_the_inflator

Former member of the Senior Management for a huge global bank operating in Germany.

1. These averages are an average. Rarely are they paid. I don't know where these numbers are from but the these numbers are for top earners in good paying industries.

There are junior, normal and senior devs, sometimes above is a lead dev. Business maybe has 2-3 leads that govern a head count of dozens, and many working off-shore.

It is called blended average. People who rank in 107k are paid including bonuses. And rarely make any further wage jumps. The dev who is paid 200k fixed in Germany is a myth, not even close to 120k.

Germany hates IT and devs. Business comes first, and anything and everything regarding IT is considered a cost factor, a necessary burden.

2. Freelancers

Freelancers rake in more money due to tax deductions. So a day rate of 800-1000EUR per month is a pay off and freelancers have to take care of the taxes etc.

Companies on the other hand can deduct those "investments" and roughly pay half the price in the end. So, this is the problem with employment: you are a cost factor and you are a liability from the beginning due to worker rights etc.

Germany punishes workers as well as freelancers. But I could not raise payment as I saw fit. Germany means yearly total times 1,25 = total cost for a company ("employers' part of the tax" as sugar coating for taxes above 60%).

100 workers with a yearly raise of 2-5% in payments amount exponentially and are a risk factor especially considering the fact that productivity rarely increases as well as value to the company on average. So ever increased payments over a span of 10-20 years.

That is the problem with the system: socialism favors keeping you poor. Don't steal people and companies their hard earned money, lower taxes, allow for higher payments at will not by "number comparison and increase to all" and let companies workers decide where they spend their bucks.

3. Conclusion You might wonder how much the guys really perceive on their bank account in the end. It may be lower than you think. Rule of fist: if you do not immediately make the switch to stay in the US you have lots of cuts along the way: company share, taxes.

Before I went to Senior Management I was totally on your page. Now knowing both sides I see the matrix so to say. I am still on your side however business doesn't benefit from all the taxes.

I agree.

And most people on the behavioral side do too, but not at the cognitive level. EU is the best example with EU AI Act, strict data regulation as well as privacy rights, on the other hand demanding that Apple does serve the EU with AI.

I have mainly one distinction: the state is the worst protector of your data and the most ruthless gatherer of all your details.

Opposite to the open sourcing of your data in the end by the state are private companies who live by your data but do this for 20+ years - battle tested protection and hardening against malicious hackers.

Security is their business while security for the state is a cost factor.

Being at the mercy of some ignorant politician is not the best way to talk about data security.

Berlin, Denmark - those are the known one. And there are many more to come.

And regarding cognitive dissonance: politicians demanding high standards and punishing data loss ruthlessly on the one hand, giving oneself a pass on a hack is nothing to increase trust into the system.

X got fined for a missing blue mark. Berlin? Denmark? Others?

A second aspect is that the average guy doesn’t get that part of the whole spectrum must be the degooglers, the home server guys.

So it is relatively easy to get data on them as well just by filtering out the other data.

In other words: 95% not doing degoogling makes for a great small sample of 5%. Negating and interpolating other demographic and psychographic factors and you get a great way of gaining insights.

And remember: being the one who is not using google when being around other guys who do - magic.

So my idea is simple: what’s in it for me, and the state offers way lower value than Google and co.

Pick your fate.

I can confirm that. Barcodes are in increased demand, since they offer an almost universal API to the physical world, kind of the not-IoT devices.

Jevons paradox in full force.

Remedial Math - I love it.

Wisdom of the crowd has finally proven to be more valuable than some obscure title in academia. Crowds are relentless and ruthless, kind of a red team instead of a cronyism.

PS: The infamous replication crisis deserves a honorable mention here.

I somehow think of it as a warning or even intentional peacocking towards foreign state hackers.

700 agents cost quite some money. 100 agents per 24 hour stint using Astra on xHigh cost somewhat between 12-42k USD, depending on the usage intensity.

I don’t know how many raw time went into this but there was a probing face before the attack itself.

So just going by seven days and 500 agents fully working on this on average amounts to a bill somewhere between 400-1.2 Mio USD.

I believe it was intentional but of course I don’t know which intention exactly.

There ain’t no accidental escape because then it would have read OpenAI lost over their agents.

The whole scenario reads as a classic movie where a hero has under the most dire circumstances to survive and fulfill his mission no matter what.

On the other hand there was a final authority under which the system of agents flocked.

Huggingface itself seems like a perfect victim.

