 

E3 2012 demo: "Probably the best VR demo the world has ever seen." 

March 2014: Zuckerberg acquires Oculus for $2.7 billion.

Palmer becomes the poster child for Silicon Valley innovation.

Then everything changed.

September 2016. Palmer, 24, is on a cruise ship with his girlfriend when his phone explodes.

The Daily Beast publishes "The Facebook Billionaire Secretly Funding Trump's Meme Machine."

The story?

Palmer donated $10,000 to "Nimble America" - a group that put up exactly one billboard saying Hillary was "Too Big to Jail."

Palmer claimed the conversation was off-the-record. The Daily Beast disagreed.

The donation sparked immediate backlash in Silicon Valley's predominantly liberal tech community.

Facebook faced internal pressure as the story spread. The Silicon Valley mob had spoken. Facebook panicked.

 Zuckerberg personally drafted a statement Palmer had to post - disavowing Trump, expressing "regret."

Palmer was ordered to stop posting on social media, stop messaging colleagues, stop coming to the office.

February 2017: Palmer gets the call while at Disney World. Facebook terminates him anyway.

He offers to forfeit ALL his remaining money just to come back.

Answer: No.

Fast-forward to 2024:  Zuckerberg and other tech leaders openly endorse Trump with minimal professional consequences. 

This termination became Palmer's catalyst for a new venture. 

 Remember, Silicon Valley had turned against defence? 

This meant VC's wouldn't fund you, engineers would join you & you'd become a social pariah. 

Luckily for Luckey, he was already a paraia. 

It's easier to go against the grain when you have nothing to lose.  

June 2017. Palmer starts Anduril Industries with former Palantir executives.

The name means "flame of the west" - Aragorn's sword in Lord of the Rings. 

If you're aware of the Thiel-verse this reference won't be a suprise.

@PalmerLuckey created Anduril with 5 principles that flipped the defense industry:

This wasn't just a new company. It was a completely different operating system for defense innovation.

At the centre of everything is Lattice AI - the nervous system of autonomous warfare.

Think of it as the ChatGPT of battlefields. 

It ingests data from thousands of sensors - drones, cameras, radars, satellites. In real-time, it identifies threats, predicts behaviours, and coordinates responses. Human operators see a God's-eye view of the battlefield with AI recommendations.

The system can autonomously launch counter-drones, coordinate swarms, or flag targets for human decision. 

It compresses the "kill chain" from minutes to seconds: Observe, Orient, Decide, Act.

This is Palmer's version of Laplace's demon - the 18th-century thought experiment about a superintelligent being that could predict the future by observing everything.

Palmer literally programmed his childhood anime fantasies into military reality.

**Principle #3 in action** - Software-First Over Hardware-Centric:

The technical arsenal that connects through Lattice:

**Ghost Drones**: 3-minute assembly, 60-minute silent flight. Navigate without GPS through AI. Coordinate in swarms.

**Anvil Interceptors**: Autonomous drone-hunters that detect and destroy enemy UAVs through kinetic impact. No explosives needed.

**Sentry Towers**: AI-powered surveillance systems that autonomously monitor borders. Originally deployed on US-Mexico border.

**Dive-LD**: Autonomous submarines that operate for months without human input. Australia ordered them for $23 million each vs $18 billion for nuclear subs.

**Fury**: Autonomous fighter jet that won the Air Force's Collaborative Combat Aircraft program, beating Boeing, Lockheed, and Raytheon.

Small, cheap, autonomous systems that swarm vs large, expensive, human-operated platforms.

2017-2021 was lonely for Palmer. 

"I was one of the only people who was already lit on fire," he said. "I could do literally anything and it couldn't get worse."

That ruined reputation became an asset - he could work on weapons when others couldn't. 

While Google employees were protesting military contracts, Palmer was hiring their best people. Being contrarian does have its advnatages 

Top engineers from Apple, Google, and Palantir came to build the future of warfare.


**Principle #5 in action** - Acquisition for Speed:

Instead of building everything in-house, Palmer went shopping.

2021-2023: 5 major acquisitions across all domains, each accelerating capabilities by 7-10 years:

**Area-I**: Air-launched drones (Altius)
**Copious Imaging**: 360-degree thermal sensors (WISP) 
**Dive Technologies**: Underwater vehicles
**Adranos Energetics**: Solid rocket motors
**Blue Force**: Fighter-scale autonomous aircraft (Fury)

The Dive deal shows the playbook perfectly: Small Boston startup with great underwater tech but no military relationships. Anduril bought them, connected to Lattice, won $100 million Australian contract within 3 months.

"Acquire and integrate" - buy the best tech, plug into Anduril's trust network.

The first major validation: Ukraine.

Small, cheap, autonomous systems were destroying Russian tanks and ships. Everything Palmer predicted was happening in real-time on CNN.

By 2024, the entire Silicon Valley establishment was changing tune. 

July 2024: Elon Musk, Bill Ackman, Marc Andreessen all endorsed Trump and Defence was no longer a conversation killer.

Palmer's reaction: "It kind of pisses me off... you guys are really a piece of work to be so mean to me about it and then all of a sudden, you all come to the conclusion I came to... 8 years ago."

