A few days ago, I got a payment reminder from my bank for my ChatGPT subscription. These renewal notifications have become a routine. For nearly three years, since May 2023, it had quietly slipped into the background and I’d stopped thinking about it altogether. And that feeling of “whatever” was precisely the nudge I needed to start thinking about it more carefully.

Your data is a treasure trove, and the conversations we have in those LLM chat interfaces linger somewhere on their servers. It’s a good principle to apply a degree of “zero trust” to anything we type online.

This isn’t about wearing a tinfoil hat, but about being more mindful. Which begged a question: am I using this tool for exploring private thoughts, or am I using it to refine something meant for public view, like this blog post?

A lot’s changed since 2023. The first open-source model I used was Meta’s Llama, but now the field’s crowded with capable alternatives. The basic tasks I rely on an LLM for, checking my grammar, searching the web, rephrasing a sentence, have become commoditised.

This sparks a question. What’s the premium you’re really paying for?

We all know the old saying: when you use a tool for free, you’re most likely the product. There’s no such thing as truly “free”, after all. So perhaps the real premium isn’t just about better features. It’s about privacy.

But the real push to reconsider came from something else entirely.

Recently, OpenAI signed a contract with the US government. This came after Anthropic was blacklisted for refusing Pentagon demands to remove its safeguards. They walked part of it back after a backlash, which is something. But the event was a stark reminder. It highlighted a question I can’t avoid asking: whose privacy is truly being protected?

Philosophically, I’ve realised that convenience shouldn’t come at the cost of principle. Privacy isn’t a conditional right. It shouldn’t hinge on your nationality or jurisdiction. If a company pledges to “protect the rights of US citizens”, that’s a red flag for everyone else. Where does that leave me? Or any user outside that jurisdiction?

This is the core of it: I want my private explorations to remain just that—private. Using an LLM as a production tool to polish public content is one thing. Trusting a centralised, commercially-driven company with my unformed thoughts and confidential chats is another.

So yes, I’m voting with my wallet. It’s a small decision, really. Just one subscription. But these choices add up. They quietly shape the digital world we’re all building.

This isn’t about being anti-American or anti-government. It’s about supporting technologies whose alignment is transparently with the individual user, not the state. It’s about wanting my AI assistant to feel like a thoughtful collaborator, not a potential informant.

So, with a quiet but firm conviction, I’m letting it go. My daily needs can be met elsewhere, and my principles feel better served.

Here I am, sitting in bed. It’s quiet, early Friday morning, just before setting off on holiday back to Latvia. I’m looking forward to enjoying the snow, the nature, and a bit of time outdoors.

Anyway, back to the topic. Something is keeping me up this early on a day when I’d usually still be asleep, so I thought I’d write it down.

I’ve been quietly observing how LLMs have changed things since they first came out publicly. People close to this field generally understand how they work and how they should be used. However, I keep seeing cases of misuse. That made me stop and reflect.

I don’t think we explain clearly enough what LLMs actually are. Vague lines like “this might hallucinate” or “it might get things wrong” aren’t strong explanations and don’t do them justice. These systems are human-made results of decades of research. During my time at university, I learned about and built a Markov chain text generator. It was nowhere near the sophistication of modern LLMs, but it still gave me a sense of what it feels like to see a machine produce readable text. Even Andrej Karpathy’s microgpt project is just a single file, around 200 lines of code, with no external dependencies, yet it can train and run a small GPT. That said, thinking of LLMs as all-knowing entities straight out of science fiction is simply wrong. LLMs aren’t unique. They’re built on longstanding concepts, repackaged for a competitive landscape.

They’re more like a shattered mirror. Thousands of tiny fragments, with quite a few pieces missing. You can sort of see your reflection, but it isn’t really you. LLMs are trained on huge datasets and contain billions of parameters, which already implies limits. They are inherently incomplete and bound by physical constraints. They’re trained on past data yet respond in real time, which can blur our perception of what is real and what isn’t. Almost like an illusion we want to believe in.

