The Age of AI: Exciting Times Ahead for Humanity… If We Do It Right

A paralyzed British Army veteran is raiding dungeons in World of Warcraft using nothing but his thoughts. A robot that learned to fold laundry by watching YouTube videos is now building BMWs. An AI designed a drug compound in weeks that would have taken human researchers a decade. And a car drove itself across San Francisco without a human touching the wheel.

All of this happened in the last twelve months.

We are living through the most concentrated period of technological advancement in human history, and it's moving so fast that breakthroughs that would have dominated headlines for weeks now get a single news cycle before the next one lands. If you're a tech enthusiast — and if you're reading TechSource, you probably are — this is the most exciting time to be alive.

But excitement and anxiety are not mutually exclusive. The same AI that designs life-saving drugs can also design cyberweapons. The same robots building cars could eventually displace the workers who used to build them. The same brain-computer interface restoring a veteran's digital life raises questions about who controls the data flowing from your neurons.

This is the duality of 2026: the future is arriving faster than we can process it, and whether it's wonderful or terrible depends entirely on the choices we make right now.

The Breakthroughs That Should Blow Your Mind

1. Robots are no longer science fiction.

I wrote about this in detail in my recent post on Linux-powered robots, but it bears repeating: humanoid robots are now building cars in BMW factories, stocking shelves in Amazon warehouses, and approaching consumer pricing. Tesla's Optimus Gen 3 entered production in January. Unitree filed a $610 million IPO with 335% revenue growth. You can pre-order a home robot for $499 a month. Prices have dropped from $85,000 to $1,400 in three years. The "robot butler" that every sci-fi movie promised us is no longer a fantasy — it's a product roadmap.

2. Self-driving is quietly becoming real.

While most people are still debating whether self-driving cars work, Waymo is completing over 200,000 paid rides per week across multiple US cities. Tesla's Full Self-Driving (Supervised) is navigating city streets, handling intersections, and making lane changes with increasing competence. The technology isn't perfect — it still requires human supervision — but the trajectory is undeniable. Within 3-5 years, the question won't be whether autonomous vehicles work, but whether humans should still be allowed to drive manually given how much safer the machines are becoming.

3. Neuralink is giving paralyzed people their digital lives back.

This one hits different. Twelve patients have now received Neuralink's N1 brain implant. Jon Noble, a British Army veteran paralyzed from the shoulders down after a car accident, received his implant in December 2025. Within weeks, he was controlling a MacBook cursor with his thoughts. By day 100, he was playing World of Warcraft using only his brain-computer interface. He described the experience as "science fiction that somehow became my everyday reality."

In May 2026, Neuralink achieved another milestone: implanting electrode threads through the brain's dura membrane without removing it — a less invasive procedure that could make the surgery safer and faster. The company is targeting high-volume production and near-fully automated surgical procedures this year, with Musk aiming for 1,000+ implants in 2026. Competitor Synchron is taking a different approach entirely, threading its device through blood vessels to avoid brain surgery altogether — lower signal quality, but dramatically lower risk.

The implications extend far beyond helping paralyzed patients (though that alone would be enough to justify the technology). If brain-computer interfaces can decode thought into digital action, the long-term possibilities — restoring speech, treating neurological disorders, even augmenting human cognition — are staggering.

4. AI is rewriting medicine and longevity research.

The global anti-aging market surpassed $85 billion in 2025, and for the first time, the science is keeping pace with the money. AI platforms are accelerating drug discovery at unprecedented speed. Insilico Medicine — an AI-driven biotech company that went public in Hong Kong in December 2025 — formed the industry's first Longevity Board in April 2026 to oversee AI-enabled aging research. Eli Lilly paid $115 million upfront (with up to $2.63 billion in milestones) for rights to develop therapeutics from Insilico's AI platform.

The numbers are striking: AI-discovered drug compounds show 80-90% Phase I success rates, compared to the historical average of 40-65%. Senolytic therapies that clear damaged cells from the body are showing strong clinical results. Low-dose rapamycin studies continue to demonstrate promising longevity effects. AI-designed compounds achieved a 70% hit rate in validating lifespan extension in laboratory organisms. Many researchers now believe adding 10-20 healthy years to the human lifespan is achievable within the next two decades.

We're not talking about immortality. We're talking about spending more of your life being healthy, active, and functional — dying at 95 feeling like 70 instead of dying at 75 feeling like 95. As someone who has run marathons and an ultramarathon and become deeply invested in personal health, this is the AI application that excites me most.

The Part Where I Stop Smiling

Now here's the thing about powerful technology: it doesn't come with a moral compass. A hammer builds houses and breaks skulls with equal indifference. AI is the most powerful hammer humanity has ever created, and the skull-breaking potential is keeping a lot of very smart people up at night.

1. Cybersecurity is becoming an AI-vs-AI arms race. 

Global cybersecurity spending surged past $240 billion in 2026 — up 12.5% from last year — and it's still not enough. AI has "democratized high-level hacking," enabling attackers to craft realistic phishing emails, generate deepfakes indistinguishable from real people, and create self-modifying malware that evolves to evade detection. Anthropic (the company behind Claude, the AI I use daily) published a report analyzing 832 banned malicious accounts and found that AI significantly amplifies attacker capabilities while making attacks increasingly autonomous.

The scariest development: autonomous cyber agents that can scan networks, exploit vulnerabilities, and make tactical decisions independently — at machine speed. These aren't theoretical. They exist now. Palo Alto Networks notes that autonomous AI agents already outnumber human operators 82:1 in some enterprise environments. One forged command from a deepfake "CEO doppelganger" can trigger an automated chain reaction before any human notices something is wrong. Organizations using AI-driven security detect breaches 98 days faster than those relying on manual methods, but the attackers are using the same AI to move faster too. It's an arms race with no finish line.

2. Autonomous weapons are no longer hypothetical.

Military AI systems capable of identifying, targeting, and engaging threats without human intervention are in active development by multiple nations. The same computer vision that lets a Tesla identify a pedestrian can let a drone identify a target. The same neural networks that help a robot learn to fold laundry can help a weapons system learn to navigate a battlefield. The ICRC has called for international regulation of autonomous weapons, but progress has been glacial while the technology accelerates.

3. Robots in your kitchen sounds great until it doesn't.

The same humanoid robot that helpfully carries your groceries is a 125-pound bipedal machine running on software in your home. Software has bugs. Networks get hacked. As any Linux user who has experienced a kernel panic can tell you, "it runs on software" is not always reassuring. What happens when a home robot malfunctions? What happens when someone hacks a fleet of delivery robots? What happens when the AI controlling a warehouse robot makes a decision that injures a human worker? These aren't paranoid fantasies — they're engineering problems that need to be solved before we put these machines in every household.

4. Job displacement is real and accelerating.

AI can now write code, generate marketing copy, create images, compose music, handle customer service, analyze legal documents, and diagnose medical images. Each of these capabilities threatens jobs held by real people. The optimistic view is that AI creates new jobs to replace the ones it eliminates. The realistic view is that the transition will be painful, uneven, and particularly hard on people without the resources to retrain. History tells us that technological revolutions ultimately create more prosperity — but history also tells us that the "ultimately" part can take decades, and the people caught in the transition often suffer.

5. The concentration of power problem.

The companies building the most capable AI systems — OpenAI, Google, Meta, Anthropic, and a handful of others — are accumulating an unprecedented concentration of technological power. The compute required to train frontier models costs hundreds of millions of dollars, creating a barrier to entry that effectively limits who gets to shape the future of intelligence itself. Open-source models (which I wrote about in my Ollama post are a crucial counterweight, but the gap between open-source and frontier models remains significant.

Doing It Right

So how do we navigate this? How do we capture the upside — the cured diseases, the restored mobility, the extended lifespans, the freed-up human potential — without the downside consuming us?

I don't pretend to have all the answers. But after nearly two decades of writing about technology on this site, watching Linux go from underdog to dominant, watching Bitcoin go from joke to asset class, and watching AI go from academic curiosity to civilization-altering force, I've noticed a pattern: the technologies that serve humanity best are the ones that are open, transparent, and distributed rather than closed, opaque, and concentrated.

Open-source AI matters. Regulatory frameworks that move at the speed of the technology (not the speed of government) matter. International cooperation on autonomous weapons matters. Investing in workforce transition programs before the displacement hits matters. Ensuring that brain-computer interface data is owned by the patient, not the company, matters.

The technology itself is neither good nor evil. It's a mirror that reflects the intentions of whoever wields it. The same AI that can generate a deepfake to steal millions can generate a drug compound to save millions. The difference is governance, ethics, and the choices we make as a society about what we build and who gets to control it.

The View from Home

I'm writing this from a small town in Bohol, Philippines. I have solar panels on my roof, a Starlink dish for internet, a Mac Mini running open-source AI in my living room, and a Tesla Model Y L now in my garage. I track my health with a Garmin and an Apple Watch. I build iOS apps using AI-assisted tools. I've been blogging about technology since 2007.

From where I sit, the age of AI feels like every other technological revolution I've witnessed and written about — thrilling, terrifying, and utterly dependent on the humans involved. Linux won because it was open and people built on it collectively. Bitcoin survived because decentralization made it resilient. The technologies that endure are the ones that empower people rather than control them.

AI will be the defining technology of our lifetime. It will cure diseases we thought were untreatable. It will give paralyzed veterans the ability to play video games with their minds. It will put robots in our factories and eventually our homes. It will extend human lifespans and redefine what it means to age. It will also be weaponized, exploited, and misused in ways we can't fully predict.

Exciting times ahead for humanity. If we do it right.

— Jun

Tesla Model Y L vs BYD Tang EV: Which 6/7-Seater Electric SUV Is Better Value in the Philippines?

Two massive electric SUVs. Both seat six or more. Both have dual motors, all-wheel drive, and enough torque to pin you to your seat. Both are available right now in the Philippines. And both want to be the EV that finally convinces Filipino families to ditch gasoline forever.

The Tesla Model Y L landed at ₱2,849,000. The BYD Tang EV has been here since 2023 at ₱3,321,000. That’s a ₱472,000 gap — in Tesla’s favor — which is not the pricing dynamic most people expected. The world’s most famous EV brand is somehow the cheaper option.