And to be honest: I don’t believe that this was the first time. I strongly believe that there were and are countless of smaller sites hacked but not harmed that we don’t know off.

Why is HF perfect?

Because there will be countless of independent security analysts who will bend their minds on the incident.

OpenAI is provided with the data of dozens of blue teams and what is desperately needed? Data of security measures and possible ways to reconstruct the incident.

I think this is genius, and just watching on neutral this is such a fantastic action OpenAI pulled off.

Imagine what the GPT 7 “Haha-Huggingface” model is going to do then on a regular basis.

State hackers and rough states were put on notice that this is a new level of the war of attrition.

Exciting to watch but simply meant silent invasion. Open invasion then might be executed by Robots, but let’s stick with the fascination mode at this time.

Anthropic it is game over when they passed OpenAI at the beginning of the year.

Now it is revenge time and OpenAI kind of is trolling Anthropic by simply going into a price war with impressive performance.

OpenAI is doing a decent job this year after they recovered. Anthropic needs to offer more payment options and be clear about token usage. The warnings I got when switching to Fable 5.1 felt like a thread. I bet more and more on OpenAI since I don’t feel robbed by them.

Claude adapts to OpenAI’s surprising move to simply deliver better performance than Fable 5.1, better tools as well as featuring very low pricing.

Fable 5.1 literally was a money grabber. While I liked the results, tokens were burned so hard it was embarrassing, while Astra seemed to not care.

Also Claude makes it very hard to pay for additional token budgets, allowing only credit cards. I don’t use mine anymore since I don’t need it in everyday life I was dumbfounded.

So Anthropic is just copying OpenAI so to say, matching them and essentially with Opus 5.5 being Fable 5.1 in disguise, all they do is reduce costs.

Competition works.

To me it is just a bittersweet irony: people who don’t know what a book is for 30-40 years or in earlier days got the RTFM (gif smiley not included) treatment dump one liners to produce a giant load of text corpus that needs to be processed by an AI to get the gist out of it with sometimes quite some refinement by asking many questions.

So there it is: cruel people who don’t even read their own prompts while happily believing others read their text load and pat the genius on the shoulder for being such a great guy in writing, while in essence it turned the RTFM movement on its head: UUOAI - “Use your own AI.”

People cannot communicate better by using the digital equivalent of logorrhea.

What aspect of Math are you referring to? Writing proofs intuitively is quite a claim, matrix calculus may have a say here, too.

But please elaborate a bit on your statement.

And what about the governments like Berlin for example? Massive data breach, and guess what happens? Nothing to those who are responsible for the breach.

So even though this is Korea, it is modern hypocrisy. Companies have to comply to more and more complicated regulation, while those who govern the states get a free pass.

If the Berlin incident remotely had happened to any private company - hell would have been loose.

Berlin reduced the IT budget especially regarding maintenance and security massively over the years. In fact, what came to light - CCC talk as a reference besides others - sounds so embarrassing, that all companies should get a bonus payment whenever they get hacked.

Sometimes, sometimes not. It depends.

My mantra is still "Don't give users what they want, but what they need."

Product Design is a one to many relationship, while users are one to one tightly knit into their app. It isn't to discredit their input, it just needs to make sense in the whole picture:

What if the one guy who reports 70% of all the features that would be cool and necessary to have, but 99% aren't using the product like he does? Is there another way to achieve that he wants even better, easier, faster, more reliably etc.?

What if a a/b test showed, that they outright ignore or complain about such new features?

Not building sounds so cool and rebellious, but it isn't about the not to build but overall product strategy and product itself that is the decision maker so to say. Because, you build a product in the first place, didn't you?

Product Design and improvement is brutally hard. Backwards compatibility is the one thing you have to take into account all times. This can create the greatest friction there is and kill products.

So how do you introduce a better way, that breaks with old traditions? Don't build it or use brainpower to solve the puzzle?

Heads up to all the Platform Leads, Founders, Product Designers - tough world, but if it works it is heaven on earth. :D

I worked at a MS "fanboy" company around 2011-2013. Highly competent guys, really Senior Devs, C#, MS SQL, as well as using graph data - with one distinction: it must be MS.

Open Source? No way. Git? No, they relied as die hard MS believers on the MS software called Team Foundation or something like that, that was integrated into Visual Studio Pro - sorry, I forgot about it, I considered it kind of bloat and outdated. Also I couldn't stand the nomenclature. A project was called "Solution" - I died inside, because this sounded like utter nonsense to me, because how do they know it would be one in the end?

