jillesvangurp

Fake news on the internet is at this point a tool that has been widely used for probably at least twenty years. The history of propaganda is much older.

A big part of the problem is social networks treating very non authoritative sources as if they are authoritative based on all sorts of factors that can be manipulated or based on financial gain (somebody pais to get their content promoted).

Users of these tools include lobby groups, nation states, politicians, criminals, scammers, etc. Anyone interested in promoting their cause, sowing FUD about other people's causes, spreading misinformation, or distracting from inconvenient but accurate information.

We'll always have people like that. But we might want to think about getting better tools to deal with them. I prefer getting my news from reputable sources instead of random clickbait distributors or whatever. Keeping track of reputation for domains, people, etc. is currently hard. I don't think it's necessarily a technical problems. We've had the tools to digitally sign stuff for decades. But we're not using them effectively. People can't be bothered to sign their emails, documents, photos, articles, etc. The tools to add and check those signatures are not there or an afterthought at best.

This where system thinking, or the lack there of, becomes an issue. You would be right if we only had exactly enough capacity for four hours of power. But of course with batteries now being produced by the multiple twh per year, each of them capable of cycling on a daily basis for a decade or longer, we soon will have quite a bit of battery connected to the grid. If you have a hundred of those four hour systems connected you'd blow fuses throughout the country if you dumped all that power on the grid all at once. That could be days of power under normal usage.

And grids can actually be connected together via cables now. The cliche is that solar panels don't generate at night. But if you connect grids via a long cable running east to west you can actually have solar power at night. And when you connect them north to south you can have solar power on a dark winter day as well. And when you add batteries to the mix (on both sides), you can keep those charged as well and use that power regardless of what time or season it is.

And that's ignoring wind (on and offshore), hydro, nuclear, and other clean forms of power that we currently don't use at scale (geothermal, fusion, etc.). If you add all that to the mix. And cables. And lots of battery. Grid connected car batteries, balcony solar, and all the rest, you end up with a lot of capacity.

Gas power is not going to way. But we probably already have too much of it and it shows in the poor utilization of existing turbines. Ironically, adding more gas turbines to the grid only makes that poor utilization worse for owners of these plants.

There's a bit more to it than that.

Battery production is being invested in heavily. There are new types of batteries (like sodium ion) that are extremely relevant to the article that it barely even mentions. The pitch for those is of course no lithium, better temperature ranges, less need for cooling/heating them, much better fire safety, better durability, etc. This all adds up to lower cost when done at scale. That all adds up to a lot less cost for especially grid storage.

I agree the 33% number is meaningless. It's probably too conservative. The article actually does not talk about new battery chemistries like sodium ion. And given that large scale factories are already producing those by the tens of gwh, I think that's more than a little oversight. Additionally, there are learning effects that explain the large differences in prices between markets that the article mentions. Those are optimizations that countries have yet to implement. Some might not get around to that any time soon (like the US seems a bit of a basket case on cost/efficiency for a lot of things).

Because of the improvements in battery cost relative to the stagnating improvements in gas turbines, gas turbine manufacturers are going to be conservative in investing in new production capacity. A lot of companies that say right now that they are going to want to buy gas turbines might no longer want to in a few years. For example the low cost and ready availability of solar+battery might be persuasive for some. Investing many billions in new production capacity that won't come online until some time next decade is highly risky if the projected demand could evaporate. The reason there is a shortage is that that's what they thought five to ten years ago as well. So, it's easy to predict that turbine shortages might continue to exist.

Good article. Not just blogging but any form of written communication can benefit from this.

I went through a phase where I had to be trained up a bit when I started out as a researcher. My written communication wasn't great. I went to high school in the Netherlands where I learned English. I am not a native speaker. But I wasn't able to write well in my native language either. I was doing my master thesis at the time and had already agreed to continue doing a PhD. I was writing what would be my first published article at the same time. My style and grammar were poor. Sentences didn't connect. The text was poorly structured. I was making all the beginners mistakes that somebody that is not used to writing professionally makes. I managed to get it published at a conference. But it wasn't great.

