Member states holding the presidency work together closely in groups of three, called 'trios'. This system was introduced by the Lisbon Treaty in 2009. The trio sets long-term goals and prepares a common agenda determining the topics and major issues that will be addressed by the Council over an 18-month period. On the basis of this programme, each of the three countries prepares its own more detailed six-month programme.

For us personally, this means that we no longer use generative AI – neither for private nor professional purposes.

To Gen or Not To Gen: The Ethical Use of Generative AI 33 minute read This blog entry started out as a translation of an article that my colleague Jakob and I wrote for a German magazine. After that we added more stuff and enriched it by additional references and sources. We aim at giving an overview about many - but not all - aspects that we learned about GenAI and that we consider relevant for an informed ethical opinion. As for the depth of information, we are just scratching the surface; hopefully, the loads of references can lead you to diving in deeper wherever you want. Since we are both software developers our views are biased and distorted. Keep also in mind that any writing about a “hot” topic like this is nothing but a snapshot of what we think to know today. By the time you read it the authors’ knowledge and opinions have already changed. Last Update: December 8, 2025. Table of ContentsPermalink Abstract About us Johannes Link Jakob Schnell Introduction Ethics, what does that even mean? Clarification of terms Basics Can LLMs think? What LLMs are good at GenAI as a knowledge source GenAI in software development Actual vs. promised benefits Harmful aspects of GenAI GenAI is an ecological disaster Power Water Electronic Waste GenAI threatens education and science GenAI is destroying the free internet. GenAI is a danger to democracy GenAI versus human creativity Digital colonialism Political aspects Conclusion Can there be ethical GenAI? How to act ethically AbstractPermalink ChatGPT, Gemini, Copilot. The number of generative AI applications (GenAI) and models is growing every day. In the field of software development in particular, code generation, coding assistants and vibe coding are on everyone’s lips. Like any technology, GenAI has two sides. The great promises are offset by numerous disadvantages: immense energy consumption, mountains of electronic waste, the proliferation of misinformation on the internet and the dubious handling of intellectual property are just a few of the many negative aspects. Ethically responsible behaviour requires us to look at all the advantages, disadvantages and collateral damages of a technology before we use it or recommend its use to others. In this article, we examine both sides and eventually arrive at our personal and naturally subjective answer to whether and how GenAI can be used in an ethical manner. About usPermalink Johannes LinkPermalink … has been programming for over 40 years, 30 of them professionally. Since the end of the last century, extreme programming and other human-centred software development approaches have been at the heart of his work. The meaningful and ethical implementation of his private and professional life has been his driving force for years. He has been involved with GenAI since the early days of OpenAI’s GPT language models. More about Johannes can be found at https://johanneslink.net. Jakob SchnellPermalink … studied mathematics and computer science and has been working as a software developer for 5 years. He works as a lecturer and course director in university and non-university settings. As a youth leader, he also comes into regular contact with the lives of children and young people. In all these environments, he observes the growing use of GenAI and its impact on people. IntroductionPermalink Ethics, what does that even mean?Permalink Ethical behaviour sounds like the title of a boring university seminar. However, if you look at the wikipedia article of the term 1, you will find that ‘how individuals behave when confronted with ethical dilemmas’ is at the heart of the definition. So it’s about us as humans taking responsibility and weighing up whether and how we do or don’t do certain things based on our values. We have to consider ethical questions in our work because all the technologies we use and promote have an impact on us and on others. Therefore, they are neither neutral nor without alternative. It is about weighing up the advantages and potential against the damage and risks; and that applies to everyone, not just us personally. Because often those who benefit from a development are different from those who suffer the consequences. As individuals and as a society, we have the right to decide whether and how we want to use technologies. Ideally, this should be in a way that benefits us all; but under no circumstances should it be in a way that benefits a small group and harms the majority. The crux of the matter is that ethical behaviour does not come for free. Ethics are neither efficient nor do they enhance your economic profit. That means that by acting according to your values you will, at some point, have to give something up. If you’re not willing to do that, you don’t have values - just opinions. Clarification of termsPermalink When we write ‘generative AI’ (GenAI), we are referring to a very specific subset of the many techniques and approaches that fall under the term ‘artificial intelligence’. Strictly speaking, these are a variety of very different approaches that range from symbolic logic, over automated planning up to the broad field of machine learning (ML). Nowadays most effort, hype and money goes into deep learning (DL): a subfield of ML that uses multi-layered artificial neural networks to discover statistical correlations (aka patterns) based on very large amounts of training data in order to reproduce those patterns later. Large language models (LLM) and related methods for generating images, videos and speech now make it possible to apply this idea to completely unstructured data. While traditional ML methods often managed with a few dozen parameters, these models now work with several trillion (10^12) parameters. In order for this to produce the desired results, both the amount of training data and the training duration must be increased by several orders of magnitude. This brings us to the definition of what we mean by ‘GenAI’ in this article: Hyperscaled models that can only be developed, trained and deployed by a handful of companies in the world. These are primarily the GenAI services provided by OpenAI, Anthropic, Google and Microsoft, or based on these services. We also focus primarily on language models; the generation of images, videos, speech and music plays only a minor role in this article. Our focus on hyperscale services does not mean that other ML methods are free of ethical problems; however, we are dealing with a completely different order of magnitude of damage and risk here. For example, there do exist variations of GenAI that use the same or similar techniques, but on a much smaller scale and restricted domains (e.g. AlphaFold 2). These approaches tend to bring more value with fewer downsides. BasicsPermalink GenAI models are designed to interpolate and extrapolate 3, i.e. to fill in the gaps between training data and speculate beyond the limits of the training data. Together with the stochastic nature of the training data, this results in some interesting properties: GenAI models ‘invent’ answers; with LLMs, we like to refer to this as ‘hallucinations’. GenAI models do not know what is true or false, good or bad, efficient or effective, only what is statistically probable or improbable in relation to training data, context and query (aka prompt). GenAI models cannot explain their output; they have no capability of introspection. What is sold as introspection is just more output, with the previous output re-injected. GenAI models do not learn from you; they only draw from their training material. The learning experience is faked by reinjecting prior input into a conversation’s context 4. The context, i.e. the set of input parameters provided, is decisive for the accuracy of the generated result, but can also steer the model in the wrong direction. Increasing the context window makes a query much more computation-intensive - likely in a quadratic way. Therefore, the promised increase of “maximum context window” of many models is mostly fake 5. The reliability of LLMs cannot be fundamentally increased by even greater scaling 6. Can LLMs think?Permalink Proponents of the language-of-thought hypothesis 7 believe it is possible for purely language-based models to acquire the capabilities of the human brain – reasoning, modelling, abstraction and much more. Some enthusiasts even claim that today’s models have already acquired this capability. However, recent studies 8 9 show that today’s models are neither capable of genuine reasoning nor do they build internal models of the world. Moreover, “…according to current neuroscience, human thinking is largely independent of human language 10” and there is fundamental scientific doubt that achieving human cognition through computation is achievable in practice let alone by scaling up training of deep networks 11. An example of a lack of understanding of the world is the prompt ‘Give me a random number between 0 and 50’. The typical GenAI response to this is ‘27’, and it is significantly more reliable than true randomness would allow. (If you don’t believe it, just try it out!) This is because 27 is the most likely answer in the GenAI training data – and not because the model understands what ‘random’ means. ‘Chain of Thought (CoT)’ approaches and ‘Reasoning models’ attempt to improve reasoning by breaking down a prompt, the query to the model, into individual (logical) steps and then delegating these individual steps back to the LLM. This allows some well-known reasoning benchmarks to be met, but it also multiplies the necessary computational effort by a factor between 30 and 700 12. In addition, multistep reasoning lets individual errors chain together to form large errors. And yet, CoT models do not seem to possess any real reasoning abilities 13 14 and improve the overall accuracy of LLMs only marginally 15. The following thought experiment from 16 underscores the lack of real “thinking” capabilities: LLMs have simultaneous access to significantly more knowledge than humans. Together with the postulated ability of LLMs to think logically and draw conclusions, new insights should just fall from the sky. But they don’t. Getting new insights from LLMs would require these to be already encoded in the existing training material, and to be decoded and extracted by pure statistical means. What LLMs are good atPermalink Undoubtedly, LLMs represent a major qualitative advance when it comes to extracting information from texts, generating texts in natural and artificial languages, and machine translation. But even here, the error rate, and above all the type of error (‘hallucinations’), is so high that autonomous, unsupervised use in serious applications must be considered highly negligent. GenAI as a knowledge sourcePermalink As we have pointed out above, LLMs cannot differentiate between true and false - regardless of the training material. It does not answer the question “What is XYZ?” but the question “How would an answer to question ‘What is XYZ?’ look like?”