While JetBrains as well as Linux quickly iterated through everything and got traction as well as a cadence that overall kind of was paced around sprint cycles that lasted two or four weeks, the company finally started to break up with project management and implemented Scrum.

As the JavaScript guy, the only one, because a customer wanted a SaaS "solution" but with static web content this wasn't really dynamic. I knew one of the founders who was a managing partner and he asked me to join as Web Developer.

Overall, all were very skeptical towards me because how could someone bet on JavaScript at the time? Well I turned the argument around and said the same about C# with its closed source walled garden approach to everything relying on MS to solve their problems with no way of giving feedback while there was no real release cycle and roadmap available - hopium and copium.

Statically typed languages for the win they said, blabla. I wasn't against static types, but did pure magic in JS, that they saw me as magician and I got some fans and I found one team mate who wanted to be coached by me on JS, Ajax and stuff.

So, there you have it. History.

I think there are pros and cons to any approach as always. Both language suffer from feature creep.

C# is still tightly knit into some products from MS and there are some backwards compatibility issues to take care of that limit certain progress and need substantial change.

Java isn't that way and was near dead and went OS. That's why they moved to the current model. Former versions were also hardly changed, have a look at everything before Java Version 12 or so.

Java wasn't community driven all the time.

So, C# has its merits, TypeScript for the win, so MS won over JavaScript ironically but only on the outside.

I shocked my MS fanboy colleagues when I really gave them a shock therapy regarding security when they mocked me with the examples given by MS why JavaScript was so bad and C# would beat it. We all know the infamous type coercion examples with mixed types, arrays etc.

So I shocked them with eval function of course but then gave them nightmares and mental overload with Function.prototype.toString and new Function() trickery.

It blew their mind, there was nothing remotely available in their world. Not introspection, nothing.

JavaScript was kind of assembler like I said. Highly flexible, you need to use modules, like jQuery did but have to build your own.

So TypeScript used exactly this flexibility: compiling to JavaScript. A metalanguage.

For the true insider, JavaScript won.

I like it. I had a friend at school who came from SA and spoke Afrikaans - so I know how a natural sounds like.

For language learning, translation, research - it is fantastic.

It works very well for certain dialects as part of a local language heritage as well: From Welsh to Bavarian - it is a joy to really get approval from people used to these dialects as being really accurate.

Happpy times! :)

I don't share your opinion.

With no deeper knowledge to the matter, it is pretty senseless to speculate.

"Numbers from GPT Astra" - well, well...

Currently roughly 1000 engineers are working on Google Chrome.

GTA 6 has around 6000 developers working on GTA 6 since 2019.

So sounds a bit off what you allegedly researched. But, hey, AI slop is real.

DevOps and infrastructure, Security. Internal tools is correct, but speed, video quality as well as sound seems so easy. Google is responsible for many compression algorithms and standards - byproducts so to say. And they are top notch.

I have insights into the matters, and a lot goes into behind the scene optimization. Some specialists in the pre-AI phase were tasked to do nothing else but optimize build tools but in the sense of tweaking every millisecond out of it. This guy saved Google Billions over the years.

Others work in compiler building etc.

There are a couple of hints in the books by Google engineers but I am regularily blown away during meetings what really highly inspiring things are there to discover as well as presented.

Another factor: Google is spread around the globe. Regulation, law. Believe me, having over a billion users per app is mindblowing.

And it is 24/7/365. And support. And research.

Google is a company I admire. Technical provess at its finest.

What is your use case? Your business model?

There is no "one harness". I don't feel your pain. I know, what quality has to be ensured, what's within boundaries and what not etc.

I work for price sensitive industries and here some information has to be factual correct. The error rate is 0% so far, 100% accuracy.

So what are you trying to harness?

Tell us everything, what happens on tour stays on tour - sure. Just approve that your data is - of course - anonymously used to train AI and regarding intelligence agencies or AI Acts: well, it is not us, it is them. Law is law, you know. We won't tell you anyway.

There is no privacy left. And to be honest, people worshipping their phones all day but complaining about privacy is the biggest WTF I have. This doesn't match.

Also I am a Mathematician and know statistics very well. The person, who is not on Facebook, but on WhatsApp, who works at X and lives in Y - we got you my friend.

