benjiro

Funny thing is that also makes it easier on LLM / AI... Tried a project a while ago both creating the same thing in Rust and Go. Go's worked from the start, while Rust's version needed a lot of LLM interventions and fixes to get it to compile.

We shall not talk about compile time / resource usage differences ;)

I mean, Rust is nice, but compared to when i learned it like 10 years ago, it really looks a lot more these days, like it took too much of a que from C++.

While Go syntax is still the same as it was 10 years ago with barely anything new. What may anger people but even so...

The only thing i love to see is reduce executable sizes because pushing large executables on a dinky upload line, to remove testing is not fun.

HDDs have also been under pressure ... There was barely a month ago a article here, from somebody who setup a cluster of like half a million, with several 1000's hdds. Just to store data for AI training.

Not even two days ago there as a article of backlog on HDDs for AI. Because everybody and their grandmother wants to store the entire internet, out of fear that AI scraping will become more difficult. Aka, they are gating data. And yes, you can train AI easily on HDD even with their lower IOPS. The fact that you got a few 1000 in parallel does the trick, and its often bandwidth issues that hit harder.

I just stockpiled a few extra 4TB NVME because i learned my lesson. NVME has not been dropping in prices after the manufacture pushed it up, and AI is going to keep eating NVME storage for a long time. Let alone HDD storage...

Welcome to the new normal ... Crypto miners killing GPU prices, HDD Crypto miners, Crypto miners again back with a vengeance, O pandemic, everybody needs hardware... Short time of benefits because of over production (on NVME especially, manufacture cut back production) AAAAND .. here comes AI.

Its something every fying year.

Easy to answer that one ... predictions are based upon accuracy. So if you have a int4 vs a float16, the chance that the prediction goes off is higher with a int4. But even with a float16, your still going to run into issues where your prediction model goes off. Its going to be a lot less, your still going to get rounding issue, what may result in a 5 being a 8 (just a example).

So while it can look like a LLM calculates correctly, its still restricted by this accuracy issue. What happens when you get a single number wrong in a calculation, everything is wrong.

While a calculator does not deal with predictions but basic adding/multiplying/subtracting etc .. Things that are 100% accurate (if we not not count issues like cosmic rays hitting, failures in silica etc).

A trillion parameter model is just that, a trillion parameters, but what matter is not the tokens but the accuracy as in, the do they use int, float16, float32, float64 ... The issue is, the higher we go, the memory usage explodes.

There is no point in spending terabytes of memory, to just get a somewhat accurate predictive calculator, when we can just have the LLM call a actual calculator, to ensure its results are accurate.

Think of a LLM more like somebody with Dyslexia / Dyscalculia... It does not matter how good you are, all it takes is to switch one number in a algebraic calculation to get a 0/10 ... The reason why i mention this, is because i often think of a LLM like a person with Dyslexia / Dyscalculia. It can have insane knowledge, be smart, but be considered dumb by society because of that less then accurate prediction (or number swiping issue).

Take it from somebody that wasted a few years in school thanks to that issue, it really does not matter if your a good programmer later in life, when you flunk a few years thanks to undiagnosed issues. And yet, just like a LLM, i simply rely on tool usage to fix my inaccuracy issues. No point in wasting good shoulder space trying to graft a dozen more heads/brains onto me, when i can simply delegate the issue away. ;)

The fact that we can get computer models, that can almost program, write texts, ... and do so much more like a slightly malfunctioning human, amazes me. And at the same time, i curse at it like my teachers did, and also call it dumb at times hehehe ... I now understand how my teachers felt loool

Thing is, a LLM is nothing but a prediction algorithm based upon what it trained. So it missing basic calculator functionality is a given. This is why tool usage is more and more a thing for LLMs. So that the LLM can from itself use a calculator for the actual math parts it needs. Thus increasing accuracy ...

Well, if you want a good reason to run LLM locally, this mail just arrived from Github ...

=======================================================================

We're writing to inform you that your GitHub Copilot usage between August 10, 2025 and September 23, 2025 was affected by a vulnerability that caused a small percentage of model responses to be misrouted to another user.

Your trust is essential to us, and we want to remain as transparent as possible about events like these. GitHub itself did not experience a compromise as a result of this event.