So which one deserves your money? I broke this down the way a tech person would: by the specs that actually matter in Philippine driving conditions. Let’s get into it.

*Price

Tesla Model Y L — ₱2,849,000 (6-seater, dual motor AWD)

BYD Tang EV — ₱3,321,000 (7-seater, dual motor AWD)

Nobody saw this coming. The Tesla — the brand associated with Silicon Valley premiums — costs ₱472,000 less than the BYD. That’s enough to install a home solar panel system and charge your Tesla for free. Or buy roughly 8,000 liters of gasoline at current prices (which, given the Iran war oil shock, is probably 6,000 liters by the time you read this).

Both vehicles benefit from zero import duties on EVs through 2028 under Executive Order No. 62, and both carry identical 8-year / 160,000 km battery warranties.

The wildcard: BYD also offers the Tang DM-i at ₱2,098,000 — a 7-seater plug-in hybrid with 110 km of electric range and a gas engine backup. It’s a different category, but if your budget is tighter or you live somewhere with zero charging infrastructure, it’s hard to ignore.

*Performance

The Tesla makes 506 hp and 590 Nm; the Tang makes 482 hp and 660 Nm. The Tang is quicker off the line at 4.6 seconds to 100 km/h versus the Tesla’s 5.0 seconds. Both are dual motor all-wheel drive. Unless you’re drag racing at stoplights, you won’t notice the difference in daily driving.

*Range and Efficiency

This is the Tesla’s biggest advantage, and it matters more in the Philippines than almost any other market.

The Model Y L gets 681 km of WLTP range from an 88.2 kWh battery — that’s 7.7 km per kWh. The Tang gets 530 km from a *larger* 108.8 kWh pack — just 4.9 km per kWh. Tesla is extracting 57% more range per kilowatt-hour. It also fast charges quicker at 250 kW versus the Tang’s 170 kW.

Why does this matter so much here? Because our charging infrastructure is still in its infancy. Tesla Superchargers exist only in Metro Manila. Fast chargers outside NCR are sparse. When your nearest reliable fast charger might be hundreds of kilometers away, every extra kilometer of range is a safety net.

With 681 km, you could drive from Manila to Legazpi and back without charging. You could loop the entire island of Bohol nine times. For daily driving of 30-50 km, you’re charging once a week at home. The Tang’s 530 km is still excellent, but the psychological comfort of that extra 151 km matters when range anxiety is the number one barrier to EV adoption in this country.

*Interior

The Tang seats seven with a bench second row and a 50:50 split third row. The interior leans premium-traditional: Nappa leather, European wood trim, space-grade aluminum accents, ambient lighting, panoramic sunroof, and a dual-tone cabin with golden orange seat accents. It’s a warmer, more luxurious feel than the Tesla’s minimalism.

The Tesla seats six with ventilated, reclining captain’s chairs in the second row with power armrests. The third row gets its own air vents. The cabin is Tesla-minimalist: clean lines, a massive screen, and very little else. Some call it elegant. Others call it sterile.

The practical reality: both third rows are tight. Comfortable for children, tolerable for adults on short trips, cramped for anything longer. But the Tesla’s captain’s chair configuration makes getting in and out of the third row significantly easier.

Cargo is a blowout: 2,539 liters for the Tesla versus 1,655 liters for the Tang. If you’re a family that travels with luggage, strollers, and the accumulated stuff of daily life, that’s a massive difference.

*Software and Tech

This is where my tech brain takes over, and where the two vehicles diverge most dramatically.

Tesla’s software is its superpower. The Model Y L receives meaningful over-the-air updates that add genuine new features — not just bug fixes. Tesla owners worldwide have woken up to new Autopilot capabilities, improved range through software optimization, and even performance upgrades pushed wirelessly. The 16-inch center touchscreen (plus an 8-inch rear display) controls everything. Sentry Mode turns the car’s cameras into a 360-degree security system. The built-in dashcam records continuously. The Tesla app lets you monitor charge status, precondition the cabin in Philippine heat, and locate your car from anywhere.

Autopilot comes standard — adaptive cruise control, lane centering, and lane keeping. The optional Full Self-Driving (Supervised) goes further with city street navigation, intersection handling, and automatic lane changes. Whether FSD will be fully functional on Philippine roads remains to be seen, but the hardware is there, ready.

BYD’s tech is more traditional but hits some marks Tesla misses. The Tang has a 15.6-inch rotating touchscreen (it physically rotates between landscape and portrait — a neat party trick), plus a 12.3-inch instrument cluster behind the steering wheel. This is significant because Tesla has no instrument cluster at all — everything lives on the center screen, which takes getting used to.

The Tang supports Apple CarPlay and Android Auto. The Tesla does not. For many Filipino buyers deeply embedded in their phone ecosystems, this alone could be a dealbreaker.

But BYD’s OTA updates are mostly incremental fixes, not transformative additions. There’s no equivalent to Sentry Mode. The ADAS is Level 2 capable but doesn’t approach Autopilot’s sophistication. And the app experience, while functional, isn’t in the same league as Tesla’s.

The bottom line on tech: if you think of your car as a gadget — a rolling computer that gets smarter over time — Tesla is in a class of its own. If you want a car that works conventionally with CarPlay and a familiar instrument cluster, BYD feels more comfortable.

*Service in the Philippines

BYD wins this one decisively.

BYD Philippines, through its distributor ACMobility, has 20+ dealerships and service centers nationwide — Manila, Cebu, Davao, and other major cities. If something goes wrong with your Tang, help is relatively accessible no matter where you are.

Tesla Philippines has one service center. In BGC, Taguig. That’s it.

If you live in Bohol (like me), Cebu, Davao, Iloilo, or anywhere outside NCR, getting your Tesla serviced means shipping it to Manila or flying there. Tesla offers mobile service for some issues and handles many diagnostics remotely through the app, but for anything physical, you’re looking at a logistical challenge.

Tesla has announced Supercharger locations for Cebu and other areas, and where infrastructure goes, service often follows. But right now, BYD’s nationwide dealer network is a concrete advantage — especially for provincial buyers.

*Charging

Tesla operates Supercharger stations at Uptown Mall, SM Mall of Asia, Shangri-La Plaza, and Opus Mall (all Metro Manila), plus destination chargers at Eastwood, Venice Grand Canal Mall, Okada Manila, Century City Mall, and Vista Mall Antipolo. Superchargers run at up to 250 kW for ₱19/kWh. Destination chargers run at 7-11 kW for ₱16/kWh.

BYD owners can use the broader third-party CCS2 charging network, including ACMobility’s Greenstrum stations, which are more spread out across the country. Since both vehicles use CCS2, they can technically use the same third-party chargers.

Tesla’s Supercharger experience is smoother though — plug in, automatic authentication, automatic billing through the app. No fumbling with RFID cards or third-party apps. The V4 Supercharger hardware is also faster and more reliable than most third-party DC chargers currently deployed here.

That said, both vehicles charge at home overnight with a wall connector, which is how most EV owners charge 90% of the time. If you have solar panels (like I do), your daily driving cost approaches zero regardless of which one you choose.

*So Which One Should You Buy?

Buy the Tesla Model Y L if you want maximum range for the lowest price, you value cutting-edge software and Autopilot, you think of your car as a tech product that improves over time, you want the most cargo space, you prefer captain’s chair seating, you’re okay with Manila-only service for now, and you can live without Apple CarPlay.

Buy the BYD Tang EV if you need seven seats, CarPlay is non-negotiable, you want a service center outside Metro Manila, you prefer a traditional instrument cluster, you value premium interior materials, and you want faster 0-100 acceleration.

Consider the BYD Tang DM-i if your budget is tighter at ₱2,098,000, you want seven seats with zero range anxiety thanks to the gas engine backup, or you want to ease into electrification gradually.

*My Pick

The range sealed it for me. Living in Bohol with zero Tesla Superchargers and limited third-party fast charging, 681 km means I can comfortably handle any trip on the island and beyond with home charging alone. My solar panels generate more than enough to keep it topped up. The tech ecosystem — Autopilot, Sentry Mode, OTA updates, the Tesla app — aligns perfectly with how I think about products as a tech person and iOS developer.

The BYD Tang EV is a genuinely excellent vehicle. It’s faster off the line, more luxurious inside, supports CarPlay, and has a far superior service network. If Tesla’s service situation gives you pause — and it should, honestly — the Tang is a fantastic alternative. And the Tang DM-i at ₱2,098,000 is arguably the smartest practical choice for Filipino families who want electrified multi-row seating without any charging anxiety.

But for me, at ₱472,000 less, with 151 km more range, and with the kind of software that makes my inner tech nerd light up every time I use it — the Model Y L was the obvious choice.

-Jun 

I Emailed Python’s Creator in 2007. The Language Now Runs the World.

In August 2007, a few weeks after launching this site, I did something that still surprises me when I think about it: I emailed Guido van Rossum — the creator of Python and the language’s self-titled “Benevolent Dictator For Life” — to ask for advice on starting a Python User Group in the Philippines.

To my genuine shock, he replied. Quickly. With actual instructions on how to get it started.

That email led to a blog post called “Will Real Python Hackers Please Stand Up,” which became a rallying cry for Filipino Python enthusiasts. Comments trickled in from across the archipelago — a math professor from Ateneo de Manila teaching Python in discrete mathematics classes, a developer from Austin, Texas who was moving to Bohol and wanted to connect, a 17-year-old IT student eager to join. By May 2008, we had Pinoy PUG— the Pinoy Python User Group — up and running on Google Groups.

It was a tiny community. A handful of enthusiasts scattered across islands, connected by a shared interest in a programming language that most of the tech world was still ignoring.

That was 19 years ago. Today, Python is the most popular programming language on the planet.

The Numbers Are Absurd

Python currently holds the #1 position on the TIOBE Index with a 21.25% share as of March 2026 — nearly double the second-place language. It surpassed JavaScript to become the most-used language on GitHub. The 2025 Stack Overflow Developer Survey shows 57.9% of respondents using Python, a 7-point jump from the previous year. It’s the most demanded language by recruiters worldwide. IEEE, RedMonk, PYPL — pick any ranking system you want. Python sits at or near the top of all of them.