. Nevertheless, many people claim that the answers that ChatGPT and alike provide for the typical what-how-when-who queries are good enough and often better than what a “normal” web search would have given us. Arguably, this is the most prevalent use case for “AI” bots today. The problem is that most of the time we will never learn about the inaccuracies, left-outs, distortions and biases that the answer contained - unless we re-check everything, which defies the whole purpose of speeding up knowledge retrieval. The less we already know, the better the “AI’s” answer looks to us, but the less equipped we are to spot the problems. A recent by the BBC and 22 Public Service Media organizations shows that 45% of all “AI” assistants’ answers on questions about news and current affairs have significant errors 17. Moreover, LLMs are easy prey for manipulation - either by the service providing organization or by third parties. A recent study claims that even multi-billion-parameter models can be “poisoned” by injecting just a few corrupted documents 18. So, if anything is at stake all output from LLMs must be carefully validated. Doing that, however, would contradict the whole point of using “AI” to speed up knowledge acquisition. GenAI in software developmentPermalink The creation and modification of computer programmes is considered a prime domain for the use of LLMs. This is partly because programming languages have less linguistic variance and ambiguity than natural languages. Moreover, there are many methods for automatically checking generated source code, such as compiling, static code analysis and automated testing. This simplifies the validation of generated code and thereby gives an additional feeling of trust. Nevertheless, individual reports on the success of coding assistants such as Copilot, Cursor, etc. vary greatly. They range from ‘completely replacing me as a developer’ to ‘significantly hindering my work’. Some argue that coding agents considerably reduce the time they have to invest in “boilerplate” work, like writing tests, creating data transfer objects or connecting your domain code to external libraries. Others counter by pointing out that delegating these drudgeries to GenAI makes you miss opportunities to get rid of them, e.g. by introducing a new abstraction or automating parts of your pipeline, and to learn about the intricacies and failure modes of the external library. Other than old-school code generation or code libraries prompting a coding agent is not “just another layer of abstraction”. It misses out on several crucial aspects of a useful abstraction: Its output is not deterministic. You cannot rely on any agent producing the same code next time you feed it the same prompt. The agent does not hide the implementation details, nor does it allow you to reliably change those details if the previous implementation turns out to be inadequate. Code that is output by an LLM, even if it is generated “for free”, has to be considered and maintained each time you touch the related logic or feature. The agent does not tell you if the amount of details you give in your prompt is sufficient for figuring out an adequate implementation. On the contrary, the LLM will always fill the specification holes with some statistically derived assumptions. Sadly, serious studies on the actual benefits of GenAI in software development are rare. The randomised trial by Metr 19 provides an initial indication, measuring a decline in development speed for experienced developers. An informal study by ThoughtWorks estimates the potential productivity gain from using GenAI in software development at around 5-15% 20. If “AI coding” were increasing programmers’ productivity by any big number, we would see a measurable growth of new software in app stores and OSS repositories. But we don’t, the numbers are flat at best 2122. But even if we assume a productivity increase in coding through GenAI, there are still two points that further diminish this postulated efficiency gain: Firstly, the results of the generation must still be cross-checked by human developers. However, it is well known that humans are poor checkers and lose both attention and enjoyment in the process. Secondly, software development is only to a small extent about writing and changing code. The most important part is discovering solutions and learning about the use of these solutions in their context. Peter Naur calls this ‘programming as theory building’ 23. Even the perfect coding assistant can therefore only take over the coding part of software development. For the essential rest, we still need humans. If we now also consider the finding that using AI can relatively quickly lead to a loss of problem-solving skills 24 or that these skills are not acquired at all, then the overall benefit of using GenAI in professional software development is more than questionable. As long as programming - and every technicality that comes with it - will not be fully replaced by some kind of AI, we will still need expert developers who can programm, maintain and debug code to the finest level of detail. Where, we wonder, will those senior developers come from when companies replace their junior staff with coding agents? Actual vs. promised benefitsPermalink If you read testimonials about the use of GenAI that people perceive as successful, you will mostly encounter scenarios in which ‘AI’ helps to make tasks that are perceived as boring, unnecessarily time-consuming or actually pointless faster or more pleasant. So it’s mainly about personal convenience and perceived efficiency. Entertainment also plays a major role: the poem for Grandma’s birthday, the funny song for the company anniversary or the humorous image for the presentation are quickly and supposedly inexpensively generated by ‘AI’. However, the promises made by the dominant GenAI companies are quite different: solving the climate crisis, providing the best medical advice for everyone, revolutionising science, ‘democratising’ education and much more. GPT5, for example, is touted by Sam Altman, CEO of OpenAI, as follows: ‘With GPT-5, it’s now like talking to an expert — a legitimate PhD-level expert in any area you need […] they can help you with whatever your goals are.’ 25 However, to date, there is still no actual use case that provides a real qualitative benefit for humanity or at least larger groups. The question ‘What significant problem (for us as a society) does GenAI solve?’ remains unanswered. On the contrary: While machine learning and deep learning methods certainly have useful applications, the most profitable area of application for ‘AI’ at present is the discovery and development of new oil and gas fields 26. Harmful aspects of GenAIPermalink But regardless of how one assesses the benefits of this technology, we must also consider the downsides, because only then can we ultimately make an informed and fair assessment. In fact, the range of negative effects of hyperscaled generative AI that can already be observed is vast. Added to this are numerous risks that have the potential to cause great social harm. Let’s take a look at what we consider to be the biggest threats: GenAI is an ecological disasterPermalink PowerPermalink The data centres required for training and operating large generative models 27 far exceed today’s dimensions in terms of both number and size. The projected data centre energy demand in the USA is predicted to grow from 4.4% of total electricity in 2023 to 22% in 2028 28. In addition, the typical data centre electricity mix is more CO2-intensive than the average mix. There is an estimated raise of ~11 percent for coal generated electricity in the US, as well as tripled emissions of greenhouse gases worldwide by 2030 - compared to the scenario without GenAI technology 29. Just recently Sam Altman from OpenAI blogged some numbers about the energy and water usage of ChatGPT for “the average query” 30. On the one hand, an average is rather meaningless when a distribution is heavily unsymmetric; the numbers for queries with large contexts or “chain of reasoning” computations would be orders of magnitude higher. Thus, the potential efficiency gains from more economical language models are more than offset by the proliferation of use, e.g. through CoT approaches and ‘agent systems’. On the other hand, big tech’s disclosure of energy consumption (e.g. by Google 31) is intentionally selective. Ketan Joshi goes into quite some details why experts think that the AI industry is hiding the full picture 32. Since building new power plants - even coal or gas fuelled ones - takes a lot of time, data center companies are even reviving old jet engines for powering their new hyper-scalers 33. You have to be aware that those engines are not only much more noisy than other power plants but also pump out nitrous oxide, one of the main chemicals responsible for acid rain 34. WaterPermalink Another problem is the immensely high water consumption of these data centres 35. After all, cooling requires clean water in drinking quality in order to not contaminate or clog the cooling pipes and pumps. Already today, new data centre locations are competing with human consumption of drinking water. According to Bloomberg News about two-thirds of data-centers that were built or developed in 2022 are located in areas that are already under “water-stress” 36. In the US alone “AI servers […] could generate an annual water footprint ranging from 731 to 1,125 million m3” 37. It’s not only an American problem, though. In other areas of the world the water-thirsty data centers also compete with the drinking water supply for humans 38. Electronic WastePermalink Another ecological problem is being noticeably exacerbated by ‘AI’: the amount of electronic waste (e-waste) that we ship mainly to “Third World” countries and which is responsible for soil contamination there. Efficient training and querying of very large neural networks requires very large quantities of specialised chips (GPUs). These chips often have to be replaced and disposed of within two years. The typical data center might not last longer than 3 to 5 years before it has to be rebuilt in large parts39. In summary, it can be said that GenAI is at least an accelerator of the ecological catastrophe that threatens the earth. And it is the argument for Google, Amazon and Microsoft to completely abolish their zero CO2 targets 40 and replace them with investments of several hundred billion dollars for new data centers. GenAI threatens education and sciencePermalink People often try to use GenAI in areas where they feel overloaded and overwhelmed: training, studying, nursing, psychotherapeutic care, etc. The fields of application for ‘AI’ are therefore a good indication of socially neglected and underfunded areas. The fact that LLMs are very good at conveying the impression of genuine knowledge and competence makes their use particularly attractive in these areas. A teacher under the simultaneous pressure of lesson preparation, corrections and covering for sick colleagues turns to ChatGPT to quickly create an exercise sheet. A student under pressure to get good grades has their English essay corrected by ‘AI’. The researcher under pressure to publish will ‘save’ research time by reading the AI-generated summary of relevant papers – even if they are completely wrong in terms of content 41. Tech companies like OpenAI and Microsoft play on