Sometimes I find myself pondering how these massive datasets were sourced, and the ethics behind that process, since it has huge implications for the results an LLM can produce. For example, some reporting, such as a piece on TorrentFreak about alleged access to large collections of pirated books for AI training, raises important questions about consent and ownership. Biases are baked in, knowingly or not. Still, it’s something I keep coming back to.

How do we measure their effectiveness? Through benchmarks designed by humans. But those are relative too, because as models improve, so do the tests. Nothing is constant. Today even the best models might score below 40% on something like Humanity’s Last Exam. Tomorrow they might not. But why does that even matter?

LLMs mimic humanity, but they are not human and never will be. That’s why I find both fascinating and unsettling the level of obsession around AI. It can feel like a kind of collective hype. In the end, though, it’s still technology: disruptive, yes, but still just a tool. People adopt tools and move on.

Much of the confusion comes from not really understanding what these systems are, how they work, or how they should be used. Like choosing the right tool for a job, you wouldn’t use a fork to hammer in a nail.

And once you see it clearly, it stops looking like magic and starts looking like a tool.

Recently, I was recommended Flip 7 - a fast, tense, and brilliant card game that’s also kid-friendly. As the father of a five-year-old, I can safely say we’ve had a lot of joy out of it.

Without going too deeply into the mechanics, the goal is simple: score the most points, without letting scorekeeping slow the game down. After using the official Flip 7 score tracker a few times, it became clear that while it works, it feels rough around the edges, more like a digital version of a paper scorepad than a companion to the game.

So that’s why I decided to build something better.

Limitations of the original score app

The official tracker covers the basics, but it comes with a few limitations that make live games harder than they need to be:

• Scoring is handled through a single round-total input per player.
• There’s no real validation or structure, so you can enter pretty much any number, up to 999.
• Player management is minimal, just initials.
• The interface feels more like a form than part of the game.

In short, it works.
It just doesn’t really help.
You might as well use pen and paper.

A score tracker designed for the table

I built this Flip 7 score tracker to feel like a natural extension of the game itself. Instead of entering a round total, you track the cards each player actually plays:

• Tap number cards and modifiers as they appear (or after the round finishes).
• Flip 7 is detected automatically and the +15 bonus applied for you.
• End a player’s round with Stayed, Freeze, or Bust.
• Once a round ends, that player’s score is locked and clearly marked.

Scoring becomes faster, clearer, and much harder to mess up mid-game.

Designed for speed, clarity, and everyone at the table

I didn’t want a score tracker that felt like admin. Game night shouldn’t pause while someone squints at numbers or double-checks the maths.

So I built it so you never type scores at all. You play the cards, tap what happened, and the app handles the calculations in the background. The focus stays on the table, not the screen.

A lot of the design ideas came from actually playing the game:

• I kept the layout clean and visual so even younger players can follow along and help keep track, mine absolutely loves being “in charge” of the scores.
• Players are automatically sorted by score between rounds, so you instantly see who’s leading without scanning a list.
• I stripped away unnecessary numbers. You don’t need a wall of stats mid-game, just the totals that matter.

It’s built to be glanced at, not studied.

More flexibility, less friction.

While I was building it, I also added a few things I always wished the original tracker had.

Setting up a game should take seconds, not a ritual. Add your players once and they’re remembered on your device. New person joining? Add them. Different group tonight? Clear the list and start fresh.

Then you choose how this game ends.

Some games are a quick sprint before everyone heads home. Others somehow turn into “just one more” long after they probably shouldn’t. So I made it easy to start fast and stretch the game when you need to.

Play to a target score when you want a clear finish line and a tidy end. Or switch to a set number of rounds if the table’s not ready to stop yet, even if it’s technically past someone’s bedtime.

No fiddling with settings, no restarting from scratch. Just pick how long tonight’s game feels like it should last, and get on with playing.

No accounts to create. No ads cluttering the screen. No tracking ticking away in the background.

Currently free to use, because I still think good things can just be free.

A simple side project, built for fun

This app came together over a few spare hours, a chance to build something playful and a bit different from my day-to-day work.

It’s built with Next.js and shadcn/ui, deployed on GitHub Pages, and designed to stay out of your way while you enjoy the game.

If you play Flip 7 and want scorekeeping to disappear into the background, this one’s for you.

Check out on https://flip7.peachy.tools.