When I was recommending Python books on this site in 2007, the language was ranked 7th on TIOBE. When I wrote about a Java developer’s experience switching to Python in 2008, Java was the undisputed king and Python was the quirky underdog that “serious” enterprise developers dismissed. When I published “How to Learn Python Quickly” in 2014, it was popular but still fighting for mainstream respect.

Now? Python doesn’t fight for anything. It is dominating.

What Happened? The short answer: AI happened.

The slightly longer answer: Python became the default language for every transformative technology of the past decade — machine learning, data science, deep learning, natural language processing, computer vision, and generative AI — and rode that wave from “popular scripting language” to “the most important programming language in the world.”

Every major AI framework is Python-first. PyTorch, TensorFlow, Keras, Hugging Face Transformers, LangChain, scikit-learn — the entire stack that powers ChatGPT, Claude, Gemini, Stable Diffusion, and every other AI tool you’ve used was built in or interfaces primarily through Python. When researchers at OpenAI, Google DeepMind, or Meta AI publish a breakthrough paper, the reference implementation is almost always in Python. When a startup builds an AI product, the prototype is in Python. When a data scientist trains a model, the notebook is in Python.

The language that Guido van Rossum created in 1991 — designed to be readable, simple, and fun — turned out to be the perfect tool for the most complex technology humanity has ever built. There’s a beautiful irony in that. The language optimized for human readability became the language that teaches machines to think.

But AI alone doesn’t explain it. Python was already climbing before the ChatGPT era. Data science and the pandas/NumPy/Jupyter ecosystem had been pulling developers in for years. Web frameworks like Django and Flask powered major applications. DevOps and automation scripts ran on Python. It was the first language taught in many computer science programs because its syntax reads almost like English. The AI boom didn’t create Python’s dominance — it accelerated a trajectory that was already inevitable.

What I Got Right (And What I Missed)

Looking back at my old Python posts, I’m struck by how much I got right about the language’s potential and how completely I failed to predict the scale.

In 2007, when I emailed Guido and tried to build a Filipino Python community, I clearly believed the language was special. I wrote about it with the same evangelical enthusiasm I brought to Linux — another underdog technology that everyone underestimated. I recommended books on Python to my readers. I shared a Java developer’s “Pythonic experience” converting to the language. I saw something in Python’s simplicity and elegance that felt important.

What I didn’t predict was that Python would become the lingua franca of artificial intelligence, that it would dethrone JavaScript on GitHub, that it would command a quarter of the entire TIOBE index, or that knowing Python would become essentially mandatory for anyone working in tech. I thought it would grow. I didn’t think it would consume the world.

I also didn’t predict my own path. The guy who co-founded Pinoy PUG and evangelized Python eventually became an iOS developer writing Swift. I still use Python — for web scraping scripts, automation pipelines, data processing, and the backend work on my projects — but it’s no longer my primary language. The irony of abandoning my first love right before it became the most popular programming language in the world is not lost on me. It’s like selling Bitcoin at $500. Which, for the record, I also… let’s not go there.

Why Python?

Python’s dominance comes down to something I recognized back in 2007 but couldn’t articulate as clearly: it optimizes for the right thing.

Most programming languages optimize for the computer — speed, memory efficiency, type safety, compilation performance. Python optimizes for the human. It prioritizes readability over cleverness. It values simplicity over syntactic gymnastics. It lets you express complex ideas in fewer lines of code than virtually any alternative.

This design philosophy — which Guido baked into the language from day one — turned out to be the perfect match for an era where the bottleneck isn’t computing power (we have plenty) but human attention and talent (we don’t). When you need millions of people to learn programming quickly, when researchers need to prototype ideas before they know if they’ll work, when AI developers need to iterate on model architectures at breakneck speed — you need a language that gets out of the way and lets humans think. Python is that language.

It’s also why AI coding assistants love Python. GitHub Copilot, Claude, and every other AI code generation tool produces better Python than almost any other language, because Python’s clear syntax maps naturally to the training data. The language designed for human readability turned out to be optimally readable for machines too. A virtuous cycle: AI makes Python easier to write, Python makes AI easier to build.

The Guido Email, 19 Years Later

I sometimes think about that email exchange with Guido van Rossum. Here was a random Filipino blogger cold-emailing the creator of a programming language, asking for help starting a user group on a tropical island. And Guido — who at the time was working at Google, managing one of the most important open-source projects in the world — just… replied. With helpful advice. Because that’s what the open-source community does.

That moment captures everything I love about technology. The barriers are low. The community is generous. A kid from Bohol, Philippines can email the creator of a language that now powers artificial intelligence, and get a response.

Pinoy PUG may not have become the massive community I envisioned. But Python did become everything I hoped for the language and more. The math professor from Ateneo who commented on my 2007 post, wanting to teach Python to freshman students — that’s now happening at universities everywhere. The 17-year-old IT student who wanted to join our group — today’s 17-year-olds are learning Python as their first language, using it to build AI projects that would have been science fiction in 2007.

The language won. And in some small, perhaps insignificant way, I’m proud that TechSource was out there telling Filipinos to learn Python before most of the world had caught on.

Now if you’ll excuse me, I have some Python scripts to refactor. Nineteen years later, I’m still hacking.

— Jun

The State of Linux-Powered Robots: From Lego Kits to World Domination

In 2009, I wrote a TechSource article called “[5 Awesome Robot Kits to Get You Started with Robotics].”The most advanced robot on that list was a LEGO Mindstorms NXT. It had three servo motors, four sensors, and the approximate intelligence of a toaster with ambitions.

Two years later, I followed it up with “[Best Robotics Software for Linux],” where we covered tools like ROS, Player, and CARMEN. At the time, the state-of-the-art in Linux robotics was getting a wheeled platform to navigate a hallway without bumping into things. We were thrilled. The robot didn’t crash into a wall! Ship it!

It’s now 2026. Humanoid robots are walking around BMW factories building cars. Tesla is converting its Fremont production line to mass-manufacture humanoid robots. A Chinese company called AgiBot just shipped its 10,000th humanoid unit. You can pre-order a home robot that folds your laundry for $499 a month. And virtually all of them — from the billion-dollar Tesla Optimus to the $1,400 Noetix Bumi — run on Linux.

The penguin didn’t just conquer servers and supercomputers. It’s learning to walk.

Linux: The Operating System of Every Robot That Matters

Here’s a fact that shouldn’t surprise anyone who read my recent post “[Linux Won, and Nobody Noticed]” : Linux is the dominant operating system in robotics, and it’s not even close.

The Robot Operating System (ROS) — which we covered on this site back in 2011 when it was a scrappy open-source project — is now the global standard for robot development. ROS 2, its mature successor, runs on Linux and provides the middleware that connects sensors, actuators, AI models, and control systems. Nearly every serious robotics company on Earth uses it. Unitree’s humanoids? ROS-compatible. Boston Dynamics’ Atlas? Built on Linux. Figure AI’s warehouse bots working at BMW? Linux. Amazon’s Digit robots? Linux. The entire humanoid robotics industry is standing on open-source shoulders.

Why Linux? The same reasons it won everywhere else: it’s free, customizable to the extreme, has the best real-time kernel support, runs on anything from a Raspberry Pi to a GPU cluster, and doesn’t require paying Microsoft a licensing fee for every robot you build. When you’re manufacturing 10,000 humanoids, that last point alone saves you a small fortune.

What’s Actually Happening Right Now

The humanoid robotics space in 2026 is moving at a pace that makes the smartphone revolution look leisurely. Here’s the state of play:

1. Tesla Optimus Gen 3 started production in January 2026 at Tesla’s Fremont factory (they literally stopped making the Model S and X to make room for robots — let that sink in). The robot stands 5’8”, weighs 125 pounds, has 22 degrees of freedom in its hands, and uses the same neural network AI that powers Tesla’s Full Self-Driving. Musk is targeting 50,000-100,000 units in 2026 and eventually a million per year, at a target price under $30,000. Reality check: on the Q4 2025 earnings call, Musk admitted the current robots aren’t doing “useful work” yet — they’re still in the learning phase. But the manufacturing infrastructure is real.

2. Figure AI’s Figure 03 completed an 11-month pilot at BMW’s South Carolina plant, helping build over 30,000 cars. It runs on Figure’s own “Helix” AI — a Vision-Language-Action model that lets you say “I spilled my coffee” and the robot understands it needs to find a towel and clean the floor. No specific programming required. It just… understands context. That’s terrifyingly impressive.

3. Unitree — the company that made robotics accessible with the $16,000 G1 — just filed for a $610 million IPO with 335% revenue growth. They shipped roughly 5,500 humanoids in 2025, targeting 20,000 in 2026. They also open-sourced UnifoLM, a Vision-Language-Action model that lets their G1 autonomously perform household tasks. Open-source AI running on open-source Linux, controlling an affordable robot. My 2009 self who was excited about LEGO Mindstorms would be losing his mind right now.

4. China dominates production. Chinese companies (Unitree, AgiBot, Fourier, UBTECH, Kepler, XPENG, EngineAI) produced roughly 90% of all humanoid robots shipped globally in 2025. AgiBot hit 10,000 cumulative units in March 2026 — doubling from 5,000 in just three months. XPeng’s IRON robot went viral because its walking gait was so human-like that people accused them of putting a person inside a suit. They had to open the robot’s casing on stage to prove it was real.

5. Prices are falling off a cliff. In 2023, the cheapest capable humanoid was around $85,000. In 2026, Unitree’s R1 costs $5,900. Noetix’s Bumi hit $1,400 — consumer electronics pricing for a humanoid robot. 1X Technologies offers their NEO home robot for $499/month. Tesla is targeting under $20,000 at scale. Within 3-5 years, capable humanoids could approach appliance pricing. Your next Roomba might have legs.

The AI Ingredient That Changed Everything

The reason robots went from “bumping into walls” to “building BMWs” in fifteen years can be summed up in two words: neural networks.

Traditional robotics programming was painstaking. You had to code every movement, every response to every sensor input, every edge case. It’s why the software we covered in 2011 — Player, CARMEN, Fawkes — was primarily about navigation and sensor control. Getting a robot to walk down a hallway without incident was a genuine achievement.