that situation by offering their ‘AI’ for free or for little money to students and universities. The goal is obvious: Students that get hooked on outsourcing some of their “tedious” task to a service will continue to use - and eventually buy - this service after graduation. What falls by the wayside are problem-solving skills, engagement with complex sources, and the generation of knowledge through understanding and supplementing existing knowledge. Some even argue that AI is destroying critical education and learning itself 42: Students aren’t just learning less; their brains are learning not to learn. The training cycle of schools and universities is fast. Teachers are already reporting that pupils and students have acquired noticeably less competence in recent years, but have instead become dependent on unreliable ‘tools’ 43. The real problem with using GenAI to do assignments is not cheating, but students “are not just undermining their ability to learn, but to someday lead.” 44 GenAI is destroying the free internet.Permalink The fight against bots on the internet is almost as old as the internet itself – and has been quite successful so far. Multifactor authentication, reCaptcha, honeypots and browser fingerprinting are just a few of the tools that help protect against automated abuse. However, GenAI takes this problem to a new level – in two ways. To make ‘the internet’ usable as the main source for training LLMs, AI companies use so-called ‘crawlers’. These essentially behave like DDoS attackers: They send tens of thousands of requests at once, from several hundred IPs in a very short time. Robot.txt files are ignored; instead, the source IP and user agent are obscured 45. These practices have massive disadvantages for providers of genuine content: Costs for additional bandwidth. Lost advertising revenue, as search engines now offer LLM-generated summaries instead of links to the sources. This threatens the existence of remaining independent journalism in particular 46. Misuse of own content for AI-supported competition. If the place where knowledge is generated is separated from the place where it is consumed, and if this makes the performance of generation even more opaque than before, the motivation to continue generating knowledge also declines. For projects such as Wikipedia, this means fewer donors and fewer contributors. Open communities often have no other option but to shut themselves off. Another aspect is the flooding of the internet with generated content that cannot be automatically distinguished from non-generated content. This content overwhelms the maintainers of open source software or portals such as Wikipedia 47. If this content is then also entered by humans – often in the belief that they are doing good – it is no longer possible to take action against the methodology. In the long run, this means that less and less authentic training material will lead to increasingly poor results from the models. Last but not least, autonomously acting agents make the already dire state of internet security much worse 48. Think of handing all your personal data and credentials to a robot that is distributing and using that data across the web, wherever and whenever it deems it necessary for reaching some goal. is controlled by LLMs who are vulnerable to all kinds of prompt injection attacs 49. is controlled by and reporting to companies that do not have your best interest in mind. has no awareness and knowledge about the implication of its actions. is acting on your behalf and thereby making you accountable. GenAI is a danger to democracyPermalink The manipulation of public opinion through social media precedes the arrival of LLMs. However, this technology gives the manipulators much more leverage. By flooding the web with fake news, fake videos and fake everything undemocratic (or just criminal) parties make it harder and harder for any serious media and journalism to get the attention of the public. People no longer have a common factual basis, which is necessary for all social negotiations. If you don’t agree on at least some basic facts, arguing about policies and measures to take is pointless. Without negotiations democracy will be dying; in many parts of the world it already is. GenAI versus human creativityPermalink Art and creativity are also threatened by generative AI. The impact on artists’ incomes of logos, images and illustrations now being easily and quickly created by AI prompts is obvious. A similar effect can also be observed in other areas. Studies show that poems written by LLMs are indistinguishable from those written by humans and that generative AI products are often rated more highly 50. This can be explained by a trend towards the middle and the average, which can also be observed in the music and film scenes film scene: due to its basic function, GenAI cannot create anything fundamentally new, but replicates familiar patterns, which is precisely why it is so well received by the public. Ironically, ‘AI’ draws its ‘creativity’ from the content of those it seeks to replace. Much of this content was used as training material against the will of the rights holders. Whether this constitutes a copyright infringement has not yet been decided; morally, the situation seems clear. The creative community is the first to be seriously threatened by GenAI in its livelihood 51. It’s not a coincidence that a big part of GenAI efforts is targeted at “democratizing art”. This framing is completely upside down. Art has been one of the most democratic activities for a very long time. Everybody can do it; but not everybody wants to do put in the effort, the practicing time and the soul. Real art is not about the product but about the process, which requires real humans. Generating art without the friction is about getting rid of the humans in the loop - and still making money. Digital colonialismPermalink The huge amount of data required by hyperscaled AI approaches makes it impossible to completely curate the learning content. And yet, one would like to avoid the reproduction of racist, inhuman and criminal content. Attempts are being made to get the problem under control by subsequently adapting the models to human preferences and local laws through additional ‘reinforcement learning from human feedback (RLHF)’ 52. The cheap labour for this very costly process can be found in the Global South. There, people are exposed to hours of hate speech, child abuse, domestic violence and other horrific scenarios in their poorly paid jobs in order to filter them out of the training material of large AI companies 53. Many emerge from these activities traumatised. However, it is not only people who are exploited in the less developed regions of the world, but also nature: the poisoning of the soil with chemicals during the extraction of raw materials for digital chips, as well as the contamination caused by our electronic waste and its improper disposal, are collateral damage that we willingly accept and whose long-term consequences are currently extremely difficult to assess. Here, too, the “developed” world profits, whereas the negative aspects are outsourced to the former colonies and other poor regions of the world. Political aspectsPermalink As software developers, we would like to ‘leave politics out of it’ and instead focus entirely on the cool tech. However, this is impossible when the advocates of this technology pursue strong political and ideological goals. In the case of GenAI, we can cleary see that the US corporations behind it (OpenAI, Google, Meta, Microsoft, etc.) have no problem with the current authoritarian – some say fascist – US government 54. In concrete terms, this means, among other things, that the models are explicitly manipulated to be less liberal or simply not to generate any output that could upset the CEO or the president 55. Even more serious is the fact that many of the leading minds behind these corporations and their financiers adhere to beliefs that can be broadly described as digital fascism. These include Peter Thiel, Marc Andreessen, Alex Karp, JD Vance, Elon Musk and many others on “The Authoritarian Stack” 56. Their ideologies, disguised as rational theories, are called longtermism and effective altruism. What they have in common is that they consider democracy and the state to be obsolete models, compassion to be ‘woke’, and that the current problems of humanity are insignificant, as our future lies in the colonisation of space and the merging of humans with artificial superintelligence 57. Do we want to give people who adhere to these ideologies (even) more power, money and influence by using and paying for their products? Do we want to feed their computer systems with our data? Do we really want to expose ourselves and our children to the answers from chatbots which they have manipulated? Not quite as abstruse, but similarly misanthropic, is the imminent displacement of many jobs by AI, as postulated by the same corporations in order to put pressure on employees with this claim. Demanding a large salary? Insisting on your legal rights? Complaining about too much workload? Doubts about the company’s goals? Then we’ll just replace you with cheap and uncomplaining AI! Whichever way you look at it, AI and GenAI are already being used politically. If we go along without resistance, we are endorsing this approach and supporting it with our time, our attention and our money. ConclusionPermalink Ideally, we would like to quantify our assessment by adding up the advantages, adding up the disadvantages and finally checking whether the balance is positive or negative. Unfortunately, in our specific case, neither the benefits nor the harm are easily quantifiable; we must therefore consult our social and personal values. Discussions about GenAI usually revolve purely around its benefits. Often, the capabilities of all ‘AI’ technologies (e.g. protein folding with AlphaFold 2) are lumped together, even though they have little in common with hyperscaling GenAI. However, if we consider the consequences and do not ignore the problems this technology entails – i.e. if we consider both sides in terms of ethics – the assessment changes. Convenience, speed and entertainment are then weighed against numerous damages and risks to the environment, the state and humanity. In this sense, the ethical use and further expansion of GenAI in its current form is not possible. Can there be ethical GenAI?Permalink If the use of GenAI is not ethical today what would have to change, which negative effects of GenAI would have to disappear or at least be greatly reduced in order to tip the balance between benefits and harms in the other direction? The models would have to be trained exclusively with publicly known content whose original creators consent to its use in training AI models. The environmental damage would have to be reduced to such an extent that it does not further fuel the climate crisis. Society would have to get full access to the training and operation of the models in order to rule out manipulation by third parties and restrict their use to beneficial purposes. This would require democratic processes, good regulation and oversight through judges and courts. The misuse and harming of others, e.g., through copyright theft or digital colonialism, would have to be prevented. Is such a change conceivable? Perhaps. Is it likely, given the interest groups and political aspects involved? Probably not