Say what you like about Palmer's politics - at least he was consistent. 

Palmer's success opened the floodgates. Suddenly, venture capital poured into defense:

**Shield AI**: Former Navy SEALs building AI fighter pilots
**Epirus**: Microwave-based counter-drone systems  
**Saronic**: Autonomous surface vessels
**Red 6**: AR/VR pilot training systems

Y Combinator and Andreessen Horowitz started actively recruiting defense startups. 

Palmer had made weapons cool in Silicon Valley again. Though the mainstream VC's had to give it a rebrand: "American Dynamism". 

**Principle #2 in action** - Fixed Price Over Cost-Plus:

The valuation explosion tells the story:

2021: $4.6B (Series D)
2022: $8.5B (Series E) 
2024: $14B (Series F)
2025: $30.5B (Series G)

From cancelled tech worker to defense prime in 8 years. Revenue hit $1 billion in 2024, growing 138% year-over-year.

The magic: 40-45% margins vs 8-10% for traditional defense contractors.

Anduril sells finished products at fixed prices vs cost-plus contracts where delays increase profit. 

Everyone benifits from efficiency, not dysfunction.

**Principle #4 in action** - Private R&D:

2024: Anduril announces Arsenal-1.

A $1 billion, 5-million-square-foot autonomous weapons factory in Ohio. 4,000+ jobs. 

Production starts July 2026—Tesla-style assembly lines for missiles and drones.

This is private capital building industrial capacity, not waiting for government funding. Venture-backed manufacturing sovereignty.

"No one asks me about discount rates or 15x EBITDA multiples" - Palmer can invest for the long term while public primes optimize for quarterly earnings.

2025: The ultimate vindication.

Meta and Anduril announce partnership for AR/VR battlefield systems. Palmer's Oculus heritage meets his Lattice AI platform. Consumer-friendly interfaces for military applications.

The irony: The same company that fired him is now his partner. 

The technology is spreading worldwide through all 5 principles:

- UK: $38 million for Altius drones supporting Ukraine (Speed)

- Australia: Ghost Shark submarines, local manufacturing (Fixed Price + Private R&D)

- Germany: Rheinmetall joint ventures (Acquisition for Speed)

"Buy European" programs while getting American AI technology. 

Meanwhile, top engineers are fleeing Google and Apple for defense startups. Y Combinator and Andreessen Horowitz are actively recruiting weapons companies.

The Anduril Framework is being exported globally.

Palmer's success raises real questions:

- Should autonomous weapons make decisions without human oversight?
- Can we trust AI black boxes in life-death situations?
- Should private companies control military technology?
- Does removing human cost lower the threshold for conflict?
- Are we trading necessary deliberation for dangerous speed?

These tradeoffs matter when technology decides who lives and dies. 

The irony: In trying to cancel one man, they accidentally catalysed a defence renaissance. Probs not what they were going for...

The Anduril Framework - Speed, Fixed Price, Software-First, Private R&D, Acquisition for Speed - is now being copied across the industry. 

Traditional primes are scrambling to adapt. New startups are raising billions.

Sometimes the best revenge isn't living well. Sometimes the best revenge is building autonomous weapons that reshape global power.

What's your take? 

Did Palmer's cancellation accidentally save American defence? 

Or are we sleepwalking into an autonomous weapons nightmare?

Revolutionizing Defense: Anduril's Rise

Most robotics startups are building variations of the same bot. 

@clonerobotics has gone completely rogue

No motors and gears. Clone Robotics have thrown out that playbook and built something that moves like us - with artificial muscles, bones, and even a beating heart.

Let's dive into the tech behind this iconic humanoid ⬇️

The tech traces back to a touching origin story.

Joseph McKibben, a Manhattan Project physicist, invented pneumatic artificial muscles in the 1950s - not for robotics, but to help his daughter regain use of her hands after losing motor function.

The "McKibben muscle" used simple rubber tubes and air pressure to mimic biological muscle contraction. Clone's breakthrough was switching from air to hydraulics for real power.

Technical specs that matter:
• 200+ degrees of freedom (vs 30-60 in motor-based bots)
• 1,000+ individual muscles
• 500+ sensors across the body
• Hydraulic pump as the "heart"
• Complete skeletal system with anatomically correct attachment point

But here's the reality check:

Protoclone still hangs from ceiling cables. Walking untethered remains their biggest engineering challenge. They're essentially trying to fit an entire hydraulic power plant inside a human-sized torso.

They're making rapid progress but  what makes them interesting to me isn't the tech specs - it's the strategic positioning.

Tesla, Figure, Boston Dynamics - they're all building variations of the same motor-driven architecture. 

Roughly 80% overlap in fundamental approach.

Clone went completely counter-positioned. 

High risk, high reward.

My take: If Clone succeeds, they win alone.

If motor-based approaches dominate, there are 10s of companies fighting for the same market with similar tech.

It's either a brilliant contrarian bet or an expensive engineering dead-end. 

But in a field where everyone is converging on similar solutions, sometimes the biggest risk is playing it safe.

What do you think? Is biological mimicry worth the engineering complexity, or should robotics stick with proven actuator technology?