### What happened

On September 23, 2025, we received multiple reports from GitHub users that some of their prompts received out-of-context responses. We immediately began investigating the reports and learned that certain responses generated by the Sonnet 3.7 and Sonnet 4 models provided by one of our upstream providers, Google Cloud Platform (GCP), could be mismatched between users. This behavior occurred due to a bug in Google's proxy infrastructure that affected how requests were processed.

As a result, between August 10th, 2025 and September 23, 2025, certain responses (approximately 0.00092% of GitHub Copilot responses served by GCP for Sonnet models 3.7 and 4 in the affected timeframe) intended for one user were misrouted to another user. Google mitigated the issue on September 26th and disclosed via a public security bulletin: https://docs.cloud.google.com/support/bulletins#gcp-2025-059.

We are writing to inform you that one or more of your prompts' responses were misrouted and sent to another user. At the bottom of this email, you will find an appendix of prompt information owned by your account that were affected by this issue.

### What information was involved

In affected cases, a user could have received a model response that originated from another user's prompt. There is no indication of targeted or malicious activity, and GitHub systems themselves were not compromised. We've assessed that a malicious actor was not able to trigger or otherwise control the outcome of this vulnerability.

### What GitHub is doing

GitHub learned of the issue on September 23, 2025 at 19:45 UTC and immediately began investigating. Upon confirming the source of the issue, we reported our findings to Google on the same day at 21:00 UTC. By 21:37 UTC, GitHub completely disabled GCP endpoints used for Copilot to prevent further ocurrences. We worked with Google throughout their investigation, verified there were no more occurrences, and have since reenabled GCP traffic on September 29th, 2025 at 10:44 UTC following confirmation of Google's fix.

We then began working to identify which customers could have been affected.

Through the available telemetry, we have identified when the impacted prompt was sent, which client was used, the client request ID, and the user ID associated with the prompt author.

We are unable to provide which user the response(s) were sent to as we do not log model responses.

### What you can do

There is no action required on your part. We've identified the affected prompt(s) and included below the client request ID, when the prompt was sent, which client was used, and the user ID associated with the prompt author. The data provided may assist in finding the impacted prompt if you or your organization log this information. GitHub does not log user prompts or model responses. GitHub is committed to transparency during events like these and are sharing as much detail as available to enable you to investigate.

GitHub Support does not have any additional logging or data about these prompts. However, if you have questions or would like to discuss this further, please contact them using this link

=======================================================================

What i do not understand is why we are not seeing specialized models that go down to single experts.

I do not need models that know how to program in Python, Rust, ... when i only use Go and Html. So we are we not seeing models that have very specialized experts, where for instance:

* General interpreter model, that holds context/memory * Go Model * Html model if there is space in memory. * SQL model if there is space in memory.

If there is no space, the GIM swamp out the Go model, for the HTML model, depending on where it is in Agent tasks or Edit/Ask code its overviewing.

Because the models are going to be very small, switching in and out of memory will be ultra fast But most of the time we get very big Expert models, that still are very generalized over a entire field.

This can then be extended that if you have the memory, models combine their output with tasks... Maybe i am just too much of a noob in the field of understanding how LLMs work, but it feels like people are too often running after large models that companies like Anthropic/OpenAI etc deploy. I understand why those big companies use insane big models. They have the money to load them up over a cluster, have the fast interconnect, and for them its more efficient.

But from the bits and pieces that i see, people are more and more going to tons of small 1 a 2B models to produce better results. See my argument above. Like i said, never really gone beyond paying for my CoPilot subscription and running a bit of Ollama at home (don't have the time for the big stuff).

As somebody from your Neighboring country, its really comes over as condescending...

"I'm sorry for how you feel" is the definition of non-appology...

Hey, even the AI overlords agree on this...

> The phrase "I'm sorry for how you feel" is often perceived as a non-apology because it shifts blame to the other person's emotions instead of taking responsibility for one's actions. Instead of a genuine apology, it can sound like you're sorry they are upset rather than sorry for your behavior, and it can imply their feelings are the problem. A true apology acknowledges your role and expresses sincere regret for the pain you caused.

In Dutch its something like "Het spijt me, hoe u hierover voelt" ... See the issue, the first part is a apology, that is then reflected to the other person.

"We want to make sure we truly understand what you're struggling with." ...

We do not understand why *you are struggling with this*. For us its perfectly normal, so why are you having a issue with this.

So in Dutch its something like "We willen echt verstaan, waarom u moeite hebt hiermee". Aka, sending the issue back to the other person.