I'm not advocating that everyone should self-host everything. But the pendulum has swung too far toward managed services. There's a large sweet spot where self-hosting makes perfect sense, and more teams should seriously consider it. Start small. If you're paying more than $200/month for RDS, spin up a test server and migrate a non-critical database. You might be surprised by how straightforward it is. The future of infrastructure is almost certainly more hybrid than it's been recently: managed services where they add genuine value, self-hosted where they're just expensive abstractions. Postgres often falls into the latter category. Footnotes They're either just hosting a vanilla postgres instance that's tied to the deployed hardware config, or doing something opaque with edge deploys and sharding. In the latter case they near guarantee your DB will stay highly available but costs can quickly spiral out of control. ↩ Maybe up to billions at this point. ↩ Even on otherwise absolutely snail speed hardware. ↩ This was Jeff Bezos's favorite phrase during the early AWS days, and it stuck. ↩ Similar options include OVH, Hetzner dedicated instances, or even bare metal from providers like Equinix. ↩ AWS RDS & S3 has had several major outages over the years. The most memorable was the 2017 US-East-1 outage that took down half the internet. ↩

Write-Ahead Logging is critical for durability and performance:

Storage Tuning: NVMe SSDs make having content on disk less harmful than conventional spinning hard drives, so you'll want to pay attention to the disk type that you're hosted on:

Making fresh connections in postgres has pretty expensive overhead, so you almost always want to put a load balancer on front of it. I'm using pgbouncer on all my projects by default - even when load might not call for it. Python asyncio applications just work better with a centralized connection pooler.

Memory Configuration: This is where most people mess up. Pulling the standard postgres docker image won't cut it. You have to configure memory bounds with static limits that correspond to hardware. I've automated some of these configurations. But whether you do it manually or use some auto-config, tweaking these params is a must.

When self-hosting doesn't make sense I'd argue self-hosting is the right choice for basically everyone, with the few exceptions at both ends of the extreme: If you're just starting out in software & want to get something working quickly with vibe coding, it's easier to treat Postgres as just another remote API that you can call from your single deployed app If you're a really big company and are reaching the scale where you need trained database engineers to just work on your stack, you might get economies of scale by just outsourcing that work to a cloud company that has guaranteed talent in that area. The second full freight salaries come into play, outsourcing looks a bit cheaper. Regulated workloads (PCI-DSS, FedRAMP, HIPAA, etc.) sometimes require a managed platform with signed BAAs or explicit compliance attestations.

The real operational complexity

I helped prove this to myself when I migrated off RDS. I took a pg_dump of my RDS instance, restored it to a self-hosted server with identical specs, and ran my application's test suite. Performance was identical. In some cases, it was actually better because I could tune parameters that RDS locks down.

But the actual database engine? It's the same Postgres running the same SQL queries with the same performance characteristics.