P.s follow @AnastasiInTech I used her video for a lot of the facts


Innovative Robot Design

Thread: How Amazon's $775M Acquisition Accidentally Launched an Entire Industry

Here's a wild story about unintended consequences: 

Amazon acquired a robotics company and nearly tanked the industry by removing it from the market. 

This seeming disaster helped catalyse a $40 billion market.

Here's the story of how the Kiva acquisition sparked one of the most dynamic markets in robotics. ⬇️

The World Before Orange Robots

In 2003, while Boston Dynamics was building walking robots for DARPA and Honda was perfecting ASIMO, warehouses were stuck in the stone age.

Human workers pushed carts through endless aisles, walking 12-15 miles per shift. They were essentially expensive GPS systems with legs—locating items, walking to them, then walking to pack stations.

The economics were brutal. 

At failed grocery startup Webvan, 70% of fulfilment costs came from human labour. And most of that labour was just... walking.

The Eureka Moment

Mick Mountz was working at Webvan when he crunched the numbers. The waste was staggering—paying people premium wages to spend most of their day trekking through warehouse mazes.

His insight was deceptively simple: What if instead of people going to products, products came to people?

It sounds obvious now. Back then, it was revolutionary.

The Dream Team

In 2003, Mountz assembled a powerhouse trio:
- Peter Wurman: AI expert from NC State who understood the software complexity

- Raffaello D'Andrea: Cornell robotics genius and robot soccer world champion who knew how to make things move

Together, they founded Kiva Systems with a radical vision: orange robots that could turn any warehouse into a choreographed dance floor.

The Technical Breakthrough

While competitors built expensive systems requiring infrastructure overhauls, Kiva's approach was elegantly simple:

- Orange, Roomba-like robots that could slide under inventory pods

- Simple barcode stickers on the floor for navigation—no expensive wiring,  magnetic strips or complex SLAM 

- Corkscrew lift mechanism that could hoist 1,000-3,000-pound shelves

- Swarm intelligence that allowed hundreds of robots to work in harmony

Result: 2-4x productivity improvement over traditional picking methods.

To the Moon 

By 2011, Kiva was growing 100% annually.

Systems cost $1-2 million for small operations, $15-20 million for large warehouses with 1,000+ robots. 

Their customer roster read like a retail hall of fame: Staples, Walgreens, The Gap, Office Depot, Crate & Barrel, Saks Fifth Avenue. 

The technology worked so well that it was almost boring—which is exactly what enterprise customers wanted.

Amazon Takes Notice

Here's where it gets interesting: Amazon was already a Kiva customer.

When Amazon acquired Diapers.com in 2010, it inherited that company's Kiva robot fleet. Amazon's logistics team was blown away by how robust and reliable the system was.

Mountz had been visiting Amazon annually to show their progress. 

The relationship was so good that when Amazon approached Kiva in 2011 about expanding their partnership...

The Offer You Can't Refuse

Amazon's message was direct: 

"We think we should do a project together, but we would need so many of your robots; it only makes sense if we own the whole thing."

March 19, 2012: Amazon announces the $775 million acquisition—their second-largest deal ever at the time and one of the biggest deals in robotics history. 

Today, this wouldn't even buy you a single AI researcher, but back then, it was massive.

The Betrayal 🗡️

Initially, Amazon promised to keep serving Kiva customers. That lasted about three years.

April 2015: Amazon rebrands Kiva as "Amazon Robotics" and stops all external sales.

August 2015: The transformation is complete. Amazon's suggestion to stranded customers? 

"Let Amazon Services handle your fulfilment using our robots in our warehouses."

The Immediate Carnage

The impact was devastating:

Existing customers were stranded — Companies like Staples and Walgreens had built entire operations around Kiva robots, only to watch their investment disappear overnight.

Years of integration work, employee training, and optimisation meant nothing when support ended.

At this point, Kiva employees started leaving to look for their next adventure.

The Accidental Revolution

This gap created instant demand, and with demand comes opportunity. 

Customers had urgent needs, investors saw market potential, and a whole cohort of experienced robotics experts had just become free to operate.

Let's focus on two key examples:

6 River Systems

Former Kiva executives Jerome Dubois and Rylan Hamilton founded 6 River Systems in 2015, creating collaborative robots dubbed "Chuck" that worked alongside humans instead of replacing them. 

Their approach was more affordable and required less infrastructure than Kiva's full automation. 

Locus Robotics

Bruce Welty of Quiet Logistics relied heavily on Kiva until Amazon's acquisition left him without a solution. "I joked, 'We'd be screwed if Amazon bought them,' but never believed it would happen."

When it did, Welty took action and founded Locus Robotics in 2014. The company developed robots that navigated warehouses without layout changes, unlike Kiva's shelf-moving method. LocusBots collaborated with human workers to optimize picking and packing.

Using advanced vision and machine learning, LocusBots adapted to fast-changing e-commerce demands. They quickly gained major clients like DHL, FedEx, and Walmart. 