The best advice I got from my professor at the time was that anything you do systematically wrong you can learn to do right. You just need to understand what you are doing wrong. He provided a lot of feedback on what I was doing wrong and why and I got better at writing rapidly. He was very patient with me, except when he caught me doing things wrong which he had corrected me on already. After a few months, I was catching most of my own mistakes. I learned to review text at the same time. I was reviewing lots of papers by others as well as my own work. So, I got a lot of examples of good and bad papers while I was also trying to figure out my own writing.

Another insight here is that writing text is a bit like programming. You can't reference things you haven't talked about yet. There's a sequence of presenting things and an expected structure. Dead code elimination is a thing as well: remove text that no longer serves a purpose. Text needs to be logical, self contained, etc. Lots of coupling is not great. Low cohesion is bad as well. Keeping text simple and minimal and flowing naturally is an art. You can challenge yourself to say the same thing with less words. Very useful if you have a hard word count or page limit.

Long term storage is a vanishingly small portion of overall investment in this area for a good reason. Most solutions in this space are a combination of deeply uneconomical and unpractical. It's why I call it a hobby horse. The people demanding we need it right now seem to be a mix of either poorly informed on the topic and people acting on behalf of the fossil fuel industry to sow fear, uncertainty, and doubt.

Long term storage has the simple economic challenge that it's simply redundant most of the time and therefore very expensive for the few times it might actually be needed. As a result, alternatives to storage tend to be more economical and practical. Having gas as a backup is a given for the next few decades. There are so many gas plants around and more are being built. Mostly short term storage is actually marginalizing a lot of those in their role as peaker-plants.

The plan for the last few percent of the renewable transition, which is what long term storage would be about is a lot less interesting than just getting the first 90-95% out of the way. Which will work fine without anything complicated or new. Burning some gas for the remaining percentages is not pretty as a stop gap solution but it's a lot better than what we currently do. We have a 3-5 decades to figure out the rest.

Towards the end of that period, most remaining gas plants will be running less and less. There are a lot of other solutions than long term storage btw. Including using cables to move power around over long distances. Connect sunny areas to dark areas. Or windy areas to calm areas. There are some plans in various stages of progress to connect e.g. Singapore with Australia, Morocco with the UK, the east coast of Canada with Europe, etc. Each of those projects adds many GW of power import and export capability. Add things like flexible demand to the mix and we gain the ability to turn things on/off selectively depending on power availability.

We got an alarm delivered via phone here in Berlin a few weeks ago notifying of an evacuation zone near our office where they had found a WW II bomb. Big enough to evacuate everyone in a 500 meter (I think) radius. Our office was outside of the evacuation zone.

It was slightly funny because I configured my phone to use a weird chipmunk voice to read stuff out loud apparently (not a feature I use) and it actually started reading out the message. Big WTF. moment until I figured out. Grabbed my attention though and it was the first real alarm I've received. We get a practice alarm message every few months.

Many people in the UK, Netherlands, Germany, etc. And even some further north have done the math and it comes up positive.

You have to be realistic of course. There are a few months in the winter where they don't perform well and returns drop well below a single kwh of power. But then you might start getting some bright winter or early spring days and you go back to getting many kwh per day worth of free power as early as February or March.

It's better if you have a battery and if your local energy supplier has favorable contracts.

Long term storage seems to be a be a hobby horse that comes up in a lot of theads like this. The reality with batteries is that the overwhelmingly vast majority is for dealing with short term supply/demand fluctuation. The supply spikes are as important as the demand spikes here. Empty batteries can absorb those.

Anything beyond four hours drops off rapidly in terms of economics. And a few hours is all most energy companies need for planning to spin up or down gas (or coal in some cases) plants or decide to import/export power via cables.

As for nuclear, it's is a lot less flexible and very expensive. And in any case very little of it is coming online for the foreseeable future (next 1-2 decades). We'll get probably quite a few TW of battery, solar, and wind deployed over that time. Nuclear is going to be a rounding error on that even in the best case.

Only partially and it depends on their contracts. And only when transporting goods. They also have to cover empty miles in between gigs when they are not getting paid. Or when idling. And of course fuel prices differ between states. So, it depends but mostly cost goes up for truckers. Some trucking companies might be very vulnerable to high prices like this.