The value proposition is operational: they handle the monitoring, alerting, backup verification, and incident response. It's also a production ready configuration at minute zero of your first deployment.

For the most part managed database services aren't running some magical proprietary technology. They're just running the same open-source Postgres you can download with some operational tooling wrapped around it. Take AWS RDS. Under the hood, it's: Standard Postgres compiled with some AWS-specific monitoring hooks A custom backup system using EBS snapshots Automated configuration management via Chef/Puppet/Ansible Load balancers and connection pooling (PgBouncer) Monitoring integration with CloudWatch Automated failover scripting

Fast forward to 2025 and I hope the pendulum might be swinging back. RDS pricing has grown considerably more aggressive. A db.r6g.xlarge instance (4 vCPUs, 32GB RAM) now costs $328/month before you add storage, backups, or multi-AZ deployment. For that price, you could rent a dedicated server with 32 cores and 256GB of RAM.

The real shift happened around 2015 when cloud adoption accelerated. Companies started to view any infrastructure management as "undifferentiated heavy lifting"4. Running your own database became associated with legacy thinking. A new orthodoxy emerged of focusing on your application logic and letting AWS handle the infrastructure.

* "Digital platforms are used for hybrid campaigns."* "EU can't compete with US tech ON THEIR TERMS."* "Post-reality US is what happens when tech is unregulated."* "Ireland is a Trojan Horse for Big Tech."* "The Digital Omnibus is sabotage."

Web2.0 seems to have been a mistake and is being rolled back entirely. The whole concept of having your own web property that other people can write to (leave comments and other things) has gone away. That means also there’s no need for a dynamic website with database anymore.

Reading back all those old posts and weeknotes I have here is super nice and reminds me:Keeping a record of things is really valuable. Just write and trust that it will come in handy at some point.I used to do so many things in a given week. Compared to what I’m doing now, my life was insanely eventful.I was consistently (too) early on a lot of things. For instance: I read myself complaining about restaurants and food in Amsterdam, something which is mostly solved now.

PHP/MySQL is losing adoption

Our backup nodes are located in Switzerland, and we use AWS (Amazon Web Services).

Media files are not directly downloaded in overall syncing to save bandwidth. Instead, when that file is requested, it is streamed to your device from the backup node or your devices on the network. For example, if you have a 4K Video, it will be streamed from the backup node or P2P devices to your device. So when you open an object with an image, it downloads. When you press play on video & audio, it begins to download. After that, this file will be stored in the application cache.

Pricing for Self-HostersSelf-hosters can manage the limits of Viewers/Editors they invite to their Spaces themselves. Those who would like to purchase a name in the Anytype naming system or access priority support, can purchase a membership at the same price as other beta testers.

Frontmatter (the --- metadata at the top) → becomes object properties.

Step 1: Export from ObsidianClean up your notesConvert embeds like ![[Note Title]] into simple links [[Note Title]]Remove or convert plugin-specific syntax, such as Dataview queries, Templater fields, etc.

Pick the plan that works for you now, add more teammates and additional storage anytime you need. Switch or cancel anytime.

Furthermore, the choice of incorporating as an Association rather than an LLC or other legal structure distinguishes ANY as a company managed in accordance with commercial practices, but which exists for the greater good. We consider ourselves a remote team, but we maintain one hub in Berlin (where our team is employed by our local GmbH, a subsidiary of the Swiss Association).

Our applications, on the other hand, are distributed under a source available license.

supported by

Single objects, infinite possibilities Visualise connections using graph & database views

Templates

Databases

Nobody is mediating the connection between your devices

Nobody can see what’s in your vault, except for you

discuss, organize, remember

A: The Calibre dedrm plugin only removes the Readium 1.0 drm. As for Readium 2.5, it employs the tougher drm, it cannot be handled with Calibre drdrm plugin.

The DeDRM plugin i

you can download the .lcpl within the Calibre directly.

All you need to open (or decrypt) LCP eBooks is the account passphrase given to you by the eBook provider - the very same passphrase you'd have to enter into your eBook reader device (once) to read LCP-encrypted books. Sometime, the passphrase can be the passwords.

drei Erkenntnisse

Zusätzlich habe ich mir jeden Freitag eine halbe Stunde Zeit genommen für einen wöchentlichen Rückblick. Dabei habe ich die Einträge durchgeblättert, Muster markiert und die wichtigsten Episoden noch einmal verdichtet. Dieser Rückblick half mir besonders, nicht nur auf einzelne Stunden zu reagieren, sondern auch langfristige Entwicklungen zu erkennen, z. B. wiederkehrende Probleme bei Gruppenarbeiten oder Fortschritte in der Diskussionskultur.

Der Zeitaufwand war überschaubar. In der Regel war ich in rund 10 Minuten fertig. Entscheidend war die Routine: Das Journal hatte in meinem Ablauf denselben Stellenwert wie das Aufräumen des Kursraums.

Ich habe mein Journal in einem einfachen A5-Notizbuch geführt. Jede Doppelseite gehörte einer Unterrichtseinheit. Links: die Events in Stichpunkten. Rechts: Episode und Analyse. Der Ablauf sah so aus: Direkt nach der Lektion: fünf Minuten für die Events – die wichtigsten Beobachtungen notieren. Episode auswählen: eine kleine Szene beschreiben, sodass ich sie auch in drei Monaten noch klar vor Augen habe. Analysis: Reflexion über Ursachen und Konsequenzen, verbunden mit einem konkreten nächsten Schritt.

Einfache Struktur: Events – Episode – Analysis Beim Start meiner Challenge habe ich bewusst eine schlanke Form gewählt. In Richards & Farrell bin ich auf ein Modell gestossen, das mir sofort einleuchtete: Events – Episode – Analysis. Events: kurze, stichwortartige Notizen zu den wichtigsten Geschehnissen der Lektion. Episode: eine kleine Szene, die ich erzählerisch ausformuliere: vielleicht eine gute Diskussion, vielleicht ein technisches Problem. Analysis: meine Gedanken dazu: Warum war diese Episode wichtig? Was hat dazu geführt? Und was nehme ich mir für die nächste Einheit vor?

ein Arbeitsinstrument. Es schafft einen Raum, um nach dem Unterricht innezuhalten, zentrale Ereignisse zu notieren, kleine Episoden festzuhalten und darüber nachzudenken, was gelungen ist – und was nicht.

Donald Schön hat den Begriff „reflection-on-action“ geprägt: Wir lernen, indem wir nach der Handlung darüber nachdenken, warum wir etwas getan haben, und welche Alternativen möglich gewesen wären [2].

Pädagogische Forschung und Praxis empfehlen es seit Jahren als Werkzeug, um den eigenen Unterricht zu reflektieren und weiterzuentwickeln. Jack Richards und Thomas Farrell definieren es schlicht als „a teacher’s written record of classroom events and of the teacher’s own thoughts and reflections about teaching“ [1].