Happy flipping ✨

I’m a massive foodie, and I love exploring different cuisines. When it comes to Korean cuisine, it holds a special place in my heart. From kimchi to bibimbap, there are so many incredible dishes to try, but one of my absolute favourites is 마약계란 (Mayak Gyeran)—soft-boiled eggs marinated in a sweet, salty, and savoury soy sauce-based brine. As the eggs soak up the marinade, they become incredibly flavourful, achieving a perfect balance of umami, sweetness, and a hint of spice. It’s no wonder they’re called “drug eggs”. Once you try them, you won’t be able to stop! Not only are they extremely delicious, but they’re also incredibly easy to make!

Origins

While the exact origins of Mayak Gyeran aren’t clear, they’re said to have gained popularity in Gwangjang Market (광장시장), one of Seoul’s oldest and most famous street food spots. The name “Mayak” (마약) means “narcotic” or “drug” in Korean, but don’t worry, it’s just a fun way to describe how addictive these eggs really are! Some people also call them Egg Jang (Gyeran-jang), and who knows, maybe that name will eventually stick.

It’s believed that marinated eggs actually evolved from Jang-jorim (장조림), a traditional Korean side dish where beef is braised in a soy marinade with eggs and served cold. At some point, someone left out the beef, and this is how Mayak Eggs were born!

Recipe

As I love making and eating them so much, I decided to mix things up a bit in my blog to share my recipe for making this irresistible Korean side dish. Trust me, once you try it, you’ll be hooked!

Mayak Gyeran is usually served as a 반찬 (banchan, side dish), but it’s also a great snack or a tasty topping for rice bowls. I love adding them to ramen dishes too. They look incredibly appealing when cut in half, showing off their golden, marinated centers.

Ingredients

• Eggs – 6
• Soy sauce – 2/3 cup (~160 millilitres)
• Water – 1/3 cup (~80 millilitres) (1:3 soy sauce-to-water ratio)
• Honey – 1/4 cup (~85 grams)
• Mirin – 2 tablespoons
• Garlic – 2 cloves, minced
• Onion powder – 1 tablespoon
• Sesame oil – 1 tablespoon
• Sesame seeds – 2 teaspoons
• Chives or green onions – finely chopped, as much as you like (you can never have too much! 😆)
• Chilli flakes or 고추장 (gochujang) – 1 teaspoon (for a bit of kick)

Instructions:

1. Boil the Eggs: Bring a pot of water to a boil and carefully add the eggs. I boil them for about 8 minutes until the yolks are slightly soft but not too runny (see the picture). Once done, transfer the eggs to an ice bath to cool, then peel them.
2. Prepare the Marinade: In a bowl, add all the ingredients and mix well.
3. Marinate the Eggs: Place the peeled eggs in an airtight container and pour the marinade over them. Make sure the eggs are fully submerged. Cover and refrigerate overnight to allow the eggs to soak up the marinade.
4. Serve: When ready to serve, cut the eggs in half and drizzle with extra marinade. Garnish with additional sesame seeds or green onions, if desired.

So, what are you waiting for? Whip up a batch of these addictive eggs and experience the irresistible allure of 마약계란 for yourself!

I hope you enjoy these delicious Mayak Gyeran. 맛있게 드세요! Let me know how they turn out!

Throughout my life, I’ve met people from all kinds of backgrounds, and I’ve learned a lot from these friendships. Some friends became idols to me—people I looked up to and thought, “I want to be like them.” But friendships change, and people move on. As I’ve experienced these changes, I’ve come to realise that these shifts are part of my own growth.

Sometimes I look back at old friendships that shaped me, friendships where I wanted to be just like them. And when those friends drifted away, it felt like a loss. But their absence also gave me motivation. I wanted to be better, not just for myself, but to reach the level I once admired in them. Looking back now, I can see that I’ve achieved it.

I do reflect and remember them, as still being myself back then—idolising, wanting to be more, experiencing. To be clear, I do not look back with regrets, but in a way that has shaped who I am today, and for that, I am ever grateful.