While that won't help most current truckers, the shift to electrical trucks is going to create some opportunities for the companies that buy them to undercut diesel trucks on price. Electrical trucks are still scarce and expensive and infrastructure is also lacking (in the US at least, EU/China is a different story). But of course with diesel prices spiking this high, it only exaggerates the benefits. And it looks like price uncertainty in general is could be here to stay with e.g. Russia probably needing many years to rebuild their oil infrastructure and the middle east also needing a lot of time if/when some sense of normality is restored there.

There are likely to be a lot more EV trucks on the market over the next years. China is already heading for 40+% of trucks sold (of all types, including heavy trucks) to be electric. They are way past the testing and experimenting phase there.

Transitioning the whole truck fleet in the US will take many years of course. But with Tesla bringing their factory online, other truck companies are likely to also increase production. This could quickly go from thousands of trucks sold per year in the US to tens of thousands. Tesla alone has the ambition to do 50K semis per year. We'll see if they manage. But it's going to disrupt truckers that are highly dependent on diesel prices when whole fleets start becoming immune to diesel price uncertainty.

I read this book and its sequel a few years ago. Interesting conceptual book that touches on some very weird ideas, which I enjoyed. But I wouldn't call it my favorite book. It's not necessarily the easiest to read. Especially the sequel.

Spoiler alert:

The protagonist is a person with severe brain damage that is AI enhanced to become this useful but also severely autistic individual that has no emotions. The novel features also features a genetically engineered vampire with psychopath tendencies. These and a few other characters end up being send on a deep space mission to combat an alien intelligence.

All this obviously requires a bit of suspension of disbelief. But as a psychological take on AI and genetically enhanced intelligence, it touches some interesting topics.

There also higher tariffs and transport cost. Shipping cars from China to the EU can cost up to a few thousand per car. Local production seems less competitive. But the bottom line is that there are now a lot of relatively affordable EVs in the market. The cheapest cars in most markets now include EVs.

For example the Dacia Spring that is selling for around 17K Euros in Germany. And then there are tax rebates, discounts, etc. that are causing some of these cars to be available for a lot less occasionally in some markets.

Markets where EVs have been available in larger numbers for a while now have a sizable EV market. Australia is relatively early with its adoption, so the second hand market is a lot smaller there. The EU as a relatively early adopter has a lot of EVs from 2020 or earlier that are still under warranty and being traded in quite large numbers now. In the last few months these been a very hot commodity as lots of people are feeling the pain of high fuel prices.

All these new small cars will become a large part of the second hand car market in a few years. There will many be millions of them on the roads in a few years. And they last quite long.

The alternative would be trying to file court cases outside the US for Qualcomm, which is where Huawei does most of its business and infringing.

There's a long history of various companies doing patent license deals like this in the telecom industry. Nokia still earns a lot of money this way, for example. Apple settled with them on a few occasions.

Speaking of which; Apple has its own 5G chips now so they are a bit less dependent on Qualcomm now. Huawei and Apple both have a lot of their own in house technology at this point. But of course Qualcomm always was a strong IP company with lots of patents and its impossible for either to make phones without infringing on at least some of those patterns. And vice versa, a lot of Qualcomm's stuff would necessarily infringe on some patents by others.

That's why these cross licensing deals like this are usually the outcome of disputes around IP.

This article tosses together bio fuels and plastics. By conflating the two, they seem to try to create some urgency for subsidizing the creation of both. And of course bio fuels are heavily subsidized already.

This article deliberately confuses things by conflating it with the discussion around bio fuels. Which, as you point out is largely redundant given that we now have decent EVs for literally anything with wheels commercially available. Anything from the tiniest scooters to the largest vehicles operated in the mining industry. And everything in between.

The fallacy with this debate is the focus on disproportionately small parts of the challenge of de-carbonization. If we manage to electrify most road and rail transport, heating (industrial and domestic), some shipping and aviation, etc. that would make an enormous dent in fossil fuel usage. After that we might focus on harder things like plastics. I'm not against people focusing on that now. More power to these people. But put it in perspective: it's relatively small part of the problem. And we have alternatives.