Modern humanoid robots don’t work that way. They learn. Companies feed millions of videos of humans performing tasks into neural networks, and the robot watches and mimics. Tesla’s Optimus uses the same end-to-end neural networks as Full Self-Driving. Figure’s Helix model processes vision, language, and physical action simultaneously. Unitree open-sourced a model that lets robots learn household tasks autonomously.

This is where Linux’s dominance in both AI and robotics converges into something genuinely historic. The AI models are trained on Linux GPU clusters. The training frameworks (PyTorch, TensorFlow) run on Linux. The robot middleware (ROS 2) runs on Linux. The robot’s onboard computer runs Linux. It’s Linux all the way down — from the data center that trains the brain to the machine that uses it.

ROS in 2011 helped your robot avoid walls. ROS 2 in 2026 helps your robot understand natural language, navigate unstructured environments, manipulate objects with 22-degree-of-freedom hands, and learn new tasks by watching YouTube videos of humans. Same open-source foundation, incomprehensibly different capability.

The Near Future (Hold Onto Your Keyboards)

Based on current trajectories, here’s what the next 2-5 years likely looks like:

1. 2026-2027: Humanoid robots become common in factories and warehouses. Amazon, BMW, Mercedes-Benz, and Foxconn are already deploying them. Tesla targets external Optimus sales by late 2027. Agility Robotics pursues the first ISO safety certification for a humanoid to work alongside humans without barriers.

2. 2027-2028: The first wave of consumer home robots arrives. 1X’s NEO is already taking pre-orders for 2026 delivery. Figure is building a “BotQ” factory specifically for consumer-grade humanoids. Expect early home robots to handle simple tasks — carrying groceries, basic cleaning, fetching items — with the grace of a helpful but slightly confused intern.

3. 2028-2030: Prices hit the $5,000-$10,000 range for capable home humanoids. The combination of mass manufacturing (mostly in China), falling component costs, and improved AI training creates a positive spiral. Robots that learn from each other across a shared network improve faster than any single unit could alone — the same fleet learning approach Tesla uses for FSD, applied to physical tasks.

4. The wildcard: Open-source humanoid platforms. Unitree is already open-sourcing AI models. If someone builds the “Ubuntu of humanoid robots” — a fully open-source hardware and software stack that anyone can build on — the pace of innovation could accelerate beyond anything we’ve seen. Given that Linux and ROS already provide the foundation, this isn’t fantasy. It’s an engineering challenge with a clear path.

Should You Be Excited or Terrified?

Both. Simultaneously. That’s the correct emotional response to watching a robot learn to fold laundry by watching humans do it on video.

The optimistic case: humanoid robots handle the tedious, dangerous, and repetitive work that humans don’t want to do. They care for aging populations. They work in disaster zones. They build things faster and cheaper, making goods more affordable. They run on Linux and open-source AI, meaning the technology isn’t locked behind any single corporation.

The concerning case: job displacement happens faster than retraining. The wealth generated by robot labor concentrates among those who own the robots. Privacy and surveillance concerns multiply when humanoid machines with cameras and microphones populate public spaces. And, as any Linux user who has experienced a kernel panic can tell you, the phrase “it runs on software” is not always reassuring when the software is controlling a 125-pound bipedal machine in your kitchen.

The realistic case: it’ll be messy, uneven, and slower than the hype suggests, but faster than skeptics expect. Just like every other technology revolution. Just like Linux itself — which took decades to go from a Finnish student’s hobby project to running 100% of the world’s supercomputers.

From LEGO Mindstorms to This

Seventeen years ago, I wrote about LEGO Mindstorms with three servo motors as a gateway to robotics. Today, a $16,000 humanoid robot runs open-source Linux, learns from neural networks, and can do backflips.

Fifteen years ago, I covered ROS as a promising Linux-based robotics framework. Today, it’s the global standard powering robots that build cars, stock warehouses, and are learning to clean your house.

The thread connecting my 2009 article to this one is the same thread that has connected every post on TechSource since 2007: open-source software, running on Linux, quietly becoming the foundation of the future while most people aren’t paying attention.

Except now the future has legs. And hands. And it’s learning to fold your laundry.

Sleep tight.

— Jun

Ubuntu 24.04 LTS vs. macOS 26 Tahoe: The Free OS That Rivals a Premium Experience

I’m writing this on a MacBook Air running macOS 26 Tahoe, and I keep glancing at my Mac Mini in the corner — the one running Ubuntu 24.04 LTS.

I’ve been a macOS user for a decade. I develop iOS apps. I’m neck-deep in the Apple ecosystem — iPhone, Apple Watch Ultra, AirPods, the whole cult membership. But last year, Apple released macOS Tahoe with its Liquid Glass redesign, and I found myself wondering: has the free operating system actually gotten *better* than the premium one?

Short answer: in some ways, yes. I’m not a fanboy for either side. I’ve lived in both worlds. Here’s what I found.

Liquid Glass: Pretty or Pretty Annoying?

You can’t talk about macOS 26 Tahoe without talking about Liquid Glass — Apple’s new translucent, depth-infused design language that makes UI elements look like layers of actual glass. It launched in September 2025. The Mac community’s reaction ranged from “interesting” to “what have you done to my computer.”

The problem? Transparency effects make text and controls hard to read on busy backgrounds. Sidebar selections sometimes vanish. One Mac blogger noted they “can’t even tell what’s a UI bug and what’s working as intended.” A Slack user reported that a yellow emoji positioned behind a Liquid Glass button made the button glow gold for no reason. Several months later, power users are still calling it a work in progress.

Now look at Ubuntu 24.04’s GNOME 46 desktop. It won’t make anyone’s jaw drop. It looks… sensible. Text is always legible. Buttons look like buttons. Sidebars are clearly sidebars. No background emoji has ever made anything glow unexpectedly.

Sometimes boring is beautiful. Right now, Ubuntu’s “boring” feels like a spa day after the visual chaos of Liquid Glass.

The Desktop Gap Has (Mostly) Closed

When I last used a Linux desktop around 2015, the experience was clearly behind macOS. Wi-Fi drivers were unreliable, Bluetooth was a coin flip, and the app ecosystem had canyon-sized gaps.

Ubuntu 24.04 LTS in 2026 is a different beast:

1. It just works now. Ubuntu 24.04.4 ships with Linux kernel 6.17 and Mesa 25.2. Wi-Fi works out of the box. Bluetooth is stable. Even NVIDIA GPUs — historically Linux’s nemesis — cooperate with drivers Ubuntu makes easy to install. The Flutter-based installer gets you from ISO to desktop in 15 minutes with full disk encryption.

2. It’s faster. On the same hardware, Ubuntu feels noticeably lighter than macOS Tahoe. My 8GB Mac Mini M1 runs Ubuntu with headroom to spare — the same machine that sometimes wheezes under Liquid Glass.

3. The apps are there. Firefox, Chrome, VS Code, Spotify, Slack, Discord, Blender, GIMP, OBS Studio, Steam — all available through the App Center, Flatpak, or Snap. Not every macOS app exists on Ubuntu, but for web browsing, document editing, development, media, and communication, the gap is essentially closed.

4. GNOME 46 is mature. The Activities overview, workspaces, Nautilus file manager, and touchpad gestures all feel polished and cohesive. It has its own identity now — no longer a macOS imitation, but a genuinely pleasant desktop in its own right.

Where macOS Still Wins

I’d be dishonest if I didn’t acknowledge these:

1. Ecosystem lock-in (the good kind). AirDrop, Handoff, Universal Clipboard, Sidecar, the new Phone app for relaying iPhone calls — if you own Apple everything, macOS ties it together in ways Ubuntu simply can’t match. This is Apple’s greatest feature and most effective trap, simultaneously.

2. Creative pro software. Final Cut Pro, Logic Pro, Adobe Creative Suite — they’re macOS-only, and if your livelihood depends on them, there’s no alternative.

3. Xcode. My personal dealbreaker. I need it for my iOS apps. If Apple ever released Xcode for Linux, I’d switch by Friday. They won’t, but a man can dream.

4. Apple Intelligence. Built-in AI writing tools, image generation, and Siri improvements. Ubuntu has nothing equivalent built-in, though running local AI through Ollama arguably gives you more privacy and control.

Where Ubuntu Wins

And these advantages are bigger than most people realize:

1. Customization. macOS lets you pick Light, Dark, or Tinted. Ubuntu lets you change *everything* — desktop environment, window manager, icons, fonts, behavior. Don’t like GNOME? Install KDE Plasma, Cinnamon, or Xfce on the same machine. Try moving your Dock to the top of the screen on macOS. I’ll wait.

2. Privacy. No telemetry by default. No ads. No account sign-in prompts. No features locked behind subscriptions. Ubuntu’s stance: your computer, your data, full stop.

3. Package management. `sudo apt install whatever-you-need`. Three words and you have it. Updating everything on the system? `sudo apt update && sudo apt upgrade`. APT is built into Ubuntu’s DNA. Homebrew on macOS is great, but it’s a third-party addition bolted onto an OS that wasn’t designed for it.

4. Dev environment parity. If you deploy to Linux servers (most web devs do), developing on Ubuntu means your dev environment matches production. Docker runs natively. No more “works on my Mac” debugging sessions.

5. Hardware longevity. Ubuntu runs on computers from 2010. macOS is about to abandon every Intel Mac ever made. For sustainability and budget-conscious users, this matters enormously.

The Verdict

In 2015, I’d have said Ubuntu couldn’t match macOS as a daily driver. In 2026, I’m saying something different: for most people — students, web developers, writers, casual users, small businesses — Ubuntu 24.04 LTS is not a compromise. It’s a genuinely excellent OS that costs nothing.

If you need Xcode, the Apple ecosystem, or creative pro software, stay on macOS. I am, for my development work. But if you just want a fast, secure, private desktop that doesn’t cost a thousand plus dollars before you even turn it on? Ubuntu isn’t just viable anymore. It might actually be the better daily experience right now — especially while macOS is going through its Liquid Glass identity crisis.

The $0 operating system is rivaling the premium experience. And in some ways, it’s winning.