See the issue how both parts flow and shift the issue to the other person. This is not a English language issue because the same way of writing is also done in Dutch if you want to do a non-apology with a dose of gaslighting.

I never see anybody write like this, beyond those that have the intention to rile people up. Its gaslighting 101 ...

> Their emissions are the emissions of Western companies for whom they are doing manufacturing.

Spoken like somebody that never stept a foot in China.

Sure, manufacturing for the West is part of it, but up to a few years ago, entering Beijing alone resulted in your naval cavities burning, the moment the airplane door opened.

Because of the usage from coal in households. It was only until a few years ago, that they banned the usage of wood/coal around the city. Outside the city, its coal everywhere for the normal class people who own their (country)house. Near other large cities its still very coal centric in the winter.

And the heating (communal for apartments) is mostly coal and while the coal may burn a bit more clean, and there is some filtration going on, its not a ton. So while open coal burning was reduced directly in the cities like Beijing, they simply moved a lot of it outside the 6th ring.

All those EV's ... great, no more gasoline/oil usage but ... wait, where does a lot of the electricity come from? Oeps...

But wait, all that crypto mining, where do you think that electricity comes from?

And now AI...

And the consumer goods.

Your statement ignore a large part of the coal consumption in the country.

> I think the issue stems from too many people making their living off reviews that require something exciting to get views.

The problem is that our hardware as we know it, has lost a lot of its stretch. Used to be that we got 100% performance gains on a generation to generation update. Then it became 50%, 30% ... Like in the GPU market, the last generation that actually got me exited was the 1000 series (1070 specific).

Now its "boring" 10 a 15% upgrades for the same generation (if we do not count naming / pricing rearrangements).

When was the last time any of use was "hey, i am exited to potentially buy this tech, really". Apple M1 comes to mind, and that is 5 years ago.

Nvidia tried to push the whole ray tracing (a bit too early), but again, its just a incremental update to graphics (as we had a lot of tricks to simulate lighting effects that had good performance). So again, kind of a boring gain if we look back.

Mobile gaming handhelds was trilling, steam deck... Then we got competitors but with high price tags = excitement became less. And now, nobody blinks with a new generation gets released because the CPU/iGPU gains are the same boring 15 a 20%... So who wants to put down 700, 900 Euro for a 15% gain.

What has really gotten you exited? Where your just willing to throw money at something? AI? And we see the same issue with LLMs ... what used to be big step/gain, in barely a years has gone from massive gains, to incremental gains. 10% better on this benchmark, 5% better there, ... So it becomes boring (GPT5 launch and reaction, Sora 2 launch and reaction).

> When updates are more evolution than revolution, it makes for a more boring article/video.

If you think about it, there is a reason why tech channels have issues and are even more clickbait then ever. Those people live on views, so when the tech they follow/review is boring to the audience, they start pushing more and more clickbait. But that eventually burning the channels.

Unfortunately, we have a entire industry that is designed around making parts smaller and smaller every generation, to make those gains. As we lost the ability to make large gains on making those smaller making parts ...

Its ironic, as we knew this was coming and yet, it seems nobody made any breakthrough at all. Quantum computing was a field that everybody knew had no road to general computing at home (materials issues).

So what is left is the same old, lets may the die a bit smaller, gain a bit, do some optimizing left and right, and call it a new product. But for customers, getting product 2.1, being named "this is our product 3.0!!!! Buy buy" ... when customers see its just 2.1, 2.2, 2.3 ...

We are in a boring time because companies sat too darn long on their behinds, milking their exiting products but never really figured how to make new products. I think the only one that took a risk was Intel years ago, and it blew up in their face.

So yes, unless some smart cookie makes a new invention, that can revolutionize how we make chips (and that can be mass produced), boring is the standard now. No matter how companies try to repackage it.

> If HTTP is involved, lets just give up and make the system as slow as possible?

Did i write that? Please leave flamebait out of these discussions.

The original author (today) answered why they wanted to use this approach and the benefits from it. This has been missing in this entire discussion. So i really do not understand where you get this confidence.

> Switching out to JSON will meaningfully degrade performance. For no benefit.

Without knowing why or how the system was used, and now we know it is used as a transport medium between the db/nodes, its more clear as to why json is a issue for them. Does not explain how you conclude it will "meaningfully degrade performance" when this information was not available to any of us.

> CDB is only a transport medium. The data originates in PostgreSQL and upon request, stored in CDB and transferred. Writing/freezing to CDB is faster than encoding JSON.