This fallacy is a recurring theme in fossil fuel lobbying materials. By spreading fear, uncertainty, and doubt, they hope to slow down the accelerating implosion of their primary revenue maker: fuel sales. So, they talk up hydrogen and bio fuels. Sponsor studies on the impact of windmills on wild life. Etc. I can't help but look at this article and wonder who payed for it.

There are several things happening:

- Retail 2D codes are being rolled out at scale from next year already in many markets; 2d codes will be on many products in the supermarket. Preparations for this have been ongoing for quite long. Mostly scanners have been ready for this for some time. Most recent hardware should handle all this fine. Scanners that don't handle 2D will rapidly disappear.

- For industrial use, QR codes are already quite common. Use of QR codes is much less about replacing existing bar codes with QR codes and more about completely new uses for QR codes. You are right that most factories won't do invasive changes in their setups unless there's a good reason. But efficiency improvements tend to be a good reason.

I was at the Logimat fair in Stuttgart earlier this year. Lots of companies demonstrating scanners, sticker printing, camera modules, and all the rest. There are a lot of companies actively pushing all sorts of solutions around 2D codes.

The big picture here is that LLMs using APIs to get into information stores works but is stupidly slow. Email is a good example. You can hookup gmail or whatever to codex or claude and it's great. But it takes ages for it to dig through your inbox and that involves doing a lot of API calls. Most SAAS tools are about gate keeping data via APIs. You need to break open those silos to get to the data. A CRM is just a contact book with a lot of meta data. Why not download all that data and provide direct access? It's a lot faster than accessing records one by one via some API.

What if email existed on your laptop? Like was common before we moved to using web based applications? What if your local email client had a local database with a proper search API. That sounds a lot like what Thunderbird, Outlook, etc. did a quarter century ago. And still do if you still use those. Digging through a local database is a lot faster than dealing with relatively slow API access. LLMs are really good at doing stuff like that.

I run a small company, I've taken a sledge hammer to many of the SAAS tools we use in our business in the last years. We got rid of Hubspot, Slack, and many other tools we use for financial planning, etc. These got replaced with self hosted open source things or simple file based repositories in some cases. This has several huge advantages: 1) it's stupidly easy to get to the data for LLMs now 2) LLMs can administer these things for us 3) we saved quite a bit in licensing fees. 4) removing obstacles means LLMs can do complex things we used to do manually in dreadful SAAS tools.

Bronze and iron weren't limited for military usage. Basically from an economical point of view, the scarcity of bronze limited the size of economies and armies.

Once people started using iron instead of bronze, the first use was probably creating tools used for farming, food production, and other important things: Axes, hoes, spades, etc. And they did not just substitute bronze tools but because iron was cheaper, they probably created more of them, and new tools. And it also enabled things like nails and other construction material.

So, the economy and power structures shifted as tools enabled more economic output. By the time steel became good enough for weaponry, the bronze age was effectively already a thing of the past as the economy would have shifted to using iron for most tools long before.

And then later of course, having cheap and good enough steel weaponry enabled the creation of larger armies too. So towards the end of the Iron age, you get e.g. the Romans creating huge legions that were simply not economically feasible in the bronze age. And given that hand to hand combat was basically the way people waged wars, the ones with the bigger armies would tend to win.

You make a good point about LLMs being (so far) great at identifying solutions to complex puzzles but not yet coming up with their own theories, questions, etc. The business of finding interesting questions to answer rather than answering them is the essence of what scientists do. Good science identifies more questions than it answers.

With Fermat's theorem, the genius was in the original theorem. Which then caused generations of mathematicians to break their heads over trying to prove it correct. Fermat didn't write down a proof. His theorem was famously just a scribble in a side line of a book. Probably it was something that he had an hunch about that he couldn't quickly falsify.

I don't think AI is being used much for coming up with new problems yet. But I don't see why that would not be possible either. It's the obvious next frontier after AI clears the backlog of existing theorems. I imagine scientists are already using AI to find new interesting problems to work on and generally explore the problem space. But fundamentally, the reflex of asking or imagining "if this is true, what else could be true" is something that distinguishes people from AIs. For now at least.

There are several moats:

- Training data, harnesses, and processes, that drives the quality of the models. Several companies have those. Some of those companies are in China.