— Jun

Why the Tesla Model Y L Is the Most Feature-Packed EV for Its Price in the Philippines

If you’re a long-time reader of TechSource, you know this site has mostly been about Linux, open-source software, and all things computing. But if you’ve been following our recent comeback, you also know we’ve expanded into covering the broader tech landscape — AI, smartwatches, crypto, and whatever else catches my persistently curious eye. Today, we’re parking (pun intended) in a topic that’s been occupying a significant amount of my brain space lately: electric vehicles. Specifically, the Tesla Model Y L, which just arrived in the Philippine market and which I am fully planning to order on April 1, 2026 — the first day online orders open in the Philippines.

A Quick Note on the Topic

I want to acknowledge upfront that talking about Tesla in 2026 comes with some unavoidable baggage. The brand’s CEO has become a polarizing figure, and I understand that many people’s feelings about Tesla the company are complicated. I’m not here to discuss politics. I’m here to discuss the technology. And purely on the merits of its technology, engineering, and value proposition, the Tesla Model Y L is the most feature-packed electric vehicle to land in the Philippine market at its price point. That’s what I want to talk about.

Full disclosure: this is not a sponsored post. Tesla Philippines doesn’t know I exist. I’m just a tech enthusiast from a little town in Bohol who has been dreaming about this car since before Tesla officially entered the Philippine market, and I want to share why.

The Elephant in the Gas Station

If there’s ever been a time to seriously consider going electric, it’s right now.

As I write this in late March 2026, global oil prices are in crisis. The US-Israeli war on Iran, which began on February 28, has effectively shut down the Strait of Hormuz — the narrow waterway through which roughly 20% of the world’s oil supply passes. Brent crude has surged over 40% since the conflict started, topping $100-$120 per barrel. The International Energy Agency has called it “the greatest global energy security challenge in history.

The ripple effects are being felt everywhere, including right here in the Philippines and across Southeast Asia. Thailand has implemented fuel rationing. Pakistan has told citizens to watch cricket games at home to conserve energy. Countries across Asia are hoarding and restricting fuel exports. Bloomberg estimates the oil shock could push global inflation significantly higher, with the US CPI for March already jumping to 3.4% year-on-year — and fuel prices are the main culprit. Analysts warn that if the Strait of Hormuz remains closed into mid-April, the world could lose up to 10 million barrels of oil per day, and prices could spike further — with some Wall Street analysts now floating the possibility of $200-per-barrel oil.

For Filipino motorists already stretched thin by living costs, the timing couldn’t be worse. Every peso increase at the pump hits harder when your daily commute is non-negotiable. And this isn’t a one-off spike — the geopolitical volatility that drives oil prices is structural and recurring. There was the Russia-Ukraine shock in 2022, and now the Iran crisis in 2026. The pattern is clear: dependence on fossil fuels means dependence on global conflicts over which we have zero control.

This is the context in which the Tesla Model Y L arrives in the Philippines. An electric vehicle powered by locally generated electricity — or better yet, by solar panels on your own roof — is immune to whatever happens in the Strait of Hormuz, the Persian Gulf, or any other geopolitical flashpoint. When you charge from the sun, no war, no cartel, and no sanctions can touch your fuel bill.

The math has never been more compelling.

What Is the Tesla Model Y L?

The Model Y L is essentially the bigger, three-row version of Tesla’s best-selling vehicle worldwide. The “L” stands for, well, long — it’s 186mm longer and 44mm taller than the standard Model Y, with a wheelbase stretched by 150mm. That extra space allows Tesla to fit a third row of seats, turning the Model Y from a five-seater compact crossover into a six-seater family hauler with captain’s chairs in the second row.

The Philippines is the third market outside China to receive the Model Y L, after Australia and Thailand. Tesla Philippines previewed it at their BGC Experience Center last week, and CarGuide.ph confirmed that online orders will be accepted starting April 1, 2026 at a starting indicative price of ₱2,849,000.

Here’s what you get for that price:

1. Powertrain: 

Dual electric motors with all-wheel drive, producing 378 kW (roughly 506 hp) and 590 Nm of torque. Zero to 100 km/h in 5.0 seconds flat. That’s faster than most sports sedans, and this is a six-seater family SUV.

2. Battery and Range: 

An 88.2 kWh nickel-manganese-cobalt battery pack with a claimed WLTP range of up to 681 kilometers on a single charge. Let me put that in context — Bohol is about 75 kilometers from end to end. You could drive across the entire island roughly nine times on a full charge. For the typical Filipino daily commute of 20-40 kilometers, you’re charging maybe once or twice a week.

3. Charging: 

DC fast charging support at up to 250 kW. At a Tesla Supercharger, you can add significant range in about 15 minutes. Home charging with a standard wall connector overnight is more than enough for daily use.

4. Towing: 

A braked towing capacity of 1,588 kg — plenty for most recreational needs.

5. Interior: 

Six seats with ventilated and reclining captain’s chairs in the second row, complete with power armrests that rise from the seat base. The third row gets its own air vents. The panoramic glass roof floods the cabin with natural light. And the entire vehicle is controlled through a massive 15.4-inch center touchscreen that handles everything from navigation to climate to entertainment.

The Tech That Makes It a Rolling Computer

This is where the Model Y L really separates itself from every other vehicle in its price range in the Philippines:

1. Autopilot and Full Self-Driving.

Every Model Y comes standard with Autopilot, which includes adaptive cruise control, lane keeping, and lane centering. But the real magic is the optional Full Self-Driving (Supervised) capability, which allows the vehicle to navigate city streets, handle intersections, make turns, navigate roundabouts, and enter/exit highways — all with the driver supervising. It uses a vision-based system with cameras mounted at the front, rear, left, and right of the vehicle, feeding data to Tesla’s onboard neural network computer. The system is constantly improving through over-the-air software updates, which means the car literally gets smarter over time. As FSD deployment expands globally, Tesla has said it will gradually make it available in select countries outside the US and Canada. Whether and when FSD Supervised will be fully functional in Philippine roads remains to be seen, but the hardware is there, ready and waiting.

2. Over-the-Air Updates. 

This is the feature that makes Tesla fundamentally different from every traditional automaker. Your car receives software updates wirelessly — just like your iPhone or your Mac. New features, performance improvements, bug fixes, and security patches arrive automatically. You go to bed with one car and wake up with a slightly better one. Traditional car manufacturers are still figuring out how to make their infotainment systems not freeze during Bluetooth pairing. Tesla is pushing neural network updates to your drivetrain.

3. The Touchscreen.

The 15.4-inch center display runs everything. Navigation with real-time traffic and Supercharger routing. A full web browser. Streaming services (Netflix, YouTube, Spotify). Arcade games (some playable with the steering wheel as a controller — when parked, obviously). Climate control with per-seat adjustments. Dashcam and Sentry Mode footage playback. It’s basically a giant tablet on wheels that also happens to take you places.

4. Sentry Mode and Dashcam. 

The vehicle’s cameras double as a 360-degree security system. Sentry Mode monitors your surroundings when parked and records any suspicious activity. The built-in dashcam continuously records your drives. In a country where road incidents and parking lot dings are facts of life, having a car that watches its own back is genuinely valuable.

5. The App. 

The Tesla mobile app lets you monitor your vehicle from anywhere — check charge status, precondition the cabin temperature before you get in (a lifesaver in Philippine heat), locate your car, lock/unlock remotely, and even summon the vehicle in a parking lot. I think the Tesla app is one of the best-designed automotive apps I’ve seen.

6. Regenerative Braking. 

Lift off the accelerator and the car slows down while feeding energy back to the battery. Most Tesla owners rarely use the brake pedal in everyday driving. It takes about a day to get used to, and then you never want to drive without it.

Why I’m Getting One (The Personal Reasons)

Beyond the tech specs, there are personal reasons why the Model Y L makes sense for me specifically:

1. My house is already solar-powered. 

This is the big one. I invested in a solar panel system for our home, and the idea of charging a car using energy from the sun — essentially driving for free — feels like the kind of future I’ve been writing about on this site for nearly two decades. The economics of EV ownership change dramatically when your electricity comes from your own roof. No more gas station runs. No more volatile fuel prices. No more watching the news about Strait of Hormuz closures and wondering how much your next fill-up will cost. Just clean energy from Bohol sunshine, which, as anyone who’s been here knows, we have in abundance. While my neighbors are anxiously checking oil price updates as the Iran war unfolds, I’ll be topping up my car from a star that’s been burning for 4.6 billion years and isn’t controlled by any cartel.

2. I’m impressed with Tesla’s ecosystem. 

We’ve been using Starlink at home, and I’ve been genuinely impressed by its reliability — consistent internet in a province where connectivity has historically been a struggle. That experience gave me confidence in the broader Tesla/SpaceX ecosystem’s ability to deliver technology that actually works in Philippine conditions. If Starlink can handle Bohol’s weather and geography, I’m optimistic about what a Tesla can do on our roads.

3. The Model Y L is the right size for a family.

The six-seat configuration with captain’s chairs in the second row is perfect. It’s spacious enough for family road trips, practical enough for daily driving, and the 2,539 liters of maximum cargo capacity means you’re not sacrificing storage for those extra seats.

4. It’s a dream car. 

I’ll be honest — I’ve wanted a Tesla since before they were officially available in the Philippines. The combination of cutting-edge technology, performance, and the simple elegance of an electric powertrain has always appealed to the tech nerd in me. The fact that it runs on software that gets better over time, that it has no traditional engine to maintain, that it’s essentially a computer on wheels — this is the car that makes sense for someone who has spent his entire adult life surrounded by technology.

The Honest Cons (Because This Is TechSource, Not a Press Release)

I wouldn’t be doing my job if I didn’t address the real challenges of owning a Tesla in the Philippines in 2026, especially outside Metro Manila.

1. The service center situation. 

Tesla Philippines currently operates one service center and it’s in Bonifacio Global City, Taguig — in Metro Manila. I live in Bohol. That’s an island in the Visayas, roughly 630 kilometers and a plane ride away. If my Model Y L needs anything beyond what Tesla’s mobile service can handle remotely, the car needs to go to Manila. That’s a significant concern, and I’d be lying if I said it doesn’t give me pause. Tesla also has two approved body shops, but again, both in Metro Manila. The hope is that as Tesla’s Philippine customer base grows, service centers will expand to major cities outside NCR. Tesla has already announced plans for Cebu Supercharger stations, and where Superchargers go, service infrastructure often follows. But for now, this is the biggest practical downside for provincial buyers.