The Startup Explosion

VCs noticed the opportunity Amazon had created. Robotics investment exploded & what followed was an entrepreneurial gold rush. Companies emerged to fill the Kiva void:
- Locus Robotics (2014) — Now valued at $2 billion
- 6 River Systems (2015) — Sold to Shopify for $450M, then to Ocado for just $12.7M
- Fetch Robotics (2014) — Collaborative picking robots, sold to Zebra
- Geek+ (2015) — China's answer to Kiva
- inVia Robotics (2015) — Goods-to-person systems

Fast forward to today: The warehouse robotics market is worth $40+ billion and growing 10-15% annually.

The Trust Problem

Although the Kiva acquisition had a net positive impact, it created lasting industry trauma. 

Established companies became terrified of partnering with robotics startups.

The fear was rational: Why invest time and resources in a partnership when your technology provider could get acquired and disappear overnight?

This "Kiva trauma" fundamentally changed how retailers approached robotics partnerships, favouring larger, more established vendors over innovative startups. 

This trust deficit remains something I have to work through in many deals today.

Amazon's Bet Paid Off

For Amazon, the acquisition was transformational:

- Click-to-ship time: 60-75 minutes → 15 minutes

- Operating costs: 20% reduction

- Space efficiency: 40% more inventory in the same space

- Total savings: Estimated $2.5 billion

- Robot fleet: 200 → 750,000+ robots today
 
They had created an insurmountable logistics moat.

The Unintended Revolution

As Peter Wurman, Kiva's co-founder, later reflected: 

"If we hadn't sold Kiva to Amazon, the number of robots deployed wouldn't have grown from 200 to 600,000 or 1 million. We helped create a category."

Industry observer Keith Shaw put it more bluntly:

 "The entire autonomous robot sector was formed because of that purchase."

Conclusion: The $775 Million Catalyst

Amazon's acquisition of Kiva ranks among the most pivotal in robotics, not just for its internal success, but for the opportunities it sparked.

Material handling has become one of the great successes in robotics

By absorbing the market leader, Amazon unleashed a flurry of innovation. It left behind a demand-driven market with eager investors and talent, igniting a shift from a single player to a $40+ billion competitive ecosystem.

Kiva's acquisition not only transformed Amazon but also redefined logistics automation—robots continue to thrive.

Amazon's $775M Acquisition Sparked Industry

☕ Coffee extraction: a $100B optimisation problem disguised as a 30-second routine.

Every shot balances 6 variables and  1000+ molecules. One parameter off? Expensive disappointment.

You'd think this has nothing to do with robotics but you would be wrong, lets dig in ⬇️

 What Is Coffee?

To understand why this has anything to do with robotics, we need to start with the basics. 

The raw material: Coffee beans are seeds from cherry-like fruits, packed with:

- Caffeine: Natural pesticide (1-2% by weight) that gives us our energy boost

- Chlorogenic acids: Antioxidants that create brightness and acidity

- Sugars: Locked in complex carbohydrates (30-35% of bean mass)

- Oils: Carry flavour compounds and create mouthfeel

- Proteins & fibre: The structural matrix holding everything together

Why roasting matters:

Raw coffee beans are essentially inedible wood. Roasting is controlled pyrolysis:

- Heat breaks down cell walls → releases trapped compounds

- Maillard reactions → create 800+ new flavor molecules

- Caramelisation → converts starches to sugars

- CO2 formation → creates the porous structure that allows extraction

The result: A brittle, porous matrix containing the concentrated essence of the original fruit - ready for grinding & extraction.

The Chemistry Challenge: Selective Dissolution

Coffee extraction is fundamentally a selective dissolution problem, where only the desirable compounds are dissolved while the undesirable ones are left behind.

The extraction physics:

- Good compounds (acids, sugars, oils) dissolve first with heat and pressure
- Bad compounds (tannins, bitter alkaloids) require longer contact time

The challenge: Stop extraction before the bitter stuff kicks in to get the perfect flavour. 

The Extraction Sweet Spot

The magic zone: 18-22% of the coffee bean's mass dissolved into water.

Under-extraction (< 18%): Only acids extracted, sugars still locked in

Tastes: Sour, sharp, thin

The tell: Makes your mouth water (like biting a lemon)

Over-extraction (> 22%): Everything good PLUS harsh compounds

Tastes: Bitter, dry, astringent

The tell: Makes your mouth dry up (like overbrewed tea)

The sweet spot: Maximum good compounds before bad compounds dominate.

The Engineering: How Espresso Machines Control Variables

An espresso machine is a precision fluid control system managing 6 variables simultaneously:

- Water temperature: 90-96°C (extraction rate)
- Brew pressure: 9 bars (flow dynamics)
- Dose: Coffee mass (typically 18g)
- Grind size: Particle surface area (THE key variable)
- Distribution & tamping: Bed uniformity and resistance
- Yield: Target liquid output (determines ratio)

How The Machine Works

Temperature control:
The machine heats water using either a boiler, heat exchanger, or a thermoblock system. PID controllers maintain temperature within ±1°C using real-time sensors.

Pressure generation:
Internal pumps (rotary or vibratory) pressurise water to exactly 9 bars. Safety valves prevent over-pressurisation while flow restrictors maintain consistent pressure.

The group head:
This is where precision matters most. A three-way solenoid valve controls water flow, while a dispersion screen with 100+ tiny holes distributes pressure evenly across the coffee bed.