It’s not always easy to lose friends, and I still miss some of them deeply. But I’ve come to appreciate that their presence—and even their departure—was necessary for my own growth. It pushed me to improve, to reach new stages in life that I might not have without those experiences. Sometimes, the best way to grow is to learn from what we lose.

If you’ve ever had friends who inspired you, think about how their impact lingers, even when they’re not around. Growth often happens when we least expect it, driven by the people we meet along the way.

“The best form of criticism is creation.” – D Watkins

Technology changes rapidly, and it can feel overwhelming. Instead of resisting, embracing creativity helps you navigate these shifts. I realised this after reading a Listen Notes article. They launched a tool to detect NotebookLM-generated podcasts, calling them ‘junk.’ This made me reflect on the power of these tools and the creativity they unlock.

I’ve been in tech for decades—from sharing files via floppy disks to using AI tools like ChatGPT and NotebookLM. When I first used NotebookLM for Friday Tech Focus News podcast, I was impressed by the quality. It sounded authentic, like a real podcast. I’ve replayed that first AI-generated episode multiple times, and it still amazes me. While there are subtle signs of AI creation, these imperfections hint at the future of this technology. Just as floppy disks seemed normal before, we’ll look back on AI tools as pivotal steps forward. The changes ahead are beyond our current imagination.

I understand the concern about AI-generated podcasts and blog posts feeling repetitive or formulaic. When everyone uses the same tools, sameness can increase. But should you dismiss all AI-generated content as junk? Poor use by some doesn’t invalidate the potential of these tools.

Here’s why I use AI:

• Efficiency: AI streamlines the creative process.
• Quality: It helps produce high-quality, ad-free content.
• Time Management: Balancing a full-time job and family limits my time. AI allows me to maintain my side project despite these constraints.

Ideally, I’d research and summarise articles myself, but that’s not feasible without AI. The technology handles much of the manual work, letting me focus on verifying and validating the output. AI acts as a powerful assistant, ensuring I deliver quality content even with limited time.

Responsibility lies with creators, not the tools. Take deepfakes as an example: the technology can be used ethically or unethically. The potential is vast, but its impact depends on how it’s applied. Just because some use AI to create low-quality or deceptive content doesn’t mean all AI-generated material is junk.

Key points to consider:

• Nuance Matters: Not all AI content is driven by greed or harmful intent.
• Thoughtful Use: When used carefully, AI can boost creativity and innovation.
• Tool, Not a Crutch: The technology itself isn’t the problem; it’s how you use it.

Rejecting AI tools outright stifles progress and creativity. Instead, use them to enhance your unique perspective.

My goal for Tech Focus is simple: provide high-quality, ad-free insights into the tech world. I’ve spent thousands of hours on this project, driven by a passion for meaningful content, not generating noise. NotebookLM helps refine my message without replacing it.

Here’s my message to Listen Notes: Do listeners gravitate towards “real human voices,” or do they value content that is meaningful, provides helpful insights, and meets their needs? At the end of the day, if the content is poor—whether created by humans or AI—most people will skip it. It’s a self-resolving issue driven by natural selection.

While AI might lead to similar-sounding content, creators like you and me can use these tools to highlight our unique perspectives. Don’t let your voice get lost in AI-generated noise. Believe in thoughtful creation because the tool’s value depends on your vision.

30th Anniversary Edition PS5 Pro, with only 12,300 units to be released, marks the end of an era. As much as it hurts me, this truly signifies the end of PlayStation for me.

Before owning a PS1, I had a Sega, and before that, an NES. I have spent countless hours having fun (Twisted Metal is one of my all-time favourites). I feel privileged to have had so much fun, to be honest.

But it saddens me deeply that now it’s all about marketing. What’s the reason for releasing only 12,300 units? This is a treasure trove for scalpers. I remember how hard it was to get my hands on a PS5 after scalpers were buying and inflating the price of the console.

But what I don’t understand is why Sony is limiting the joy that could have been shared. At the end of the day, isn’t this about the PLAYERS? It doesn’t seem so anymore.

Sunsetting disc drives, paywalling to play games online. You can colour things grey as much as you want (like they did for the 25th anniversary), but those days are gone when PlayStation was, to me, about having a lot of fun.