30 Tage lang wollte ich nach jedem Unterricht ein handschriftliches Lehrjournal führen. Kein grosses Projekt, einfach ein Notizbuch und der feste Vorsatz, nach der Lektion zehn bis fünfzehn Minuten zu schreiben. Am Anfang war ich skeptisch, ob sich dieser zusätzliche Aufwand lohnen würde. Heute weiss ich: Es war eine der produktivsten Routinen, die ich mir als Dozent angewöhnt habe.

the Language Processing Unit (LPU), a new category of processor. Groq created and built the LPU from the ground up to meet the unique needs of AI. LPUs run Large Language Models (LLMs) and other leading models at substantially faster speeds and, on an architectural level, up to 10x more efficiently from an energy perspective compared to GPUs.

The company's stock has declined approximately 34% from its December 2024 peak,

the raw intelligence that LLMs provide

Benioff had recently told Business Insider that he's drafting the company's annual strategic document with data foundations—not AI models—as the top priority, explicitly citing concerns about "hallucinations" without proper data context.

Phil Mui described as AI "drift" in an October blog post. When users ask irrelevant questions, AI agents lose focus on their primary objectives. For instance, a chatbot designed to guide form completion may become distracted when customers ask unrelated questions.

Home security company Vivint, which uses Agentforce to handle customer support for 2.5 million customers, experienced these reliability problems firsthand. Despite providing clear instructions to send satisfaction surveys after each customer interaction, The Information reported that Agentforce sometimes failed to send surveys for unexplained reasons. Vivint worked with Salesforce to implement "deterministic triggers" to ensure consistent survey delivery.

Chief Technology Officer of Agentforce, pointed out that when given more than eight instructions, the models begin omitting directives—a serious flaw for precision-dependent business tasks.

The company is now emphasizing that Agentforce can help "eliminate the inherent randomness of large models," marking a significant departure from the AI-first messaging that dominated the industry just months ago.

admission comes after Salesforce reportedly reduced its support staff from 9,000 to 5,000 employees

All of us were more confident about large language models a year ago," Parulekar stated, revealing the company's strategic shift away from generative AI toward more predictable "deterministic" automation in its flagship product, Agentforce.

would take seriously the fact that intelligence is now being scaled and distributed through organizations long before it is unified or fully understood

examine power as an emergent consequence of deployment and incentives, not intent.

Such a work would treat alignment as institutional design rather than a property of models alone.

Empirical grounding. In 2015, scaling laws, emergent capabilities, and deployment‑driven feedback loops were speculative. Today, they are measurable. That shift changes the nature of responsibility, governance, and urgency in ways that were difficult to justify rigorously at the time.

Political economy and power. The book largely brackets capital concentration, platform dynamics, and geopolitical competition. Today, these are central to any serious discussion of AI, not because the technology changed direction, but because it scaled fast enough to collide with real institutions and entrenched interests.

Alignment as an operational problem. The book assumes that sufficiently advanced intelligences would recognize the value of cooperation, pluralism, and shared goals. A decade of observing misaligned incentives in human institutions amplified by algorithmic systems makes it clear that this assumption requires far more rigorous treatment. Alignment is not a philosophical preference. It is an engineering, economic, and institutional problem.

somethingnew.davidorban.com

what it feels like to live through an intelligence transition that does not arrive as a single rupture, but as a rolling transformation, unevenly distributed across institutions, regions, and social strata.

As a result, the debate shifted. The central question is no longer “Can we build this?” but “What does this do to power, incentives, legitimacy, and trust?”

The book’s central argument was not about timelines or machines outperforming humans at specific tasks. It was about scale. Artificial intelligence, I argued, should not be understood at the level of an individual mind, but at the level of civilization. Technology does not merely support humanity. It shapes what humanity is. If AI crossed certain thresholds, it would not just automate tasks, but it would reconfigure social coordination, knowledge production, and agency itself. That framing has aged better than I expected, not because any particular prediction came true, but because the underlying question turned out to be the right one.

America’s former role is gone. And I think that Americans themselves are having the hardest time of all coming to terms with what that might actually mean in the long run.

The officer then said that even a swift return of America to its former role won’t matter. Because “we will never fucking trust you again.”The Americans at the table seemed somewhat startled by the heat of that pronouncement. I agreed with it entirely. So, it seemed to me, did most of the non-Americans. This wasn’t the only such moment at the forum this year, but it was, to me, the most interesting. And it was still being talked about the next day. “Thank God,” one allied official said to me. “Someone had to tell them.”

But before I could worry about it too much, a senior military officer from a major (non-American) allied nation drove a stake right through the heart of the matter.America has blown 80 years of accumulated goodwill and trust among its allies, our American moderator was told. A rock-steady assumption of allied defence and security planning for literally generations has been that America would act in its own interests, sure, but that those interests would be rational, and would still generally value the institutions that America itself worked so hard to build after the Second World War. America’s recent actions have destroyed the ability of any ally to continue to have faith in America to act even within its own strategic self-interest, let alone that of any ally.

And the damage to America’s soft power — the shutting down of aid programs and things like Voice of America — can’t be undone rapidly no matter who wins the midterms. U.S. troops that are pulled out of bases where the U.S. no longer sees a strategic reason for their presence aren’t likely to come back.And, this is the critical part, wouldn’t necessarily be welcomed even if they did.

The damage is already done. I know firsthand that a great many Americans who really do believe in the post-1945 global order, and of America’s prior role in the world and the value of that role to America and Americans, are still inside the U.S. government. But I also know that many of them are retiring, or seeking early retirement, or switching to consulting gigs. They can’t stomach what U.S. foreign policy is becoming, and they won’t be a part of it.Good for them. But every single person who departs is being replaced by someone who is totally fine with the new U.S. foreign policy. And sometimes is actually quite enthusiastic about it. That will accelerate the process that’s already underway. And those new people are going to have long careers, shaping things both in public and behind the scenes.

The session, over dinner, was a small group. It was about America’s moral leadership in the world. Our moderator was a now-former American official. She was pretty frank and clear-eyed about how America’s allies currently view the country’s place in the world, but also expressed some hope that after the midterms next year or maybe the next presidential election, things would start to get more back to normal. We were assured that a lot of people in America are still with us. Some of the other Americans present nodded their heads.

Amazon or any other U.S.-based company will then make the decision that best promotes and protects long-term shareholder value. And that decision will be, in every case, to submit and comply. Everyone in the room knew that. America is different, now. It’s inescapable.

There were two fascinating things about that exchange (it starts around the 17-minute mark of that video). The first was the question itself; it alone was a signal of how much things have changed. The second interesting thing is that Zapolsky’s answer was, with respect, bullshit. I can see why he’s a legal officer! He gave an answer that was legally correct — the only way that the U.S. government can officially bar Amazon from providing cloud services for a foreign military, for example, would be by sanctions or some comparable legislation.

What would happen is that someone senior at Amazon, maybe Jeff Bezos himself, would get a call from some golf partner or drinking buddy in the administration, and the message will be simple: “Stop, or you won’t get contracts. We’ll arrange some hearings into your operations. Your little spaceflight company will find itself under way more levels of regulatory review than your Musk-owned competitor. This is what the boss wants. Make it happen.”

Tarabay dropped a humdinger of a question on Zapolsky. Here’s the quote (slightly cleaned up for clarity): “We’re in an age where there’s a government that puts pressure on companies [and] people for [Trump’s] own gain. You have been so steadfast in your support for Ukraine. What will Amazon do if your government says ‘Stop’?”Zapolsky replied that the company has contracts with foreign governments and NATO allies and said that Amazon would only change those relationships if it was legally forced to do so via something like a sanction.

but it raised a much deeper point — America has “walked away” from its allies. And the leader of the CODEL took no issue with that characterization.