Might have been interesting to actually include this in the article, do you not think so? ;-)

The way the article is written, made it seen that you used cdb on edge nodes to store metadata. With no information as to what your storing / access, how, why ... This is part of the reason we have these discussions here.

> You pay for it because you want backups and replication to be somebody else's problem.

Or you just use something like CockroachDB, YugabyteDB etc that auto replicate, auto rebalance if a node goes down, and have build in support for backups to and from S3...

Or if your a bit more hands on, multigress seems to be closing to completion ( https://github.com/multigres/multigres ) from the guy that make Vitess for Mysql.

The idea that managing hardware and software is hard, is silly yet, people (mostly managers it seems ) think its the best solution.

People are so focused on the mmap part, and the latency, that the usage is overlooked.

> The last couple of weeks I've been working on an HTTP-backed filesystem.

It feels like this is micro optimizations, that are going to get blocked anyway by the whole HTTP cycle anyway.

There is also the benchmark issue:

The enhanced CDB format seems to be focused on a read only benefits, as writes introduced a lot of latency, and issue with mmap. In other words, there is a need to freeze for the mmap, then unfreeze, write for updates, freeze for mmap ...

This cycle introduces overhead, does it not? Has this been benchmarked? Because from what i am seeing, the benefits are mostly in the frozen state (aka read only).

If the data is changed infrequently, why not just use json? No matter how slow it is, if your just going to do http requests for the directory listing, your overhead is not the actual file format.

If this enhanced file format was used as file storage, and you want to be able to fast read files, that is a different matter. Then there are ways around it with keeping "part" files where files 1 ... 1000 are in file.01, 2 ... 2000 in file.02 (thus reducing overhead from the file system). And those are memory mapped for fast reading. And where updates are invalidated files/rewrites (as i do not see any delete/vacume ability in the file format).

So, the actual benefits just for a file directory listing db escapes me.

> Was on a road trip last summer, around Norway, in VW id.Buzz. Charging time of 5min vs 20min doesn't matter. When you're on that long trip, you need time to eat, go to bathroom, walk a little so your legs/back doesn't hurt.

That is often the argument that i see but people forget a lot:

Fast charging if often 20 > 80% for 20min. If you want 80 to 100, its a lot less fast (think how your phone slows down).

While you can drive down to 5% of less, it can become a issue if you do not find a charger. When the summer vacation happened, a lot of north traffic goes south, over France ( and reverse).

What happened? People ended up waiting 15 a 25min at the chargers traffic jam in their cars. Then they charged up to 80% (because the stations had people to manage the flow), and needed to drive out (or pay more/fine).

This resulted that your range was already reduced by 20%. You ended up wasting 15 a 25 minutes stuck in your car. And with airco's on because its summer, so more battery drain. Aka, you did not really tank X km range, but X - waiting usage - 20% less limit.

Its always fun to compare a gasoline engine 5 min tank job, vs "not a issue, we need to stretch our legs", but when the reality of long trips that often coincide with vacation periodes... Yea, then the disadvantages of that statement come into play.

So the irony is that, a EV with a realistic 500km range, got hit with a 20% at chargers, then another hit from the waiting, o and you had no choice, it was fast charge or no charge.

I remember warning people to not see EVs based only on short trips or long "out of season" trips but also on those typical school vacation trips that many take. Ironic part, if you drove a hybrid, 5 min tank job for 100% fuel at the regular price.

> Here in europe cars have become so expensive compared to just 10y ago.

Check the ownership of the different brands and you will discover from the dozens of brands in Europe, there are really only a few companies making cars. And the "mother company" then shares chassis between models, engines between models. So you can have a a Lada with a Renault engine, with a Dacia chassis. We see 3 brands, but in reality, its one brand.

* Groupe Renault's owned brands include Renault, Dacia, Alpine, and Lada.

But wait ...

* Stellantis was formed from the 2021 merger of two major auto groups: PSA Group (which included Citroën and Peugeot) and FCA (Fiat Chrysler Automobiles). Therefore, the brands Stellantis owns are now a part of Citroën's parent company, which include brands like Peugeot, Fiat, Jeep, Chrysler, Dodge, and Opel.

You see the issue very fast. If you own many brands, do you compete against each other, or do you "stabilize prices for max profit". Same issue with other brands...