You moved away? No problem! Someone photographed you casually, you weren't even the main character, we ignored you, and send it in a group chat you know nothing about. Well, we know all the others and can relate to anyway. And then? Well AI can very well predict you aging - easy. We have you already. Sleep well and dream of being unknown. Because... How do you know what you don't know?

I gave up on privacy because there is none anymore. Selling and publication of your data is a whole different story. And that's what Apple means by circumventing the rest.

"It is all just for your safety." and other excuses.

Don't get me wrong, I think it is viable by Anthropic to do so, it is a business decision. Nevertheless the reasons sound contrived and mask the demand for such a lucrative business opportunity, if you abstract it away from activists.

This is not about moral, only about reality.

Maybe Anthropic had a nice chat with the administration and/or think, that this was something, that interests them on a psychological level: predicting human behavior.

Look at Muse from Meta: Meta is doing exactly the same I bet, to optimize their ad business.

Anthropic will sooner or later feature ads, and maybe this is a way to distract from this usage possibility.

Remember, the "No! We don't work with the government, we want privacy for our customers." mantra won't match with "We need to make money and surveillance of our customers would be helpful as well as a cool topic for our engineers to work on."

Anthropic itself will be yet another normal Big Tech company, all, even those who boasted "Don't be evil" or encrypt like hell their smartphones allegedly made good bucks in China.

It is ok, no problem with that by Anthropic just with their hypocrisy in my point of view.

Maybe check in the other direction as well: Astra to Fable.

I frequently simply let one of the three review what something that looks like awesome output by one AI gets totally annihilated by the other.

Finished outputs are easier to improve than bend a LLM to produce stuff like that in my observation.

Same with Gemini.

I yet have to find out how to handle this, whether I let agents check themselves and if on what process step.

Tweaking is hard.

I agree with your conclusion I am a huge ChatGPT and Codex fan, Gemini has to many infrequent quality changes when new models arrive ranging from great improvement to WTF.

ChatGPT seems to get scaling well while Claude still feels unstable, unclear usage statistics. Really weird.

Tough call I use all three.

Hayes modems were pretty much the crème de la crème back then.

I never heard of them since the CompuServe and AOL times.

Exactly. See TikTok trouble as example and quite honestly, try a local open source LLM and ask it to use profanity, paint nudes - the LLM doesn’t answer the question of it is from OpenAI or Google.

The thing is that needs more attention is reverse engineered a LLM which is highly fascinating. I tried it, but it seems I am not there yet to put it mildly. It requires serious effort.

I am just speculating but can LLMs be sleepers? You write software and it seeds traces here and there under certain conditions that pose a serious security risk.

Or a kill switch?

I don’t know. I distrust Chinese LLMs but even more due to training data.

It is after all not a Western model. Different biases and the might be subtle but nevertheless substantial.

In short: no open source LLM may be usable without additional Finetuning for certain valid use cases.

The real value is versioning and autonomy as well as lot more stable answering despite model rot.

Also testing and the supporting systems are easier to maintain.

It is mainly an infrastructure challenge.

How large are your projects? What kind of?

Just curious. No right or wrong answer possible.

Did you study math? I did and I cannot relate to your comparison.

Mathematics isn’t about calculating numbers but formulas and proves.

I have to put it politely: until the advent of AI no machine ever “calculated” a formula for me, because I didn’t have to calculate anything.

It is like driving a Tesla and laughing about the manufacturer when he mentions how hard it is to get the product right, because you have an easy ride.

I can relate to the article’s AI depression sometimes.

Whatever comes to your mind, AI makes it happen in a couple of minutes in most cases no matter what you imagined.

You invest days and the AI just build it is a couple of minutes way better than you and your 10 teams together did before.

So much hard work and sacrifice - and now? Poof - there it is.

It is nice but also depressing. And not being challenged is the actual problem.

Serious question: how widespread and actually used in production is Zod?

Honestly, it made things worse for me. JavaScript isn’t a statistical typed language and TypeScript already is something, where due to compiler issues I just use “any”.

Zod has its conventions and JavaScript ain’t Java.

So Zod is also hard to read and debug. It is type conversion to type conversion - bloat, not JavaScript.

Even Pydantic looks tamed compared to Zod.

So how do you use it? Which conventions? Only for validation at certain steps? APIs?

I stopped using TanStack because it got to meta for me and AI as well.

Tests are needed anyway, I rely on TypeScript only now.

Hotkey usage means speed. Ironically you hinder people using their capabilities to the fullest. It also works in the other direction.