Kucher said this to Shaheen: “We’ve talked about allyship. What should the allies, who uphold democratic values, in the reality that the United States has walked away from them … what should the allies do?”

I will not confirm or deny that that is happening, but there is nothing in 12333 to prevent that from happening.

Hypothetically, under 12333 the NSA could target a single foreigner abroad. And hypothetically if, while targeting that single person, they happened to collect every single Gmail and every single Facebook message on the company servers not just from the one person who is the target, but from everyone—then the NSA could keep and use the data from those three billion other people. That’s called 'incidental collection.'

Executive Order 12333 has been regarded by the American intelligence community as a fundamental document authorizing the expansion of data collection activities.[9] The document has been employed by the National Security Agency as legal authorization for its collection of unencrypted information flowing through the data centers of internet communications giants Google and Yahoo!.[9]

Part 2.3 permits collection, retention and dissemination of the following types of information along with several others. .mw-parser-output .templatequote{overflow:hidden;margin:1em 0;padding:0 32px}.mw-parser-output .templatequotecite{line-height:1.5em;text-align:left;margin-top:0}@media(min-width:500px){.mw-parser-output .templatequotecite{padding-left:1.6em}}(c) Information obtained in the course of lawful foreign intelligence, counterintelligence, international narcotics or international terrorism investigation ... (i) Incidentally obtained information that may indicate involvement in activities that may violate federal, state, local or foreign laws[1]

Amerikaanse wetgeving zoals de CLOUD Act (Clarifying Lawful Overseas Use of Data Act), de Foreign Intelligence Surveillance Act (FISA)5 en Executive Order 12333 komt te vallen?

Indien de Algemene Rijksvoorwaarden bij IT-overeenkomsten 2022 (ARBIT-2022)6 van toepassing verklaard is op de overeenkomst is er een ontbindingsgrond als er sprake is van een ingrijpende wijziging in de zeggenschap (wat het geval kan zijn bij fusies en overnames) met betrekking tot de onderneming van de wederpartij/opdrachtnemer. Door de landsadvocaat wordt momenteel onderzocht in hoeverre dit het geval is. Indien door de overname de nakoming van de verwerkersovereenkomst en de naleving van de AVG wordt bemoeilijkt of zelfs onmogelijk wordt, kan dit een grond vormen om de dienstverleningsovereenkomst respectievelijk de verwerkersovereenkomst te ontbinden. Dit laat overigens onverlet de mogelijkheid om in een concreet geval een overeenkomst op basis van de Algemene Rijksvoorwaarden op te kunnen zeggen.

Het versterken van digitale autonomie vergt daarom een Europese aanpak.

De overeenkomsten tussen de Staat en Solvinity bieden aanknopingspunten om ten minste van Solvinity te verlangen dat er technische en organisatorische maatregelen worden getroffen om te waarborgen dat de gegevens waartoe zij toegang heeft op een wijze worden verwerkt die voldoet aan de in de EU geldende regels, zoals die uit de Algemene verordening gegevensbescherming. Welke maatregelen dat zullen zijn vormt onderwerp van de gesprekken tussen de Staat en Solvinity.

De drie genoemde wettelijke instrumenten maken het, in ieder geval in theorie, mogelijk dat autoriteiten in de VS onder de in deze wetgeving genoemde voorwaarden toegang kunnen krijgen tot de gegevens waarover een onderneming in de VS beschikt, óók wanneer de gegevens zich bevinden onder een dochtervennootschap en op servers buiten de VS. Als Solvinity wordt overgenomen door een onderneming in de VS brengt dit Solvinity onder de reikwijdte van deze wetgeving. Het gevolg daarvan kan, in ieder geval in theorie, zijn dat autoriteiten in de VS in voorkomend geval toegang krijgen tot de gegevens die door Solvinity in opdracht van de Staat worden verwerkt.

An aspect of the human use of information that has generally been overlooked in the automation of information services is the human tendency to locate information spatially. Computer-based systems do not necessarily assign any unique role to spatial tags, and so a feature of considerable importance for the organization of the user's memory seems to have been largely overlooked. The spatial dimension of human memory is discussed, and some suggestions are offered for exploiting it more effectively in the context of information retrieval services.

The economics are far from certain though, and competition will be fierce. Even if NASA is able to spur a private orbital economy, there may not be enough business to support multiple private space stations.

All these projects hope to have NASA as an anchor tenant. But they are also heavily reliant on the idea that there are a broad range of potential customers also willing to pay for orbital office space.

In addition, Blue Origin, founded by Jeff Bezos, is working with Sierra Space and Boeing to build Orbital Reef,

Meanwhile, Voyager Space and Airbus are designing a space station called Starlab, which recently moved into “full-scale development” ahead of an expected 2028 launch. The station can host four astronauts, features an external robotic arm, and is designed to launch in one go aboard SpaceX’s forthcoming Starship rocket.

Axiom Space, one of the companies vying for this funding, plans to piggyback on the ISS to build its space station. The company will first launch a power and heating module and connect it to the ISS. The module will be able to operate independently starting in 2028. They’ll then gradually add habitat and research modules alongside airlocks to create a full-fledged private space station.

The agency has paid out about $415 million in the program's first phase to help companies flesh out their designs. But next year, NASA plans to select one or more companies for Phase 2 contracts worth between $1 billion and $1.5 billion and set to run from 2026 to 2031.

Development of Vast’s second station relies on funding from NASA’s Commercial Low Earth Orbit Destinations program,

Haven-2, a larger modular station that Vast hopes could succeed the ISS.

May 2026, when California-based startup Vast plans to launch its Haven-1 space station

ISS is nearing the end of its planned lifespan and NASA’s been clear that it doesn’t intend to replace the space station.

The ISS was humanity's only permanent outpost in space for nearly a quarter of a century, until China’s Tiangong station was permanently crewed in 2022.

Main entry point[edit] Main article: Entry point As in C, C++, C#, Java, and Go, the entry point to a Kotlin program is a function named "main", which may be passed an array containing any command-line arguments. This is optional since Kotlin 1.3.[26] Perl, PHP, and Unix shell–style string interpolation is supported. Type inference is also supported.

The name is derived from Kotlin Island, a Russian island in the Gulf of Finland, near Saint Petersburg. Andrey Breslav, Kotlin's former lead designer, mentioned that the team decided to name it after an island, in imitation of the Java programming language which shares a name with the Indonesian island of Java

On 7 May 2019, Google announced that the Kotlin programming language had become its preferred language for Android app developers.[7] Since the release of Android Studio 3.0 in October 2017, Kotlin has been included as an alternative to the standard Java compiler.

Prerequisites Basic Kotlin knowledge

Once you have multiple enhancement prompts, you can switch between them on the fly without opening the main app window. This is done using keyboard shortcuts when the Mini Recorder is active. For a detailed guide, see Quickly Switching Enhancement Prompts. By configuring an AI provider, you unlock the full potential of VoiceInk's enhancement features, allowing you to transform your speech into perfectly formatted and context-aware text.