Its like in the US, where if you trace back the brands, like 3 or 4 companies, own like 90% of the known US food brands. And when you have only a few companies, gentlemen's agreements about not competition too much are a thing.

Hey, why are HDD not dropping in prices for the last 10+ years. Well, there are only 3 brands left (with two that are huge). Hey, why did SSD/NVME prices suddenly skyrocket. Well, only a few brands make the stuff, and when one decided to increase prices, all of a sudden the rest also followed. Strange is it not...

We are in the age of monopolies again, and that is linked to a lot of the price issues.

Your also forgetting that Chinese cars have a large import tariff.

Current rates are BYD (17%), Geely (18.8%), and SAIC (35.3%).

> And much like the US found out with rare earths:

Rare earths is not rare at all. China has a cornered the market on processing those minerals for cheap. The problem is that we in the West have no appetite to actually support companies to process those "rare" earths.

And the US is really the last country to even talk about batteries, when the arrested 300+ Koreans that worked to get a US battery plant going.

> They’re playing the long game and western nations seem unable or unwilling to do the same.

Its not just a government issue, its a companies issue. You think that car companies in Europe do not get subsidies? Have you ever looked at how much subsidies they get opening new plants or renovating plants?

Reality is that a lot of car companies enjoyed their little monopolie positions in Europe. Sure, we got a lot of brands, but in reality its a illusion of competition, as most brands are owned by a few big companies.

China had a different situation, where yes, there was subsidies from the government (and a lot of misuse of those subsidies), but you had 100's of car manufactures entering the markets. I mean, we talk about BYD often but BYD was not even a car manufacturer until 2005 ... The result is a strong competition between car makers, what resulted in a lot of technological development and a push to be better then the next guy(s). Constant copy and innovation work. Work to reduces prices to be able to compete better, as remember, their competitor also go those subsidies.

We have this issue with our car industry only to thank to our own car industry. Frankly, i am happy to see the Chinese enter the EU market, for the simply fact that it pushed the EU car makers to actually start innovating again and offering more to the clients.

The issue with competition to dead was a issue in new upcoming industries, like solar. Because their both sides started with the same level vs +subsidies. The car industry is a totally different beast with deep pockets and manufacturing capabilities. So lets not act like "poor EU car makers".

> It is network latency.

People do not realize for that fancy infinite storage scaling, that it means that AWS etc run network based storage. And that, like on a DB, can be a 10x performance hit.

Not the person you asked the question but i can answer this partially.

I see from the EV my parents bought, they got for 22k euro, a car that does 240km on paper. Now a few years later, you can buy a BYD for 26k, that does 450km (ext version) on paper. That same BYD is WAY more fancy, tech gadgets etc...

But that is comparing a lower market segment where this effect is stronger.

On the higher end market, where the Tesla's used to dominate, we see less movement as those cars already came with tons of tech, and large batteries.

Thing is, i am still not buying a EV despite them being better. The same arguments we had with EV disadvantages is still present today. I rather buy a second hand gasoline engine simply because of the convenience of the default 600 a 700km range on most vehicles, the 5 minute tank job, the lack of charging in the city, the prices despite the heavier depreciation in the second hand market are still worse.

Ironically the gasoline car we bought 10 years ago, for 7.5k, still sells today for 8k in todays market (added 50k kms on the car). So even if we take a cut, the actual gasoline engine depreciation is strangely less strong. I track several gasoline cars where i had a interest in from years ago, same phenomenon.

For some reason you expect gasoline cars to drop more in pricing because now in EU 17% of new cars sales are EVs and 36% are hybrid. So more gasoline cars on second hand market? No ... because people are also buying less cars because of the economy, more work from home etc. Resulting in actually less second hand cars as people hold on to their vehicles longer, waiting out this transition periode.

DB lookups + extra index are way more expensive then hardware assisted decoding.

If your UUIDv4 is cached, your still suffering from extra storage and index. Not a issue on a million row system but imagine a billion, 10 billion.

And what if its not cached. Great, now your hitting the disk.

Computers do not suffering from lacking CPU performance, especially when you can deploy CPU instruction sets. Hell, you do not even need encryption. How about making a simple bit shift where you include a simple lookup identifier. Black box sure, and not great if leaked but you have other things to worry about if your actual shift pattern is leaked. Use extra byte or two for iding the pattern.

Obfuscating your IDs is easy. No need for full encryption.