- Infrastructure and funding for running those. That's a scarce commodity currently, training the latest frontier models cost billions of dollars apparently. And if you don't have the infrastructure already and don't have the suppliers on speed dial, good luck getting anything.

- Infrastructure for running inference for running what comes out of those. Several of the key providers of this infrastructure are using their own in house chips for this now. At scale this means huge cost savings.

If you start from scratch without infrastructure, there are a bunch of open source things you can find. But beyond that, you'll have a lot of catching up to do. That's the very definition of a moat.

Here in Germany, you can just print your own codes or even just write those on the envelope. You can buy these via a website.

I tend to just use one of the many late night shops doubling as a post office these days in the rare case I need to send something. Thankfully, the amount of paper arriving in my mailbox has been declining steadily. It's next to nothing at this point. Dedicated post offices are very rare now. But there are many of these late night shops doing post and packages on the side. They take your envelope, they print a label with some kind of scanable code, slap it on there, and take care of delivery.

The codes are basically short hashes that correspond to a database record confirming the correct amount was paid. Very hard to forge those.

No need to handle stamps anymore. Only for international postage.

The key point with small modular reactors vs. reactors like this is that construction happens in factories rather than on site. Theoretically, you might get some economies of scale from series production in a factory that is much harder to get doing bespoke construction projects. Which is why historically, nuclear projects tend to blow through their cost estimates and why having larger reactors makes that a bit more tolerable. Of course until somebody actually does this and scales to hundreds/thousands of reactors production, this is all theoretical.

This particular reactor is already estimated at over 5 billion $ I think. That's a lot of money for just 300MW and it probably could end up being a wildly optimistic estimate as well. I think effectively much of the construction is still on site and not in a factory. Hence the need for a construction permit. So, you get all of the downsides of complex on site construction such as including high cost, permitting overhead, lots of bureaucracy, associated delays & uncertainty, etc. without the upside of actually delivering a lot of power like you would with a larger reaction. Calling it "modular" might be overstating things a bit.

Probably cheaper, faster, and easier to just plonk down 5GW of solar, wind, or battery (or combinations of those). Especially if you calculate in the 200-300% time and dollar budget that many nuclear projects seem to end up having. Maybe this one will be different. A lot of people have a lot riding on Nuclear projects breaking this trend. But then that has been the case for decades.

That gets you a long way actually. This actually sounds a bit like a tamed version of openclaw. A self learning and adapting always on agent that takes care of stuff for you. And of course they do employ Peter Steinberger who created OpenClaw since early this year.

So, I expect it might be borrowing some ideas from that. Probably/hopefully not any of the code.

I ran an OpenClaw instance for a few months on a vm. It was fun but also very flaky. These days I already have chatgpt connected to gmail and drive, which is very useful for me and saving me lots of time already. I have some scheduled tasks running as well. This would just be the next step up from that. I'll probably give this a try when they roll it out in Germany.

It will be interesting to see if Anthropic will launch a competing feature as well. Also, there is now growing competition from Google, MS, Meta, and Apple that are each doing their own versions of AI Agents.

They usually roll things out in the EU after a few days/weeks. It's too juicy of a market to ignore for them. It's an easy way to phase in a new product without having to commit to world wide coverage on day 1.

I like Störtebeker Atlantic Ale, IPA style beer. Very drinkable. Available in most super markets and Späti's (late night shops). Here in Berlin there's also the Brlo Nakid which is similar in taste.

If you want a pils style beer, Jever Fun is pretty hoppy. Krombacher has a decent 0.0 beer as well. A bit on the sweet side for me but I have that sometimes. I also recall seeing a beer that was labelled as Feinherb at some point that was pretty tasty. The exact name escapes me.

Several of the Bavarian beer hall style helles beers now also come in alcohol free versions. Very drinkable. They taste very similar. Bayreuther is one that you might find in some shops.

Many of the Hefeweizen varieties also exist in alcohol free versions and are pretty drinkable. Most places will have that available.

Not German but some bars now have Guinness 0.0 in cans. Haven't seen that in the supermarket but it's a near drop in replacement for normal Guinness as far as I'm concerned and I tend to order that when I see it available.