The portafilter:
Acts as both a filter and a reaction chamber. The basket has precisely sized holes (0.3-0.5mm) that create the right flow resistance. Heavy metal construction maintains temperature stability.

The Extraction Process
Pre-infusion: Low pressure (2-4 bars) wets the coffee bed evenly
Full extraction: 9-bar pressure forces water through coffee particles
Flow monitoring: Sensors track volume and timing
Automatic shutoff: When target yield is reached

Real-time feedback through pressure gauges, flow meters, and timers lets you monitor every variable.

Every shot is a controlled chemistry experiment in 25-30 seconds.

The Automation Paradox

You could automate every single variable. Push a button, get espresso.

But here's the problem: Fully automated machines consistently produce mediocre coffee.

Why? Because perfect extraction requires constant adaptation. Beans change daily as they age. Humidity affects grind consistency. Equipment drifts over time. Customer preferences vary.

Machines follow predetermined protocols. They can't taste the difference between sour under-extraction and balanced sweetness.

The breakthrough insight: The best coffee comes from strategic human-machine collaboration.

Controlling the Chaos: The Master Barista Approach

Elite baristas use a constraint-based optimisation strategy that turns 6 variables into a manageable system:

Step 1: Let machines handle precision
- Modern espresso machines excel at temperature and pressure consistency. Set them once, trust them completely.

Step 2: Standardise the controllables

- Fix your dose (18g becomes your constant)
- Perfect your tamping technique (consistent pressure, level surface)
- Nail distribution (even coffee bed every time)

Step 3: Choose your target ratio
- Pick your desired coffee-to-liquid yield. Common target: 1:2 ratio (18g - coffee → 36g liquid)

Step 4: Optimise through grind size
- Now you have ONE variable to adjust: grind fineness
- Start medium, pull shot, stop at target yield (36g)
- Taste: Sour = grind finer | Bitter = grind coarser
- Repeat until balanced
- Lock in the grind setting

Step 5: Use time as your consistency check
- Perfect shots take 25-30 seconds. 
- If tomorrow's shot runs 35 seconds with the same settings, something changed (bean ageing, grinder drift, humidity). 
- Time becomes your early warning system.

Why Humans Still Win

The gap: Machines are precise but not adaptive. Current espresso machines only have basic sensors for temperature and pressure - they're essentially blind to extraction quality.

The magic happens in what humans sense that machines can't (yet).

Multi-sensory feedback is required:
- Visual: Flow color, crema formation, stream consistency
- Sound: Extraction rhythm, grinder motor load changes
- Smell: Aroma shifts indicating extraction phases
- Taste: Complex flavor compound detection and memory

Rule-breaking intelligence:

A skilled barista knows when to break the standard rules. 

Want a ristretto? They'll decrease the ratio to 1:1.5 without causing under-extraction by adjusting grind size and timing simultaneously. 

Want a lighter roast to shine? They'll drop the temperature by 2°C and extend the extraction time.

Machines follow protocols. Humans adapt in real-time based on sensory feedback and experience.

That's why the best coffee still comes from expert human-machine collaboration, not full automation.

The Broader Lesson

Perfect espresso demonstrates that complex optimisation problems are sometimes best solved through strategic human-machine collaboration, not full automation.

When machines do have the right sensors, they dont have the right data. 

The winning formula becomes: 

Machine precision + human adaptability + constraint-based design.

In robotics: 

Don't automate everything. Design systems where humans and machines each handle what they do best.

Next time you drink great coffee, remember: You're experiencing a masterclass in human-machine collaboration.

Coffee Extraction and Robotics

Thread: How China bought Germany's robotics crown jewel 

Quick story: I recently called Kuka German in a thread and got called out - "Kuka is Chinese now."

This sent me down a rabbit hole. 

What I found was straight out of Succession, and it reshaped how the West thinks about tech sovereignty.

This is the story of how a Chinese appliance company bought Kuka & kick-started China's robot dominance. ⬇️

The Setup:

In May 2015, China announced "Made in China 2025" - an ambitious 10-year industrial strategy targeting robotics as a key sector for technological dominance.

At the time, the global robotics landscape was dominated by a handful of established powers.

 - Japan led with Fanuc and Yaskawa
 
- Germany & Switzerland represented Europe with KUKA & ABB

China? Largely absent from the top tier. They were consumers of robotics technology, not creators.


The Hunter:

Midea Group wasn't an obvious candidate to acquire Germany's robotics crown jewel. 

Founded in 1968 as a bottle cap producer, the company had evolved into the world's largest air conditioner manufacturer, with $21 billion in revenue by 2015.

But CEO Paul Fang saw beyond appliances. China's home market was maturing, and Midea needed its next act. 

Fang identified industrial robotics as the strategic opportunity that could transform Midea into an automation powerhouse.

The Target:

KUKA represented everything Midea needed: 118 years of German engineering excellence and one of the top 4 robotics companies globally.

The Augsburg-based company employed 13,000 people across 25 countries and held a commanding position in automotive robotics - exactly the expertise China's Made in China 2025 strategy required.