> I wish there was a way to buy windows 11 LTSC with a version of MSOffice stripped of copilot, OneDrive, and all the modern windows fluff.

Its called Windows 10 LTSC + cleanup script + any old office ... I am still using Office 2007 + pdf plugin. Thing uses no resources, still works perfectly.

Do you really have a reason to upgrade? Probably not... So unless Microsoft start forcing people with DX13 or whatever, there is no reason to give up on Windows 10.

> the browser based Word, PowerPoint, Excel, and Outlook are getting better all the time

Again, why? Same privacy issues anyway and AI will get pushed in those also, you can bet on it.

Just find a old version of Office, and use that. Or LibreOffice or whatever. Word, Excel etc are things that really did not change in the last 20 years beyond a ton of corporate fluff and spyware/"cloud".

> It’s wild how unreliable and complex windows has become

Because Microsoft wants to keep all the users locked into their eco system. So the more used you are to their products, the harder it is for the average person to jump ship (for any product, not just windows).

> I've never set up a highly available Postgres and Redis cluster on dedicated hardware, but I can not imagine it's easier than doing it in AWS which is only a few clicks

I haven ever setup a AWS postgres and redis, and know its more then a few clicks. there is simply basic information that you need to link between services, where it does not matter if its cloud or hardware, you still need to do the same steps, be it from CLI or WebInterface.

And frankly, these days with LLMs, its no excuse anymore. You can literally ask a LLM to do the steps, explain them to you, and your off to the races.

> I don't have to worry about OS upgrades and patches

Single command and reboot...

> Or a highly available load balancer with infinite scale.

Unless your google, overrated ...

You literally rent from places like Hetzner for 10 bucks a load balancer, and if your old fascion, you can even do a DNS balancing.

Or you simply rent a server 10x the performance what Amazon gives (for the same price or less), and you do not need a load balancer. I mean, for 200 bucks, you rent a 48 core 96 thread server at Hetzner... Who needs a load balancer again... You will do millions or requests on a single machine.

> but just, how can you stand possible breadcrumbs IN THE JAR!?

Bacteria can do a number on people... Kept telling the wife to not cross contaminate jars, and other products. But you know, she knows better...

Wife was eating some fish in can, puts spoon back (with lovely saliva bacteria), puts in fridge "i wil eat tomorrow", 2 days later she eats the leftover.

She enjoyed a hour+ of "fun" muscles contracting stomach cramps to the point it was almost hospital time. Learned her lesson, well, ... for fish.

Its always like "but i do not like to keep using fresh utensils". I am always: "the dishwasher does not care if it 20 or 40 utensils". Its the same amount of wash and cost. So stop trying to recycle utensils!

> I use GitHub Copilot Pro+ because this was my main requirement as well.

Frankly, i do not even get how people run out of 1500 requests. For a heavy coding session, my max is around 45 requests per day, and that means a ton of code / alterations and some wasted on fluff mini changes. Most days is barely in the 10 a 20.

I noticed that you can really eat your requests if you just do not care to switch models for small tasks, or constantly do edit/ask. When you rely on agent mode, it can edit multiple files at the same time, so your always saving tokens vs doing it yourself manually.

To be honest, i wish that Copilot had a 600 token version, instead of the massive jump to 1500. Other option is to just use the pay per request.

* Cheapest is Pro+, 1500 requests , year paid, at around 1.8cent / request * The 300 requests Pro, year paid is around 2.4cent / request. * The overflow tokens (so without subscription) is at 4 cent / request.

Note: The Pro and Pro+ prices assume you use 100% of you tokens. If you only use 700 tokens on the Pro+, your paying the same as the overflow 4 cent / request one.

So ironically, you are actually cheaper with a Pro (300 requests ) subscription, for the first 300, and then paying 4 cent / request between your 301 ~ 700...

I really do not want to respond to this but your information is so wrong, and clearly off.

Foreign investment was already in a down trend from 2013, investment kept somewhat stable in the 2015-2019 periode. Until a little unknown virus hit the world, resulting in a total shutdown / disrupted market.

There was a peak in 2021 as of the resulting reopening of the Chinese market, you know, kind of hard to do business when your companies employees need to spend a month in a hotel in quarantine. So there was a strong recovery in those years, a known effect from the 2020 pandemic and the eventual recovery.

I do not even know where you get the peak in 2022 when it was 2021. 2022 was barely on the level of te 2015-2019 periode. In 2023 it ended up crashing to 1990's level of investment.