It truly hurts. I don’t enjoy playing my PS5 like I enjoyed playing my PS1, and I won’t enjoy the PS5 Pro or any other limited edition because that feeling is gone, marking the end of an era of fun. The PS1 will forever be the console that was new and felt different.

This week has been full of announcements, which has made me think and eventually prompted me to write this post.

Yesterday, I watched the announcement of Apple’s iPhone 16. Most of the information had already been leaked before the event, so there wasn’t much of a surprise. Although the announced AirPods Pro 2 software update now provides clinical-grade Hearing Aid features, it can be viewed as reasonably priced compared to some mid- to high-end hearing aids, which, to me, seems pretty awesome. However, after the Apple event, Huawei upstaged Apple by releasing a new phone—a phone that I had predicted and envisioned as the next logical step after the first folding phone was announced back in 2018: the triple-folding Mate XT Ultimate Design phone by Huawei. So the question arises, if you could see this coming, does it still count as something “new”?

Today, Sony announced the PlayStation 5 Pro—another logically predictable iteration that doesn’t feel like something “new”, similar to the iPhone 16. Something new to me means something I couldn’t imagine or predict would exist in the future, like the first PlayStation 1 console or the first iPhone. Everything after that is just an enhancement of its predecessor’s success.

Innovation has changed the trajectory, rather than inventing something “new”, we are stuck in a loop of enhancing the old. But how could a visionary with novel ideas compete with the big overlords of the tech world?

Everything that’s new seems bleak to me now, possibly due to my growing older and becoming a father. However, I do find old things quite enjoyable, which no longer appears boring at all.

By the end of this month, there will be two new announcements: a cheaper Meta Quest 3S VR headset and the Nintendo Switch 2 console. We will see what this new iteration brings.

Okay, this will be a long one. I sacrificed my sleep over this post, so you better read the whole thing and write me some AI-generated response!

“The pessimist sees difficulty in every opportunity. The optimist sees the opportunity in every difficulty.”

– Winston Churchill

Last week, I read that Google pulled an ad after some criticism before airing it during the Olympics, where a Gemini AI was used to write a fan letter. Google stated:

“We believe that AI can be a great tool for enhancing human creativity, but it can never replace it.”

The ad showed a father using AI to write a fan letter to his daughter’s favourite athlete, Olympic track star Sydney McLaughlin-Levrone. You can see the video on YouTube. I didn’t want to embed the iframe, sorry for the extra hop.

Takeaways

After watching the ad, these were my key takeaways:

• A fictional story from a father’s point of view about his aspiring young daughter idolising Sydney McLaughlin-Levrone.
• AI giving tips on how to teach hurdle technique.
• AI composing a fan letter with some guidance from the father.
• No true evidence that the young aspiring girl is a fan of Sydney McLaughlin-Levrone.

Perspective

As a father myself, I know how proud we dads are of our little ones; spending time together means the world. But what is wrong with using AI to generate a fan letter? Surely, she is too young to write a full-blown essay about how inspiring Sydney McLaughlin-Levrone is to her. Does this make her less of a fan? No. Obviously, Google needs to sell AI, but at the same time, even if your intentions are good, there is a general notion that everything AI is bad.

On the other hand, how genuine are fan mail responses? Most famous celebrities have management teams that handle fan mail. More fans mean more revenue from sold merchandise, books, TV deals, and tickets. Can we be 100% certain that these management teams don’t use AI? How impersonal is it for famous people not to respond to fan mail personally? Is it even humanly possible for celebrities to personally read and respond to every single fan letter?

To that end, does it really matter that some percentage of fan mail is generated by AI? Has a fan-written letter ever completely changed someone’s life after reading it? Maybe, I don’t know. But we are living in a dystopian world where having synthetic interactions is becoming the new normal.

Criticism

Critics like Linda Holmes, host of NPR’s Pop Culture Happy Hour podcast, and Shelley Palmer, a professor at Syracuse University, voiced concerns. Holmes wrote, “Who wants an AI-written fan letter?” Palmer warned of a “monocultural future where original human thoughts become increasingly rare.” All I know is that AI is here to stay, and we need to evolve our thinking and perception of what genuine human thoughts even mean.