The Approach:

Fang's first meeting with KUKA executives revealed the challenge ahead. 

Only a sales manager was there to greet the Chinese delegation - clearly not taking them seriously as potential partners, let alone acquirers.

Instead of protesting, Fang began executing a methodical acquisition campaign.

The 3 Phase Campaign:

1️⃣ Phase 1: Silent Accumulation (2015):
Midea quietly accumulated KUKA shares, staying below the 5% disclosure threshold. 

Once they crossed to 5.4%, disclosure requirements kicked in and everything changed. 

"As soon as it was out, their CEO came knocking, financial statements in hand, ready to negotiate," Fang later recalled.

2️⃣ Phase 2: Revealing Intent (Early 2016):
Accumulation continued, reaching 10.2% by early 2016, making Midea KUKA's second-largest shareholder. 

KUKA went from dismissing Midea to desperately needing them. 

3️⃣ Phase 3: The Shock Offer (May 2016):
The bid - €4.5 billion total, €115 per share, representing a massive 60% premium over February 2016 stock prices. 

This was an offer to good to refuse. 

The Kill:

When German Economy Minister Sigmar Gabriel searched for European "white knights" - courting Siemens, ABB, and others to counter-bid - none were willing or able to match Midea's offer.

European companies simply didn't believe KUKA was worth that much.

The Regulatory Reality:

Germany tried to throw up barriers but discovered a fundamental gap: they lacked the legal apparatus to block the acquisition. 

They couldn't even conduct formal inquiry unless it touched sensitive sectors like telecommunications or utilities.

KUKA's industrial robotics, however advanced, didn't fall under existing security frameworks. Ouch. 

Strategic Concessions:

Midea sidestepped remaining concerns with targeted concessions:
- Signed agreements restricting corporate reorganisation for 7.5 years
- Carved out US aeronautics subsidiary to satisfy American CFIUS requirements
- Secured management and supervisory board support

These moves neutralised regulatory pushback while preserving the strategic value of the acquisition.

By January 2017, Midea controlled 94.55% of KUKA.

Too little too late:

The acquisition transformed how Europe viewed outside investment.

Immediate responses:
- Investment review threshold lowered from 25% to 10-20% (making Midea's stealth accumulation strategy impossible - future acquirers would face government scrutiny before building momentum)
- Extended screening periods
- Broader definition of "strategic" sectors

The Strategic Payoff:

The acquisition wasn't just about Kuka - it was about transplanting decades of German expertise.

Made in China had staggering results. China's 2024 vs 2015 Transformation:

- Robot density: 49 → 470 units per 10,000 workers (+861%) 
- Global ranking: 25th → 3rd most automated country, only behind Singapore and South Korea
- Domestic market share: 28% → 47% —from dependency to dominance
- Global installations: 30% → 51% of worldwide total. More than Europe and the Americas combined
- Annual robot production: ~69k → 429k units (+523%). 6x growth in 8 years


Was KUKA solely responsible for China's robotics advancement? No. But it accelerated what might have taken decades into just a few years.

Robotics isn't just another industry - it's the foundational technology of the next economic era. 

When we sell 100+ years of expertise for short-term profits, we're not just selling companies - we're selling our technological future.

The lesson: In an era where technology equals power, some capabilities shouldn't be for sale at any price.

China's Robotics Takeover

What can robotics learn from watch design?

- Hundreds of parts packed into a coin-sized case, 
- keeping time within ±20 seconds daily
- enduring decades of use

Mechanical design at its most optimal.

Let's explore the mechanisms that make watches tick ⬇️

Start with the rotor - a semicircular weight that spins freely inside the watch case.

Every time you move your wrist, this weighted rotor swings like a pendulum, driven by inertia from your natural arm movements.

This is your watch’s power source, converting motion into stored energy.

The rotor’s spin winds the mainspring—a coiled steel strip housed in a barrel.

Once tensioned, it strains to unwind, storing potential energy. 

This is the watch’s battery. 

But this raw power would spin chaotically without control.

The mainspring’s energy flows through the gear train that transforms the mainspring’s slow, powerful rotation into faster, controlled motion.

This regulated energy reaches the escape wheel, ready for precise timekeeping.

The escape wheel receives all this energy and wants to spin freely. But it can't. It's blocked by the pallet fork.

Since all the gears are connected, this blocks the entire gear train - nothing can move until the pallet fork releases a tooth. But what controls this gate? 

Meet the gatekeeper: the balance wheel. This weighted wheel oscillates back and forth, controlled by its hairspring. 

Think of it as the watch's metronome - it sets the rhythm for everything else.

Each swing of the balance wheel unlocks the pallet fork to release exactly one escape wheel tooth. 

Each release lets the entire gear train advance by a precise amount. 

To show time, the gear train connects to the dial train—a separate set of gears (second wheel, minute wheel, hour wheel) under the dial.

These gears convert the gear train’s pulses into the correct speeds for the watch hands: 

The second hand sweeps, the minute hand moves 60 times slower, and the hour hand 12 times slower than that, typically through ratios like 1:60:720.

All this precision requires minimal friction. 

Jewelled bearings (synthetic rubies) at pivot points let steel shafts spin with virtually no wear. 