2. Charging infrastructure in the Visayas. 

Tesla currently operates Supercharger stations exclusively in Metro Manila — at Uptown Mall, Shangri-La Plaza, SM Mall of Asia, and Opus Mall. They’ve announced plans for two Cebu locations plus stations in Clark, Baguio, Olongapo, and Taguig for 2026. But in the Visayas? The charging landscape is still in its early stages. Third-party EV chargers from ACMobility and others are slowly appearing in malls and commercial centers, but they’re sparse. For daily driving within Bohol, home charging from my solar setup will more than suffice — the 681 km range means I’d charge at home a couple of times a week at most. But road trips to Cebu, Dumaguete, or further afield will require careful planning. The Department of Energy has set a target of 7,000 EV charging stations nationwide by 2028, so the infrastructure is coming, but it’s not here yet.

3. Resale uncertainty. 

The Philippine EV market is still young. While Teslas hold their value well globally, the local resale market for EVs is uncharted territory. This is less of a concern if you plan to keep the car long-term (which I do), but it’s worth noting.

4. The learning curve.

Everything is controlled through the touchscreen. There are no traditional buttons for the climate, the headlights, or the mirrors. The gear selector is on the screen. If you’re coming from a traditional car (like I am), there’s an adjustment period. Most owners say it takes about a week before the Tesla way feels natural and everything else feels antiquated.

5. No Apple CarPlay or Android Auto. 

Tesla uses its own infotainment system and doesn’t support Apple CarPlay or Android Auto. The built-in navigation, music streaming, and communication features are good, but if you’re deeply attached to CarPlay, this will annoy you.

 

How It Compares at Its Price Point

At ₱2,849,000, the Model Y L competes with vehicles like the BYD Tang EV (₱3,321,000) and various hybrid and ICE SUVs in the ₱2.5-3.5M range. What sets the Tesla apart is the sheer density of technology packed into the price.

You’re getting dual-motor AWD, 506 hp, a 681 km range, Autopilot, over-the-air updates, Sentry Mode, a 15.4-inch touchscreen, six seats with ventilated captain’s chairs, a 5-star safety rating, access to the Tesla Supercharger network, and a vehicle that improves through software updates for years after purchase — all for under ₱3 million. You’d need to spend significantly more with traditional luxury brands to get even a fraction of these features.

The 4-year or 80,000 km bumper-to-bumper warranty and 8-year or 160,000 km battery warranty also provide peace of mind. And with zero import duties on EVs in the Philippines through 2028 thanks to Executive Order No. 62, the pricing is as good as it’s going to get.

April 1 Can’t Come Soon Enough

I’ve already bookmarked the Tesla Philippines website. My finger is ready to click “Order” the moment it goes live on April 1, 2026 — two days from today as of this writing.

Will it be nerve-wracking to be an early Tesla owner in the Visayas, with the nearest service center a plane ride away and Superchargers still confined to Metro Manila? Absolutely. But being an early adopter has always been part of who I am. I started writing about Linux when most Filipinos had never heard of it. I ran a Bitcoin node on a Raspberry Pi when crypto was still considered fake money. I built a local AI hub on an old Mac Mini when most people thought you needed a supercomputer to run AI.

Getting a Tesla in Bohol or in the Philippines in 2026 feels like the natural next chapter of that same story — embracing technology that most people think is “not ready yet” for places like ours, and proving that it absolutely is.

The sun is shining on my solar panels. The Starlink dish is humming on my roof. And soon, if all goes according to plan, a Tesla Model Y L will be sitting in my driveway, charging from that same sunshine, ready to silently drive me around the island I call home.

The future isn’t coming. It’s already here. It just needs to be plugged in.

— Jun

Linux Won, and Nobody Noticed

The tech industry has failed to properly acknowledge this for years: Linux won. Not "Linux is doing fine." Not "Linux is making progress." Not "maybe next year will be the year of the Linux desktop." No. Linux won. Decisively. Overwhelmingly. In nearly every category of computing that actually matters, Linux is the dominant operating system on the planet — and it happened  quietly that most people, including many who use it every single day, have absolutely no idea.

I've been writing about Linux on this site for a long time now. I've reviewed dozens of distros, compared desktop environments, hosted Distrowar battles between distributions, and written passionate articles about why Linux deserved more attention. For nearly two decades, the narrative around Linux has been the same: "it's great, but it'll never go mainstream." That narrative is wrong. It's been wrong for years. And it's time someone said it clearly.

The Scoreboard:

Let's look at the actual numbers, because the scoreboard tells a story that the "Year of the Linux Desktop" jokes have been drowning out.

1. Supercomputers: Linux owns 100%. Every single one of the world's 500 fastest supercomputers runs Linux. Not 99%. Not most of them. All of them. This has been the case since 2017, and there's no sign of it changing. The last non-Linux system dropped off the TOP500 list years ago. When humanity needs raw computational power — for climate modeling, genomic research, nuclear simulations, AI training — it runs Linux. Full stop.

2. Servers: Linux dominates. Linux commands roughly 63% of the server operating system market globally. Over 96% of the top one million websites run on Linux. The web servers you interact with every day — Nginx, Apache — run almost exclusively on Linux. When you load a webpage, stream a video, check your email, or buy something online, the odds are overwhelming that a Linux server handled your request.

3. Cloud: Linux is the foundation. About 49% of all global cloud workloads run on Linux. AWS, Azure, and Google Cloud — the three pillars of cloud computing — all use Linux as their foundational operating system. Over 90% of public cloud workloads operate on Linux. The entire cloud revolution was built on top of a free, open-source kernel.

4. Containers: Linux is the only game in town. Docker, the dominant container platform, is used by over 108,000 companies. Kubernetes, the container orchestration standard, holds 92% market share with 5.6 million developers using it. Nearly all of this runs on Linux. The modern DevOps pipeline — the infrastructure that builds, tests, and deploys the software the world depends on — is a Linux pipeline.

5. Mobile: Linux is in your pocket. Android, which is built on the Linux kernel, powers roughly 71% of all smartphones globally. That's approximately 3.9 billion active devices. Every time someone says "Linux has no users," they're ignoring the billions of people carrying a Linux-based operating system in their pocket right now. It's the single most widely deployed operating system family on Earth by device count, and it's not even close.

6. Embedded systems and IoT: Linux is everywhere you don't see it. Over 44-46% of embedded systems run Linux. Your smart TV probably runs Linux. Your router almost certainly runs Linux. Many cars, medical devices, industrial controllers, and smart home devices run Linux. The invisible infrastructure of modern life is quietly humming along on a kernel that Linus Torvalds started as a hobby project in 1991.

7. Developer adoption: Linux is the default. About 78.5% of developers worldwide use Linux as a primary or secondary operating system. Among cloud-native developers, that number jumps to over 90%. The people building the future of software overwhelmingly choose Linux as their platform.

Now compare all of that to the one metric everyone obsesses over: desktop market share. Linux sits at roughly 4-5% on the desktop globally, around 5% in the United States. And because of this single number, the prevailing narrative remains "Linux hasn't made it."

That's like saying a basketball team lost the game because they didn't win the coin toss. The desktop is one court. Linux is winning the entire league.

How We Got the Narrative Wrong

The "Year of the Linux Desktop" has been a meme for over two decades. Every year, someone declares it, and every year, Linux's desktop share barely moves. The punchline writes itself.

But here's the thing the meme got wrong: it defined Linux's success by the wrong metric. Desktop computing is no longer the center of the tech universe. It hasn't been for years. Mobile, cloud, servers, IoT, containers, supercomputing — these are where computing lives now. And Linux owns all of them.

The fixation on the desktop is a relic of the early 2000s, when the desktop was the primary way most people interacted with computers. That world doesn't exist anymore. Today, most people's primary computing device is their smartphone (Linux, via Android). The software they use is served from the cloud (Linux). The websites they visit are hosted on servers (Linux). The apps are built and deployed using containers (Linux). The AI models they query are trained on supercomputers (Linux).

The average person in 2026 interacts with Linux dozens of times a day without knowing it. They just don't see a penguin on their screen, so they assume Linux isn't relevant.

I fell into this trap too. Back in 2011, I wrote an article on this very site titled "Why the Linux Desktop is Still Not #Winning." I argued that Linux's lack of focus was holding it back on the desktop. I wasn't wrong about the desktop, but I was wrong about what winning looked like. Linux didn't need to conquer the desktop to win. It conquered everything else.

The Quiet Victories Nobody Talks About

Some of Linux's most important wins happened so gradually that they never got a headline.

Linux ate the corporate data center. Red Hat Enterprise Linux holds 43.1% of the enterprise Linux server market. Over 90% of Fortune 500 companies use Red Hat products. The business world runs on Linux, and most employees have no idea. They sit at their Windows desktops, interacting with business applications that are served, processed, and stored on Linux infrastructure behind the scenes.

Linux runs the financial system. Stock exchanges, banking systems, payment processors — the financial infrastructure that moves trillions of dollars every day runs predominantly on Linux. The New York Stock Exchange switched to Linux years ago. When you swipe your credit card, the transaction almost certainly touches a Linux server somewhere in the chain.

Linux powers space exploration. NASA's Mars rovers and helicopters run Linux. SpaceX uses Linux for its flight software. The International Space Station has Linux computers on board. When humanity reaches beyond Earth, it does so on the back of open-source software.

Linux runs AI. The AI revolution — ChatGPT, Claude, Gemini, Stable Diffusion, all of it — runs on Linux. The NVIDIA GPUs that train these models run Linux drivers. The data centers that house them run Linux. The containers that deploy them run Linux. Every time someone marvels at what artificial intelligence can do in 2026, they're marveling at something that Linux made possible.

Governments are switching. Germany's Schleswig-Holstein replaced Microsoft tools with Linux across all public offices. France runs over 103,000 government computers on a custom Ubuntu distribution. Denmark is transitioning from Microsoft to open source. The EU is considering a standardized "EU-Linux." Switzerland mandated that government-developed software be released as open source. These aren't experiments. These are institutional commitments.