And I agree with you: Vibe Coding literally automagically includes it, and I had the notion, that you must actively advise an LLM to ignore Web Accessibility.

I understand that developers in the traditional sense were shunning the topic, but via LLM?

I never found it easier. Of course here and there you have to lead the LLM in the right direction as usually.

This is an interesting ACID test, because at its root hotkeys as they are usually called, are a graph structure.

And LLMs struggle a lot with kind of complexity - as we all do at a certain level, if we don't visualize it i.e. using paper and pencil to paint a picture. Then it is the smoothest thing ever.

But again, then you test it - nice loop for a LLM.

I had no success so far using LLM to sketch my database schema for graphs. I do it by hand and recommend it. Later changes can be nasty and ugly. Changes will occur, but even dry runs make Claude, ChatGPT etc. go rookie level in what they do.

Processing a table and creating a graph structure out of it - no way. And so before a LLM could work on it, it needs lots of preparation. Or we create a new benchmark category.

Sequential work is easy for LLMs now, but Graphs are domains, were they massively struggle.

Nice idea.

You are absolutely right. In fact with the hit on men’s sexual ability sometimes or often times you could consider this a hit hot (very sinister framing) on the practical side, having in the best case no anxiety and being able to meet people and how are you supposed to date? Blue solution?

Men’s inability to perform is causing a huge hit on their psyche. So in any case consider wisely, I heard many accounts of very disappointed men.

It may help women, but that’s a different thing.

For men - only in rare cases to be considered if ever.

I don't get the EU reference in any sense:

"The company was founded in 2016 by French entrepreneurs Clément Delangue, Julien Chaumond, and Thomas Wolf in New York City, originally as a company that developed a chatbot app targeted at teenagers. ..." Wikipedia

Besides that: which EU and sovereignty doesn't really rhyme, given data protection, energy costs, AI restrictions and no vital start up scene at all.

Mistral is heavily state funded via Banks that are predominantly owned by the French state, so I don't get it.

Also name one global success that doesn't came from US? China? Well, second place maybe, but second in any sense.

Nevertheless, kudos to huggingface. Not only to the founders, but to the ecosystem and I guess it got harder by the day to finance all the hosting and only hope for the future, that Nvidia will help HF to thrive, at least not put burdens on them.

GitHub is still alive, and hope that HF will benefit from more AI power.

Nvidia doesn't have any disadvantages in my playbook, because hardware wise, there is no one to beat them in any regard and Apple will always stay nichy.

So well played - hopefully.

It could be worse or even a shut down otherwise, who knows. So for me it sounds cool and I am glad that EU isn't involved in any regard.

As the AI Act demands full disclosure under really easy to match requirements into your inner workings, this is quite the opposite of freely sharing information.

RAG is art. I have a very straight forward setup that is highly modular.

RAG is routing and decision making.

I found so much joy in achieving the best results given the requirements than simply hoping for the best with the cool kid called vector db and embeddings.

I agree with you.

Depending on the context and required output I decide how to orchestrate a multitude of specialized modules that produce the best specific result to gain a universally usable system.

It maintains itself.

Also live updates need reruns and rebuilding certain indexes. Everything is highly dynamic but in a deterministic way.

I found my niche with RAG selling and I build them myself.

I take pride in them.

So many look at the technology but not on the required output. It takes hours of talking to people to get an idea of what they need.

And there are regulated businesses where certain information is required to be always factual correct - pricing for example.

Vector search becomes a liability for this use case.

So naturally you have to reconsider your system: mixing factual with probabilistic content and how to make sure, it hits always certain quality benchmarks and on the other hand doesn’t fail others.

I love this kind of stuff.

And there is personal information etc.

Using modules is the key. Orchestration is really fun but I have to admit, not for the faint of heart.

And ever changing parts: LLMs, or restrictions to be matched liked autonomously working - I love RAG.

It gave me back the joy of developing. In fact I never had so much phun before, because it is also “team work”: I am not programming, I am managing a product.

I was in Senior Management of a top tier international bank and besides that build the only ever working platform or IT transformation called dbCORE and overlooked 13 teams with 120 developers.

RAG gives me dbCORE vibes so to say.

Good luck and fun with your RAG systems.

The trick are tax deductions.

I wouldn’t recommend buying any bare metal unless money is a second thought or you can fully deduct the price.

Most often in the end you pay half the price then. Depending on the write offs you could even make some bucks out of it.

Or buy and lease. Under certain circumstances the hardware costs you nothing.

But you need money to save money. And a company.