Those tiny red crystals aren't decoration - they're precision engineering for decades of accuracy.

Those are the basics, but things get interesting with complications. 

In watchmaking, a complication is any function beyond basic hours, minutes, and seconds. 

These mechanical features require additional gears, springs, and levers - often tripling the component count.

Each complication presents unique engineering puzzles. 

I'll cover three of my favourites:

- How do you add a stopwatch without disrupting timekeeping? 
- How do you mechanically encode calendar irregularities? 
- How do you deal with the effects of gravity?


The chronograph tackles a tricky problem: how do you add a stopwatch without disrupting the main timekeeping?

The solution. A separate gear train runs parallel to the main movement, connected only when needed.

Press start: a clutch wheel drops down, meshing the chronograph train with the main gears. The chronograph hand begins sweeping.

Press stop: the clutch lifts away. The main movement continues unaffected while the chronograph hand freezes.

Reset uses a heart-shaped cam. A hammer strikes this cam, snapping the chronograph hand back to zero through pure geometry.

Multiple hands each get their own gear ratio: minute counter advances every 60 seconds, hour counter every 60 minutes.

Two independent systems sharing one energy source, triggered by button presses while keeping perfect time throughout.

The perpetual calendar is the ultimate challenge, mechanically encoding the Gregorian calendar’s quirks.

A program wheel rotates once every four years (1,461 days), its edges shaped to signal each month’s length.

For 30-day months, a lever advances the date twice on the last day.

February’s 28 or 29 days are tracked by a 48-month wheel, with a feeler lever reading its profile. 

Gears with precise ratios sync the day, date, and month, tracking multiple cycles purely mechanically.

The tourbillon tackles gravity’s effect on timekeeping.

When a watch is vertical, gravity pulls unevenly on the balance wheel, causing timing errors.

The solution? Rotate the entire escapement (balance wheel, pallet fork, escape wheel) in a carriage that spins once per minute, averaging out gravitational errors.

The carriage acts as a gear in the gear train, driven by a fixed pinion. 

The balance wheel must be perfectly balanced to avoid centrifugal forces.

In pocket watches, held vertically for hours, tourbillons improved accuracy. 

In wristwatches, which move constantly, they’re more about showcasing mechanical mastery.

This is where a few 100 years of mechanical engineering refinement can get you.

No batteries, no electronics, no external power source. 

Just human motion converted into measured time.

Now the question is, where will our robots be in 100 years?

Timeless Engineering: Watch Mechanisms

The robotics revolution is here - insights from over 1,092 job listings reveal significant industry growth. 

1,092 active robotics positions from 123 companies. 34% are new & recent 

Here's what the data indicates about the industry's future direction


🎯 Hottest Sectors Right Now:

Humanoid Robotics (206+ jobs)
Autonomous Vehicles (130+ jobs)
Warehouse Automation (120+ jobs)
Agricultural Robotics (45+ jobs)


🤖 Humanoid Robotics = still hot 

Tesla: (Optimus): 49 jobs
Figure: 50 jobs
Agility Robotics: 30 jobs
The Humanoid AI: 27 jobs

🌍 Geographic Hotspots Revealed:

Silicon Valley: 312 jobs (28% of market!)
European Tech Hubs: 93 jobs
East Coast (Boston/NYC): 75 jobs

The Valley still dominates, but Europe is emerging as a serious robotics powerhouse 🇪🇺


 🔥 Pro tip for job seekers: 

The keywords "manipulation" and "perception" appear in 40%+ of senior robotics roles. 

Upskill in these areas to ride the wave.

Want access to the latest robotics jobs?

📧 Newsletter subscribers get exclusive access 

Stay ahead of the curve →https://robotiq.substack.com/subscribe

Robotics Revolution Explosion

🤖 We analysed 1,141 robotics jobs across the industry this week. 

The data reveals where the smart money is going.

Here's what is happening in robotics hiring 👇

Software is still the king but hardware is catching up. 

The USA is still dominating roles but Europe is hold its own! 

Get the latest robotics jobs

🌐 Our website updates with new roles every Friday 
📧 Newsletter subscribers get early access on Wednesday

If you need early access but missed the letter  - comment 'access' on this thread

Stay ahead of the curve → https://www.robot-iq.org/

Robotics Hiring Trends

Actuators eat up 30-50% of a humanoid robot's BOM cost. Mass adoption? Not at these prices.

This is one of the most persistent challenges in robotics, but @IMSystemsNL may have cracked the code with a novel drive. 

To understand it, we need to dive into the world of drives & transmission, with all their demands and trade-offs.

Shoutout to @MarwaEldiwiny and @GoingBallistic5 for the original video.

The actuator dilemma that's plagued industrial robotics for 50 years:

Electric motors are most efficient at high speeds (5,000-20,000 RPM). But robot joints need slow, controlled movement (~60 RPM max).

This forces massive gear reductions (50:1 to 320:1) - and that's where everything breaks down.

Why massive gear reductions suck:
- Backlash multiplies: Every tooth adds play - multiply by 50x-320x reduction and your robot can't draw a straight line
- Efficiency drops: Each stage loses 2-5% efficiency
- Weight explodes: More stages = more mass to accelerate
- Back-driveability vanishes: High ratios make manual joint movement impossible - a safety nightmare

The bigger the reduction, the worse these problems get.