I stopped drinking alcohol a few years ago but I still enjoy going out with friends who do drink. Mostly people barely notice what other people drink. And the lack of a hangover after a good night sleep is a nice bonus. Although you can overdo it. I always joke about all of the downsides of drinking beers without any of the upsides.

If you are not in the market for a new car, there are plenty of second hand EVs now. Including some that are close to 15 years old. Most of those will easily beat your ICE car on maintenance and fuel consumption. And they aren't that much more expensive even.

> typically more expensive to purchase and insure

The cheapest car for sale in the UK is an EV at this point. The market is rapidly filling with EVs in essentially all price segments. Most EVs are still under warranty when they hit the second hand market for the simple reason most EVs on the market were produced in their typical warranty period.

What you are referring to is usually referred to as an EREV (Extended Range Electrical Vehicle). The reason legacy manufacturers market their less efficient plugin hybrid and non plugin hybrids (100% of the power comes from petrol) instead is because that's what they have and want to milk for as long as possible. In terms of drive train these are more complex to manufacture and more costly to maintain than an EREV. EREVs typically have slightly larger batteries than a typical PHEV as well. Typical electrical ranges are at least 80-100 miles or more in some cases. Usually the cars are available as without the range extender as well. This is less common for PHEVs and hybrids. Though of course some legacy manufacturers are still trying to push compromise cars. Like Nissan in this case.

Nissan has been really struggling for the last few years, this new hybrid engine doesn't look like it will turn things around for them to me.

EREVs are an idea that is not new and also relevant in shipping and aviation. Many modern large ships (oil tankers, cruise ships, etc.) are already EREVs effectively: an electrical motor drives the propeller and a generator powers the motor.

And there are are a few plane designs that are using small batteries + generators to power flight. It makes a lot of sense: electrical motors are a lot simpler, lighter and cheaper to maintain. Engine maintenance is very costly in aviation. And so is fuel. The recently announced X1 is a good example of a plane that does this. There are a few other manufacturers working on this as well. Early days but the potentially large fuel savings probably ensure this is going to happen one way or another.

People forget that pull requests and doing code reviews in the context of those is still a fairly recent thing. People did some code reviews before that of course but nowhere near as strictly. Same with testing practices, static code analysis, etc. Most of that wasn't all that common until beginning of this century. I remember using findbugs with Java around 2004. It actually found bugs in my code the first time I used it. No review had caught those. And we got lucky not finding them in production. But they were definitely bugs. Our system didn't have unit tests; it was all manual. Junit was a fairly recent system that hadn't been around for that long yet. Our build was done with Ant. There was no test phase. Our tech lead would of course check my work and correct & educate me (I learned a lot). But a lot of bugs slipped through as well.

Git did not exist either, I migrated out cvs to a beta release of Subversion. We only used branches for releases. We'd cut a branch just before a release. Test it (manually) and then ship. That was a process I helped put in place actually. After release, master would diverge quickly so back porting fixes was not really a thing. We'd support releases for as long as our customers used them. Often that involved just upgrading them to the recent version. We shipped when things were good enough.

I think the notion of people reviewing any meaningful amount of generated code is simply delusional. As you say, we do need alternative means to replace those checks. And a lot of that is going to be AI driven as well. AI driven testing, code reviews, and all the rest. Essentially all the stuff we used to do manually (poorly).

And we do have an important new tool as well: clean room code replacement. That used to be prohibitively expensive but now it's not. If you have something that is well specified through documentation, APIs, specifications, tests, etc. replacing it is fairly straightforward now. There are some early examples of people using LLMs to generate functioning replacements for things like Postgresql, browsers, compilers and similarly large and complex systems. While not perfect, these things seem to work, pass their tests, and generally not be completely horrible. It's only going to get better from here.

The notion that people are going to ever manually review code that was generated for such systems in mere hours/days is beyond imagination. How? When? Who? Why? It simply does not scale. It's only going to be more and more code. The amount of code no person will have ever looked at will soon dwarf the amount of code that is still manually inspected/created pretty rapidly.