In the Enhancement settings, you can also enable Clipboard Context and Context Awareness. These features provide the AI with additional information from your clipboard or screen to produce more accurate and relevant results.

Select a Model: Once connected, you can choose from any of the models you have pulled in Ollama.

Supported Providers Ollama (Free & Local): Run powerful open-source models locally on your machine. This is a great option for privacy and offline use.

Enhancement ModelsTransform your transcriptions with AI-powered enhancement and correction

Here are the best transcription models you can use with VoiceInkAI transcription modelDescriptionParakeet modelthe real-time offline model with the best accuracyWhisper large v3 turbofast and accurate Whisper model from OpenAI

Made with ❤️ by Pax

Open Source & Privacy-FirstThe source code of VoiceInk is available on GitHub

VoiceInk works only on Apple Silicon Macs and requires macOS 14.0 or later. The local models require Apple Silicon's Neural Engine for fast, local AI processing. While we offer cloud models as an alternative, the local version is designed specifically for Apple Silicon. We recommend having at least 8GB of RAM for optimal performance.

The Cloud Enhancement feature is entirely optional - if enabled, only the transcribed text (not your voice) is processed by third-party providers for improved accuracy.

SoloWorks on 1 macOS device

Personal DictionaryTrain AI with your custom words, phrases, and replacements for faster, more personalized responses

VoiceInk is a macOS dictation app that uses local AI models to convert speech to text with complete privacy since all processing happens offline on your device.

Offline Mode: Works without internet connection

MacWhisper is a Mac-exclusive transcription app that runs OpenAI's Whisper AI model directly on your computer for speech-to-text conversion.

The local processing makes SuperWhisper stand out for privacy-conscious users who want dictation without cloud dependency

Apple Dictation is the built-in speech-to-text feature that comes with all Mac and iOS devices.

Flow uses a combination of open-source models (i.e. LLAMA 3.1) and proprietary LLM providers (such as OpenAI) to provide its services. Wispr has agreements with all third party generative AI providers to ensure no data is stored or used for model training (zero data retention).

Privacy built-inWith Privacy Mode enabled, zero dictation data is stored on our servers. To enable it, go to Settings → Data & Privacy → Privacy Mode.

Backed by the bestWe are fortunate to work with some of the best investors in the industry. Our backers include the top venture firms and some of the world’s most exceptional founders and product builders.

We are entering a time when the ability to create software is no longer a specialized skill. It is becoming a basic form of digital literacy, like writing a document or making a spreadsheet. Not everyone will do it. But everyone could, if they wanted to.

Apple faces choices. They could try to restrict sideloading further, but that means fighting against a tide of users who simply want to run software they or their friends created.

The implications are significant. Apple's control over iOS software distribution has always rested on two pillars. First, the App Store as the only legitimate channel. Second, the high barrier to creating software in the first place.The second pillar just collapsed. And without it, the first pillar looks different. The App Store is not going away. It will remain the home of professional, polished applications. But it will no longer be the only place where iPhone software lives.

You need three things. A Mac with Xcode, which is free to download. A $99 per year Apple Developer account. And an AI tool that can write code based on your descriptions.

This way I made a lot of existing apps that I happily paid for absolutely obsolete. The stuff that I created was simply doing more of what I wished for, building on the ideas of all the apps I have seen before. A next iteration, but just for me.

So far I have built six different apps this way in the past few weeks. A personal transit tracker. A task manager tailored to my exact workflow. Small tools that solve specific problems in my life.

a voice app called Wispr Flow to talk to my computer in Dutch

Much of it feels like a monopoly hiding behind rhetoric about security and user experience.

The base case isn't resolution. It's controlled escalation with higher compliance spend, modest margin drag, and forced substitution as the biggest platforms build moats from regulatory complexity itself.

The wild card remains a behind-the-scenes "Digital Bretton Woods"—standardized frameworks for transparency, due process, and appeals that let both sides claim victory while lowering uncertainty.

Second-order effects create opportunities: vendors selling compliance plumbing (audit trails, policy ops, transparency systems) gain; European "sovereignty stack" providers (cloud, identity, data governance) benefit if retaliation shifts to procurement preferences over fines.

The splinternet thesis gets its Western chapter. Not US-China separation, but US-EU divergence inside allied markets—subtler, but more margin-destructive. Big Tech that can operationally bifurcate wins near-term. X-style political defiance loses because EU enforcers smartly choose process violations over content disputes.

That raises fixed costs and favors scale—paradoxically advantaging the largest platforms that can afford regional bifurcation while crushing subscale competitors.

Expect geo-fenced product design: "EU mode" platforms with different algorithmic defaults, transparency flows, and researcher access versus "US mode."

U.S. willingness to treat compliance as hostile action changes the calculation.

That increases risk premium on firms whose EU revenues depend on algorithmic distribution: social platforms, digital advertising, app stores, marketplace ranking.

The irony both sides miss: this conflict could entrench the very platforms Trump claims to defend and Europe claims to regulate. Compliance burden becomes incumbent moat.

The DSA doesn't mandate content removal based on viewpoint; it requires transparency in algorithmic curation, researcher access to platform data, and accountability for enforcement decisions. What the Trump administration calls "censorship," Europe frames as democratic governance of the digital public square—the same principle that makes "what is illegal offline illegal online."

a sharper conflict emerges: this is about who owns the distribution layer of democracy—who sets the rules for how speech gets amplified, throttled, demonetized, and made discoverable on platforms where most political discourse now occurs.

The EU had just levied a €120 million fine on X for DSA violations—the first major enforcement action under rules requiring platforms to moderate illegal content

Secretary of State Marco Rubio framed it as combating a "global censorship-industrial complex" targeting American platforms and speech.

visa restrictions on Thierry Breton, architect of the EU's Digital Services Act, alongside four anti-disinformation advocates: Imran Ahmed of the Center for Countering Digital Hate, Clare Melford of the Global Disinformation Index, and Anna-Lena von Hodenberg and Josephine Ballon of HateAid

The cost goes beyond simple inefficiency and becomes a mountain of invisible labor, usually absorbed by the most junior person in the room or whoever has the misfortune of being labeled as “good with computers.” It becomes a drag on every collaboration, the friction in every workflow, the meetings that take an extra ten minutes while someone (who is often paid twice the average salary of the other people in the meeting) figures out why they can’t access the shared folder the rest of us have been using for months. It’s the quiet erosion of patience and goodwill among people who are constantly expected to know and fix things that shouldn’t need fixing in the first place.

Imagine a carpenter who couldn’t figure out how to adjust their table saw, or a surgeon who shrugged and said something like, “I’m just not a scalpel person.” We would never accept that. But in the field of knowledge work, “I’m just not a tech person” has become a permanent identity instead of a temporary gap to be filled.

I’m talking about the basics: keyboard shortcuts that save hours per week. Understanding the difference between “reply” and “reply all.” Knowing how to search your own inbox or switch between work and personal accounts. Reading the words on your screen or an error message before throwing your hands up and declaring “something’s broken.” Learning how to unmute yourself or share your screen after years of being forced to do all our meetings on Zoom.

The number of professionals in journalism, media, communications, and academia who still don’t understand how to use the very tools they depend on for their livelihood is, frankly, staggering

I assumed we would all finally reach a baseline level of digital fluency. I could not have been more wrong.