The Desktop Is Finally Moving Too

And here's the twist: even the desktop — that one stubborn metric — is finally showing real momentum.

Linux desktop market share hit 4.7% globally in 2025, representing a 70% increase from 2.76% in just three years. The United States crossed 5% for the first time. India leads major economies at over 16%. These are still small numbers compared to Windows, but the trajectory is unmistakable.

Several factors are converging. Windows 10 reached end of life in October 2025, and Windows 11's strict hardware requirements (TPM 2.0, specific CPU generations) left millions of perfectly functional PCs unable to upgrade. Many of those users are discovering that Linux can give their hardware a second life at zero cost. Valve's Steam Deck — a handheld gaming device running SteamOS (Arch Linux) — sold millions of units and proved Linux could be a consumer gaming platform. Proton, Valve's compatibility layer, now makes roughly 90% of Windows games playable on Linux. At CES 2026, Lenovo announced a handheld powered by SteamOS. The taboo of Linux gaming is officially over.

Ubuntu 26.04 LTS drops next month with GNOME 50, Wayland-only graphics (X11 is finally gone from core components), Rust-based system utilities, and improved NVIDIA performance. It's shaping up to be the most polished Ubuntu release ever — arriving at exactly the moment when the most people are looking for a Windows alternative.

The desktop isn't the finish line. But if it were, Linux is finally within sight of it.

What Winning Actually Looks Like

Linux didn't win the way anyone expected. There was no dramatic moment where Ubuntu overtook Windows on the desktop. No press conference. No champagne. Linux won the way open source always wins — gradually, relentlessly, by being better at the things that matter most to the people building the future.

It won because it was free, and startups with no budget could build their entire infrastructure on it. It won because it was customizable, and engineers could tune it for everything from a tiny IoT sensor to the world's fastest supercomputer. It won because it was open, and thousands of companies and millions of developers could contribute to and benefit from the same shared foundation. It won because it was reliable, and system administrators could trust it to run for years without a reboot. It won because it was secure, and organizations handling sensitive data needed something they could audit and verify.

It won not despite being open source, but because of it.

Why This Matters

I'm not writing this to gloat (okay, maybe a little). I'm writing this because I think there's an important lesson in how Linux won — a lesson that applies far beyond operating systems.

When I started TechSource in 2007, advocating for Linux and free software felt like shouting into the void. The dominant narrative was that open source couldn't compete with well-funded proprietary alternatives. That free software was inferior. That you get what you pay for, and if you pay nothing, you get nothing.

Linux proved all of that wrong. And it did it not by being a charity project, but by being genuinely, measurably, demonstrably better for the use cases that mattered. Corporations didn't adopt Linux because they loved the philosophy of open source. They adopted it because it was the best tool for the job. The philosophy was a bonus.

Today, I see the same dynamic playing out with open-source AI. Models like Llama, Mistral, Qwen, and DeepSeek are challenging proprietary AI systems the same way Linux challenged proprietary operating systems. The tools like Ollama and Open WebUI that make local AI accessible are following the same playbook that Ubuntu followed to make Linux accessible. The pattern is the same. The lesson is the same. Openness wins, eventually, because it enables the most people to build on the same foundation.

So the Next Time Someone Asks...

The next time someone asks, "when is the year of the Linux desktop?" — gently remind them that they're asking the wrong question.

The year of Linux already happened. It's been happening for over a decade. It happened when Linux took over supercomputing. It happened when the cloud was built on Linux. It happened when Android put Linux in 3.9 billion pockets. It happened when containers and Kubernetes became the standard way to deploy software. It happened when every major AI model was trained on Linux infrastructure.

Linux didn't win the desktop war. It won the computing war. And it did it so quietly that most of the world still doesn't know.

But now you do.

What do you think? Is the desktop still the metric that matters, or has Linux already won where it counts? 

— Jun

How I Built a Local AI Hub Using Free and Open Source Software on My Old Mac Mini

I’m going to tell you something that would have sounded absolutely insane five years ago: I’m running artificial intelligence on a computer the size of a lunch box, it works offline, my data never leaves my house, and it costs me nothing beyond the electricity to keep it running.

No monthly subscription. No API fees. No sending my private documents to some server farm in Virginia. Just me, a Mac Mini M1, and a free and open-source software called Ollama that has quietly become one of the most important pieces of software I’ve used — and I say that as someone who has been reviewing software on this site since 2007.

If you’ve been curious about running AI locally but thought you needed a $5,000 GPU rig and a computer science degree, this post is for you. I’m going to walk you through exactly how I set up my local AI hub, what I use it for, and why I think every tech enthusiast should consider doing the same.

Why Local AI? Why Now?

Let me give you some context. Like most people, I’ve been using cloud-based AI tools — ChatGPT, Claude, Gemini — and they’re incredible. But there are situations where sending your data to a cloud service isn’t ideal.

When I’m working on business documents for my offline ventures, I don’t necessarily want those financial projections living on someone else’s server. When I’m brainstorming ideas for my apps, I’d rather keep those early concepts private. When I’m processing data for my web projects, I want the flexibility to run queries without worrying about rate limits, usage caps, or monthly bills that scale with every prompt.

The privacy argument alone is compelling, but it’s not the only reason. Local AI is also fast — there’s no network latency, no waiting for servers, no “we’re experiencing high demand” messages. It works offline, which means I can use it on a plane, in a coffee shop with terrible Wi-Fi, or during one of those delightful Philippine internet outages that build character.

And perhaps most importantly for a guy who has been writing about free and open-source software for nearly two decades: local AI puts the power back in your hands. You own the hardware, you own the model weights, and there are no terms of service to violate. That’s the open-source philosophy I’ve been preaching since my Linux days, applied to the most transformative technology of our generation.

What You Need (Less Than You Think)

Here’s my setup:

Hardware: Mac Mini M1 (8GB Unified Memory)

That’s it. That’s the hardware. No dedicated GPU. No server rack. No liquid cooling. An old Mac Mini M1 — the base model with just 8GB of RAM — that I bought a few years ago and that sits quietly on my living room table consuming roughly the same power as a light bulb.

Now, let me be upfront: 8GB is the bare minimum for local AI. It’s not ideal. After macOS takes its share of memory (roughly 3-4GB for the operating system and background processes), you’re left with about 4-5GB of usable space for AI models. That means the popular 7B and 8B parameter models that most guides recommend are either too tight to run comfortably or will cause constant memory pressure and slowdowns on my machine. I learned this the hard way after watching my Mac Mini struggle and swap memory like it was reliving its Intel days.

But here’s the thing — you don’t need the biggest models to get genuinely useful results. The smaller models, in the 1B to 3.8B parameter range, run beautifully on 8GB machines. They’re fast, responsive, and for many everyday tasks, surprisingly capable. Are they as good as GPT-4 or Claude? Not even close. But for quick drafts, summarization, code snippets, brainstorming, and general Q&A, they get the job done without sending a single byte of your data to the cloud.

The secret sauce that makes even my base model Mac Mini viable is Apple Silicon’s unified memory architecture. Unlike traditional PCs where the CPU and GPU have separate memory pools, the M1’s unified memory means the GPU can directly access whatever RAM is available for AI inference. Even with just 8GB, the M1’s efficiency means small models can generate tokens at 30-60+ tokens per second — fast enough that responses feel nearly instant.

Could you do this on a Windows PC or a Linux machine? Absolutely. If you have a desktop with an NVIDIA GPU (even a used RTX 3060 for around $150), you’d get excellent performance with even bigger models. But for Mac users with older Apple Silicon hardware gathering dust, Ollama gives that machine a second life.

Minimum specs to get started:

Any Apple Silicon Mac (M1 or newer) with 8GB of RAM can run small models (1B-3.8B parameters). Think of these as quick, lightweight assistants good for summarization, simple coding help, and general Q&A. With 16GB, things get significantly better — you can comfortably run 7B-8B models at good speed and even some 14B models. With 32GB or more, you’re in serious territory — running models that rival cloud-based services for many tasks.

On the PC side, 16GB of system RAM plus a GPU with at least 8GB of VRAM is the sweet spot. More VRAM means bigger, better models.

Installing Ollama: Easier Than Installing Most Apps

Ollama is the foundation of my local AI setup. It’s a free and open-source tool that handles downloading, managing, and running large language models with absurd simplicity. If you can type a command into a terminal, you can run local AI.

Step 1: Install Ollama

On Mac, you have two options. The easiest is to download the app directly from ollama.com. (). Download the DMG, drag it to Applications, and launch. Done.

If you prefer Homebrew (and if you’re a developer, you probably do):

    brew install ollama

On Linux:

    curl -fsSL https://ollama.com/install.sh | sh

On Windows, simply download the installer from the Ollama website.

That’s the entire installation. No Python environment management. No dependency hell. No CUDA driver nightmares. It just works.

Step 2: Pull Your First Model

Open your terminal and type:

    ollama pull llama3.2:3b

This downloads Meta’s Llama 3.2 3B model — one of the best small open-source language models available and the sweet spot for 8GB machines. It’s about 2GB on disk and runs comfortably without choking your system.

If you want something even lighter to start with:

    ollama pull phi4-mini

Microsoft’s Phi-4 Mini (3.8B parameters) is another excellent choice for 8GB systems — strong instruction following and surprisingly good at code for its size.

Step 3: Start Chatting

    ollama run llama3.2:3b

That’s it. You now have a local AI assistant running entirely on your machine. Ask it questions, have it summarize text, help with code, draft emails — whatever you need. Type your prompt, get a response. No account required. No internet required after the initial download.

The first time I ran this and got a coherent, helpful response from a model running entirely on my Mac Mini, I had the same feeling I had back in 2007 when I first booted Ubuntu and realized an entire operating system could be free. That feeling of “wait, this actually works, and it’s free?” — that’s the open-source magic I’ve been chasing for nearly 20 years.

The Models I Actually Use on 8GB

Ollama gives you access to a growing library of models. Here are the ones that work well on my 8GB Mac Mini and what I use them for:

1. Llama 3.2 3B — My go-to daily driver. This is the model I reach for most often. For a 3B model, the quality is genuinely impressive — it handles summarization, drafting, general Q&A, and brainstorming surprisingly well. On my M1, it runs at roughly 30-50 tokens per second, which means responses feel nearly instant. It’s the perfect balance of quality and speed for an 8GB machine.