So, how do engineers solve this? Three dominant approaches:

• Planetary gears: Stack multiple stages for high ratios
• Harmonic drives: Achieve huge ratios in single stage via flex
• Cycloidal drives: Use eccentric motion for compact reduction

(Excludes belts, tendon, worm gears, ball screws, fluid)

Each solves some problems while creating others.

Planetary gears work like a solar system - central "sun" gear surrounded by "planet" gears inside a ring.

Great efficiency (95-98%) but the flaw: tiny gaps between teeth multiply by 50x reduction and your robot can't draw a straight line.

Harmonic drives flex a thin metal cup with an oval cam to create massive reductions in one step.

Zero slop, but they cost $1,000-$3,000 each and only 3-4 companies worldwide can make them properly. Supply chain bottleneck.

Cycloidal drives use an off-centre disc that "wobbles" against pins for reduction.

Super rigid and strong, but complex to manufacture. You need aerospace-level precision machining.


This is an even bigger problem for humanoids that have even tougher constraints:

1. Safety (must not injure humans during interaction)
2. Cost (sub-$20,000 robot requires ~$500 actuators)  
3. Performance (zero backlash, high bandwidth, force transparency, low weight)

Current transmissions force trade-offs. Humanoids need all three.

This is why the industry is chasing direct drive - motors directly powering joints with no transmission at all.

Direct drive would solve everything: zero backlash, perfect force transparency, infinite stiffness, no gear complexity. It's the holy grail.

But no one has cracked it due to motor size and cost constraints. 

The compromise? Quasi-direct drive (QDD) with low gear ratios (6:1 to 20:1). 

Good but not perfect. 

Enter the Archimedes Drive from IMSystems—a departure from gear-based thinking.

Instead of metal teeth, they use controlled friction between rolling cylinders under compression. 

Think: train wheels on tracks, but in a planetary configuration.

The physics are elegant:
• Hollow rolling cylinders replace gears
• Pre-compression creates spring-loaded contact
• Force transfers through traction, not gear mesh
• Variable geometry enables ratios from 6:1 to 100:1+

Think: infinite gear teeth at the microscopic level.

Three breakthrough characteristics:
• Zero backlash from continuous rolling contact
• Inherent slip protection - overtorque causes slipping, not breaking  
• Perfect back-driveability for human-safe interaction

You can hit it with a hammer and it just slips safely at preset limits.

Manufacturing advantages:

Traditional gears need specialised grinding equipment ($70K-$100K per tool) with limited suppliers.

The Archimedes Drive uses bearing manufacturing - a mature, scalable industry. Early estimates suggest 50-80% cost reduction vs harmonic drives.

The counterintuitive insight: what made this drive "unsuitable" for industrial robots makes it ideal for humanoids.

@IMSystemsNL spent 10 years being "too early" - industrial robots saw slip as a liability.

Now humanoid companies specifically request slip protection for safety. Yesterday's weakness became today's advantage.

Prediction: Within 5 years, two distinct ecosystems:
• Industrial: High-ratio drives optimised for precision
• Humanoid: Low-ratio friction drives optimised for safety and cost

The Archimedes Drive represents a step in this direction.

The bigger picture: We're at an inflection point similar to the transition from mechanical to electronic systems.

Is the next chapter isn't about making better gears or about eliminating gears?

Here's the link to the full video: 
https://www.youtube.com/watch?v=XqOumFdvZps&t=573s&ab_channel=MarwaElDiwi

Want access to the latest robotics jobs?

🌐 Our website updates with new roles every Friday

📧 Newsletter subscribers get early access on Wednesday

Stay ahead of the curve → https://www.robot-iq.org/

Cracking Actuator Cost Challenge

ARK believes that if humanoids can achieve just 75% of human productivity, they pay for themselves. This thesis boils down to a simple economic argument: the cost equation fundamentally shifts in favor of robots.

**Economic Analysis**

Their NPV analysis shows that if Tesla hits its projected Optimus pricing (~$25K), robots only need 75% of human productivity to break even over 10 years. That's a lower bar than traditional industrial automation, which typically requires 200%+ gains. The math assumes $46/hour total human labor costs and standard work schedules. At Tesla's price point, robots become profitable even while underperforming humans—they just need to work consistently without benefits, sick days, or turnover.

**Market Implications**

Why this matters: It expands the addressable market from "jobs robots do better" to "jobs robots can do adequately for less." That's potentially most manual labor roles.

**The Bull Case**

AI breakthroughs, falling component costs, and global labor shortages create perfect conditions for mass adoption.

**The Sceptical View**

The analysis assumes Tesla hits aggressive cost targets, robots achieve reliable 75% productivity, and ignores integration/maintenance complexity. Plus, 10-year payback periods are long—most manufacturers want ROI under 5 years. Real-world deployment often reveals hidden costs.

**Conclusion**

Bottom line: ARK's thesis isn't about robots being amazing—it's about robots being so cheap that "good enough" becomes transformative. This is a lower bar by far! 