And has been a negative FDI for the past 3 years. That means more investment is being pulled out of the country, not in anymore.

So unless you want to dispute macroeconomic data, your information is highly off. I am not even going to touch the whole IP comment "massively overblown" as there are plenty of reports on how it absolute not overblown.

No offense but your comments read like a propagandist. I let the readers make up their minds, as my information can be easily google on.

> suddenly things started breaking. We eventually tracked it down

Amateur level ... Around 2006, we enjoyed some clients complaining why information on our CMS was being duplicated.

No matter what we did, there was no duplication on our end. So we started to trace the actions from the from the client (inc browser, ip etc). And low and behold, we got one action coming from the client, and another from a different IP source.

After tracing back the IP, it was a anti virus company. We installed the software on a test system, and ... Yep, the assh** duplicated every action, inc browser settings, session, you name it.

Total and complete mimic beyond the IP. So any action the user did + the information of the page, was send to their servers for "analyzing".

Little issue ... This was not from the public part of our CMS but the HTTPS protected admin pages!

Sure, our fault for not validating the session with extra IP checks but we did not expect the (admin only) session to leak out from a HTTPS connection.

So we tried to see if they reacted to login attempts at several bank pages. O, yes, they send the freaking passwords etc. We tried on a unused bank account, o, look, it was duplicating bank actions (again, bank at fault for not properly checking the session / ip).

It only failed on a bank transfer because the token for authorization was different on their side, vs our request.

You can imagine that we had a rather, how to say, less then polite talks / conversation with the software team behind that anti-virus. They "fixed it" in a new release. Did they remove the whole tracking? Nowp, they just removed the code for the session stealing if the connection was secure.

O, and the answer to why they did it. "it a bug" (yea, right, your mimic a total user behavior, and its a "bug"). Translation: Legal got up their behinds for that crap and they wanted to avoid legal issues with what they did.

Remember folks, if its free your the product. And when its paid, you are often STILL the product. And yes, that was a paid anti-virus "online protection". And people question why i never run any anti-virus software beyond a off-line scan from time to time, and have Windows "online" protections disabled.

Companies just can not stop themselves from being greedy. Same reason why i NEVER use Windows 11... You expect if you paid for Windows, Office or whatever, to not be the product, but hey ...

> China has been moving towards more openness to foreign investment over time

After they lost most of those same foreign investors that got tired of handing over their IP, the "local partnering" (what often resulted in state/court backed judgements against the foreign investors), ...

The disasters "wolf warrior" politics, combined with the above mentioned issues. And the stringent monetary policies limitation on moving money out of China did not help (translation: We want you to put money in, not being able to pull it out).

This has resulted in a reduction of foreign investment by almost 70% (given the 15 year average). What used to be easily be 200 a 250B investments per year, has now dropped to barely 100B, and is still declining.

> randomized grains appear smoother to the human eye than clean uniform pixels

Does this explain why i dislike 4K content on a 4K TV? Where some series and movies look too realistic, what in turn gives me a amateur film feeling (like somebody made a movie with a smartphone).

Take a chill pill sudodevnull ... As i stated before, turn down the rhetoric. Your not helping yourself with these "!" reactions.

> its not bait and switch, its main has features we are willing to continue to support given we did a whole rewrite and this is what we think you should use. Don't like it? Fork it, code is still there. I hope your version is better!

It sounds like your are the dev of Datastar...

Let me give one piece of advice. Drop the attitude because this is not how you interact in public as the developers of a paid piece of software.

You can get away with a lot when its free/hobby project, but the moment you request payment, there is a requirement for more professionalism. These reactions that i am reading, will trigger responses that will hurt your future paycheck. Your already off on a bad start with this "bait and switch", do not make it worse.

I really question your future client interactions, if they criticize your product(s) or practices.

> I hope your version is better!

No need for Datastar, my HTMX "alternative" has been in production (with different rewrites) over 20 years. So thank you for offering, but no need.

The problem is that the developer of datastar did a bait and switch. Releasing the beta for free, and then removing features into a pro version with a price tag.

Nothing wrong with people making money on their software but you need to make it clear from the start, that it will be paid software and what price range.

Bait and switch is often used to get people to use your software, you spend time into it, and then if you need a Pro feature, well, fork up or rework your code again. So your paying with your time or money. This is why its nasty and gets people riled up.

Its amazing how many people are defending this behavior.