Surely, if the ad had shown a hand-drawn picture of Sydney McLaughlin-Levrone attached to the AI-generated letter, it would have been okay; we would let it slide this time. Or maybe we should expect meaningful, poetic, prolific essays from young fans? I would be more heartbroken knowing that the fake response I got from my idol was not a genuine response, even if it came with a photo and an autograph. Knowing that a heartfelt letter to an idol was never actually read would be devastating. But this is why we humans have imagination; we believe our made-up stories. We want to believe that our idols do read our letters. Our idols read them and personally respond with a thoughtful signed photo, which gives us motivation for whatever it is we dream to achieve.

Concluding thoughts

Labeling content as “generated by AI” would suffice, but would it? But obviously, “Who wants an AI-written fan letter?” Right? NOBODY! Who wants an AI-written recruiter reach-out letter? NOBODY! Does this mean that recruiters will suddenly stop using AI to generate reach-out letters? NO!

It’s down to ethics and whether it will be accepted by general consensus.

People can have their opinions and their lofty, unrealistic standards, but not everything is as straightforward as it might seem.

We need to accept that every communication bit might be AI-enhanced, and there’s no sure way of telling otherwise. Receiving a fan letter, whether written or enhanced by AI, is still meaningful to some point, at least from the sender’s perspective, and doesn’t make the sender less of a fan. If you want a genuine experience, go out and talk to people.

If you are here, you are probably wondering what a beer glass gadget is, just as I did. A colleague, after work, was telling a pub story about how he had heard of this beer glass gadget, located at the bottom of the beer glass, that makes beer fizzy. Apparently, if you have been served a dirty glass, you won’t see bubbles forming at the bottom of the glass. Now you know. I found this very fascinating, which is why I decided to write this post about it.

Yeah, this sounds like some crazy conspiracy theory. Surely, we would all know about this type of invention that makes beer bubbly, right?

The mystery of the beer glass gadget

I decided to figure out what was going on. What is this mysterious gadget at the bottom of the glass? I had no clue what could make beer fizzy.

After some googling, I found the answer. The process of making beer fizzy is referred to as the beading effect.

A nucleation site (e.g. scratch on a glass or an engineered device) that allows bubbles to form and rise in the beer, sometimes referred to as “beading.”

So apparently, these purposefully engineered beading gadgets provide nucleation points that are etched at the base of the glass to:

1. Maintain a frothy head to help the beer maintain its fizziness through controlled gas release, giving it a fresh and crisp taste.
2. Make the beer look more appealing. Who wouldn’t enjoy some bubbles floating from the bottom of the glass?

How do the bubbles form?

But how and why are bubbles forming with the help of these nucleation points? Well, these points help to release carbon dioxide (CO2) bubbles from the beer, which enhances the carbonation.

Nucleation points

The nucleation points act as places for bubble formation. On a side note, this is the same way Mentos candy provides nucleation points for the formation of CO2 bubbles from carbonated soda. Once the beer is poured into the glass, the dissolved CO2 in the beer attaches to these nucleation points, as the energy barrier for bubble formation at nucleation points is lower.

Bubble release process

The bubble at each nucleation point keeps growing as more CO2 accumulates. When a bubble reaches a certain size based on the characteristics of the nucleation site, the size, shape, and roughness, its buoyancy (the adhesive force keeping the bubble attached to the nucleation site) causes it to detach from the nucleation point and rise to the surface

Nucleation characteristics

The characteristics of nucleation points, such as their size, determine the frequency of bubbles being formed. Smaller points will produce a finer stream of bubbles, while larger and fewer nucleation points can create larger bubbles that rise less frequently. Another interesting fact is that while the bubble rises to the top of the beer, it will keep growing by collecting more CO2.

Helps to surface aromatic compounds

As these bubbles in a stream keep rising to the top of the beer, they also carry volatile aromatic compounds from the beer to the surface, enhancing the flavour for that crisp taste.

Experimentation

I got my hands on one of those beer glasses. It wasn’t that hard, to be honest. Also, what a great opportunity to experiment and record some video, not that I wouldn’t enjoy a cold beer on a hot summer day. 😉

Cheers! 🍻