2. Phi-4 Mini (3.8B) — My coding companion. Microsoft’s Phi-4 Mini punches well above its weight for code generation and technical tasks. When I’m working on my iOS apps or web projects and need a quick SwiftUI snippet, JSON formatting help, or a debugging nudge, this model delivers at around 15-20 tokens per second. It won’t replace Claude for complex architecture decisions, but for quick code help during focused development sessions, it’s remarkably useful.

3. Gemma 2B — My speedster for trivial tasks. Google’s smallest Gemma model is ultra-lightweight and blazing fast. I use it for simple reformatting, quick translations, and tasks where I just need a fast answer and don’t care about nuance. Think of it as the Puppy Linux of language models — tiny, fast, and gets the basics done.

4. Llama 3.2 1B — My offline emergency model. At just around 1.3GB, this model loads almost instantly and runs so fast it feels like autocomplete. The quality is basic, but when I need something working on minimal resources or want to run alongside other applications without memory pressure, it’s there.

Here’s the honest truth about running local AI on 8GB: you’re operating within constraints. Multi-turn conversations get noticeably weaker after several back-and-forth exchanges because the limited memory means shorter context windows. Complex reasoning tasks will sometimes produce mediocre results. And you’ll occasionally notice responses that are clearly “smaller model quality” compared to what you get from cloud services.

But for single-turn tasks — summarize this, draft that, reformat this JSON, explain this concept, help me with this code snippet — these small models are fast, private, and genuinely useful. It’s like having a competent junior assistant who works for free and never sleeps.

To switch between models, I just run a different command. Different models for different jobs — just like how I used to keep different Linux distros for different purposes back in my distro-hopping days.

Adding a Proper Interface: Open WebUI

Running Ollama from the terminal is fine for quick tasks, but for extended sessions, it gets clunky. You lose chat history, you can’t easily compare models, and scrolling through terminal output isn’t exactly a delightful user experience.

Enter Open WebUI — a free, open-source web interface that connects to Ollama and gives you a ChatGPT-like experience running entirely on your local machine.

If you have Docker installed, the setup is one command:

docker run -d -p 3000:8080 \

  -v open-webui:/app/backend/data \

  --name open-webui --restart always \

  ghcr.io/open-webui/open-webui:ollama

Open your browser, go to `http://localhost:3000`, create an account (this is local — nobody else sees it), and you’re in. Every model you’ve pulled with Ollama automatically appears in the interface.

Open WebUI is where the magic really happens. You get persistent chat history so you can pick up conversations where you left off. You can switch models mid-conversation to compare outputs. There are system prompt templates, temperature controls, and per-chat configuration settings. You can upload documents and use RAG (Retrieval Augmented Generation) to ask questions about your own files — PDFs, text documents, code files. It even supports web search integration, image generation, and voice input.

The interface looks and feels remarkably similar to ChatGPT, except everything is running on your own hardware. No cloud. No subscription. No data leaving your network.

I access Open WebUI from my Apple devices like my MacBook, my iPhone, and my iPad — all pointing to the Mac Mini sitting quietly on my living room table. It’s like having a private ChatGPT server for my household.

My Actual Workflows

Let me get specific about how I use this setup in real life, because “run AI locally” sounds cool in theory but means nothing without practical application.

1. For my blog (this site). When I’m researching topics for TechSource, I’ll dump my raw notes into a chat, ask the local model to identify the most interesting angles, suggest outlines, or flag gaps in my research. The model doesn’t write the posts for me — my writing voice is my own — but it’s an incredibly useful brainstorming partner.

2. For my iOS apps. I use Phi-4 Mini for quick SwiftUI help, JSON formatting, and debugging. Having a coding assistant that responds in under a second with no internet dependency is genuinely useful during focused development sessions.

3. For my offline businesses. I process business documents, draft communications, and analyze data without any of that information touching a third-party server. This is the use case where local AI’s privacy advantage matters most.

4. For website automation. I’ve built an automated pipeline that scrapes information from various sources and publishes curated content to my niche site. Ollama plays a role in processing and formatting that data. Having this run locally means the pipeline works even if my internet connection is spotty.

5. For learning. I feed technical articles, documentation, and research papers into the RAG system and then have conversations with the content. It’s like having a study partner who has perfect recall of everything you’ve uploaded.

How Does Local AI on 8GB Compare to ChatGPT and Claude?

I’m going to be honest with you, because that’s what TechSource has always been about.

On an 8GB machine running 3B models, local AI handles roughly 60-70% of the simple tasks I’d otherwise use cloud AI for. Summarization, quick drafts, code snippets, reformatting, basic Q&A — the small models get these done fast and privately.

For the remaining 30-40% — complex multi-step reasoning, nuanced creative writing, deep code architecture analysis, long conversations that require extensive context, and tasks requiring broad world knowledge — cloud models like Claude and GPT-4 are in a completely different league. There’s no sugarcoating this. My 3B model running locally isn’t competing with a 400B+ parameter model running on a data center full of A100 GPUs. That would be like comparing my Raspberry Pi to a supercomputer.

But that’s not the point. My approach is hybrid: local for privacy-sensitive work, quick tasks, and offline use. Cloud for complex, high-stakes tasks where quality matters more than privacy. The two complement each other perfectly. And if I ever upgrade to a Mac with 16GB or more RAM, those 7B-8B models become available and the quality gap narrows significantly.

What This Costs

Let’s do the math, because this is one of my favorite parts.

My setup costs:

Mac Mini M1 8GB (already owned and started to gather dust in my drawer): $0 additional cost. If buying used today, base M1 Mac Minis go for roughly $250-350 on resale markets — they’ve depreciated significantly, which makes them incredible value for a dedicated local AI server.

Ollama: Free, open-source.

Open WebUI: Free, open-source.

All AI models: Free, open-source.

Electricity: My Mac Mini draws about 20-39 watts during AI inference. Running it 8 hours a day costs roughly $2-3 per month in electricity.

Total monthly cost: About $3.

For comparison, ChatGPT Plus is $20/month. Claude Pro is $20/month. Running API calls at scale can easily cost $50-100+ per month depending on usage.

Even with the limitations of 8GB, my local setup handles enough daily tasks to reduce my reliance on paid subscriptions. Over a year, that adds up to meaningful savings — while giving me unlimited usage, complete privacy, and offline capability.

Tips I’ve Learned the Hard Way

After months of running this setup daily on constrained hardware, here are some practical lessons:

1. RAM is king. No, seriously.  On 8GB, every megabyte counts. Close unnecessary applications before running models. Safari with 20 tabs open and Xcode running simultaneously will leave almost nothing for Ollama. I’ve learned to treat my AI sessions like focused work blocks — close everything else, then chat.

2. Smaller models, faster results.  Don’t try to squeeze a 7B model onto an 8GB machine. I tried. It technically loads, but the constant memory swapping makes it painfully slow and the system becomes unusable for anything else. Stick to 3B and under for a smooth experience. A fast 3B model that responds instantly is infinitely more useful than a struggling 7B model that takes 10 seconds per response while your fans sound like a jet engine.

3. The 60-70% rule. Your model file should be no more than 60-70% of your total available memory (after macOS takes its share). On 8GB, that means model files of about 2-3GB maximum. This leaves enough room for the operating system, the context window (KV cache), and Ollama’s overhead.

4. Set Ollama as a network service. By default, Ollama only accepts connections from the local machine. If you want other devices on your network to access it (like I do with my MacBook and iPad), set the environment variable `OLLAMA_HOST=0.0.0.0` to allow connections from your local network. Just don’t expose it to the internet without authentication.

5. Different models for different jobs. I keep three to four small models installed and use them contextually. Phi-4 Mini for code, Llama 3.2 3B for general tasks, and Gemma 2B for quick throwaway queries. Specialization matters, even at the small model tier.

6. Keep an eye on model updates. The open-source AI community moves incredibly fast. Small models are improving at a staggering rate — the best 3B model today is dramatically better than the best 3B model from even six months ago. Check Ollama’s library periodically for new models. Pulling an update is just `ollama pull model-name`.

7. Plan your upgrade path. If local AI clicks for you (and I think it will), the single best upgrade you can make is more RAM. A used Mac Mini M1 with 16GB runs 7B-8B models comfortably and the quality jump from 3B to 8B is enormous. Consider it the best investment in your local AI future.

The Bigger Picture: This Is the Open-Source Revolution, Again

I started this site in 2007 writing about Linux because I believed free and open-source software could change the world. It did — Linux now powers 100% of the world’s top 500 supercomputers, 77% of web servers, and roughly half of all cloud workloads.

Now I’m watching the same thing happen with AI. Open-source models like Llama, Mistral, Qwen, Phi, Gemma, and DeepSeek are making AI accessible to anyone with a decent computer. Tools like Ollama and Open WebUI are making it easy. The barriers are falling fast.

A few years ago, running a useful AI model required cloud infrastructure and enterprise budgets. Today, you can do it on an old Mac Mini with 8GB of RAM that costs less than a pair of sneakers on the secondhand market. That trajectory reminds me of the early days of Linux, when something that was once the domain of server rooms gradually became something anyone could run on their desktop.

The fact that I can run a functional AI assistant on the most basic Apple Silicon Mac — the cheapest, lowest-spec model they ever made with an M1 chip — tells you everything about where this technology is headed. If this is what’s possible on 8GB today, imagine what the next generation of small models will do on the same hardware a year from now.

If you’ve been reading TechSource since the Ubuntu days, you already understand why this matters. The same principles that made open-source software transformative — transparency, control, community, freedom — are now being applied to artificial intelligence. And just like with Linux, you don’t need anyone’s permission to get started.

Pull up a terminal. Install Ollama. Run your first model. Welcome to the revolution. It’s local, it’s private, it’s free, and it could talk to your Linux-powered robot soon :) 

For those of you who are curious, below is a photo of my old Mac Mini (named Murdoc) lying on my living room table, looking like a metal brick that does nothing:

Mac Mini (Murdoc)

— Jun