> Cars that are designed from the ground up to be EVs tend to be much better

Better here means more efficient, cheaper to assemble, more spacious on the inside, etc. Having to repurpose space left when you remove the combustion engine + fuel systems and other supporting infrastructure to host an electric drive train plus battery typically means compromising on these things. It works but it won't be as good as it could be.

And of course in a market full of competitively priced EVs from vendors that designed them from the ground up to be pure EVs that means fighting with your hands tied behind your back.

This thing will probably still do alright for Toyota. But they'll have to discount it to compensate for it just not being all that great compared to its more optimized competition.

Toyota is very good at selling things at a bargain though. They are actually holding up pretty well compared to other legacy manufacturers (they even managed a bit of growth last year in units sold). But it's in a market that except for EVs isn't growing much where the Chinese manufacturers are expanding their market share globally at the cost of pretty much everyone else. Toyota is kicking the can a bit down the road here. Probably to buy some time to build something better in a few years. But they might be milking this platform for quite some time.

Good point. Procrastination is a big part of the issue for a lot of people. I'm a hopeless procrastinator. Some people are better at this than others.

Focusing your attention on things that matter is a good skill to have. Even if you only succeed part of your time, it's better than nothing.

I look at product features through a lens of value adding features, commodity/basic non value adding features, and negative value adding features (commodity features done poorly).

A good example is sign up and sign in functionality. You have to have it. It doesn't do anything valuable. But if you do it poorly it removes value from your product.

I like this as an example because lots of product owners try to innovate with this and then somehow botch the job and create needless onboarding friction or otherwise annoy their users.

And other product owners try to play it safe and simply copy what others do without wasting much energy on the topic. Which is why in 2026, you still see companies insisting on good old email+password based signup processes. Because it is what everybody else does.

Both are creating negative value for their products by either over or under engineering something that is pretty important but essentially a commodity thing. Worst case, this can result in a lot of potential users abandoning an attempt to use the product before they ever even experience the amazing feature set it has to offer.

You see this a lot in product design. Most projects are 90% copy paste jobs with a lot of common, non differentiating and completely bland/unremarkable stuff. LLMs are very good at copy pasting. But there are a lot of decisions that when you challenge them a little bit would result in a very different and arguably better product. A good onboarding experience can make all the difference.

And then there's the simple notion of what is the quickest way to validate that your value adding features are good enough. Which mostly boils down to finding a way where you get those in the hands of users without wasting most of your time on the non value adding stuff that is nice to have but otherwise won't make a difference. LLMs kind of allow you to compress the time you waste on this stuff and do a better job of it. Which is a different way of saying that you will create negative value if you don't use LLMs or otherwise botch the job.

It's not that hard to bring old buildings (altbau in German) up to modern standards via renovation. Here in Berlin, there are lots of buildings from early last century that are very desirable property not just because of their exterior but because of their interiors. You get modern insulation, wiring, bathrooms, kitchens, etc. And high ceilings, wooden floors. When I moved here 18 years ago, there were still a lot of these buildings in their original state with no insulation, coal burners for heating, etc. You could smell the coal in the winter. All that has been renovated at this point. The communists actually attempted to cover up the nice facades by just plastering over it as that was considered decadent. A lot of that property has been restored to its original state at this point.

A challenge with some old construction is that it was (at the time) cheap construction. Which means that after a century there can be lots of structural issues, mold, leaky roofs, problems with foundations, etc. But you should compare that to modern cheap construction, which isn't all that great either and doesn't age well.

A lot of West Berlin post war construction is particularly grim and ugly because it was obviously built on the cheap with little or no attention to how things looked. Surviving, pre-war construction is a lot nicer. Especially if somebody invests in fixing that up. You can see the contrast throughout the city because heavy bombing left a lot of gaps that were later filled in with some typically very bland and sometimes very ugly and now very dated construction. It seems some architects had some weird notions of modern. Especially some seventies and eighties architecture is a combination of cheap and very ugly. Cheap concrete does not age well. And painting it some weird shade of yellow or purple doesn't make it better. More recent construction is a lot less offensive, thankfully. The DDR mostly didn't bother with filling in the gaps and as a result, a lot of recent construction has been addressing that using more premium materials and architects that actually know what they are doing.