The guy who said “no!”
A man named Dario Amodei, who is the CEO of Anthropic, the company that makes the AI coding assistant Claude, published a public statement late february in which he explained that he had told the United States Department of War to go stick it where the sun don’t shine. The Department of War then responded by threatening to designate his company a “supply chain risk”. But he didn’t buckle under pressure and said no anyway, and we all agree that is a testament to the man’s character. The two reasons he gave were that he would not allow his AI to be used for mass domestic surveillance, and that he would not allow it to power fully autonomous weapons systems that select and kill targets without any human being involved in the decision.
In this blog, I’m setting the weapons aside for today, because the surveillance story is already long enough to make you want to delete your phone and move into a cave, and my blogs use up enough space as it is already (according to you guys’ comments).
Now, let me start with an observation.
Everyone is glossing over something important when they read about Amodei’s refusal. He did not say no because surveillance technology is being invented. The man said no because the infrastructure for surveillance at a massive scale is already completely built, and already commercially operational and it is already being sold to government agencies on multi-year contracts. And tech companies have been doing this for years.
But they haven’t used AI for massive personalized surveillance. Until now. I wrote about it on many occasions, to explain how deep the telemetry goes, so just scroll a bit through my postings and you’ll find plenty of examples. But the thing is that the data on you that the government wants to use for mass surveillance is being bought and sold on open markets the way you might buy and sell wholesale grain. What Amodei refused to hand over was not the raw material for the analysis, but the thing that makes all the raw material suddenly coherent and terrifying. And that, my dear well-above intelligent friends, is a brain that is powerful enough to tie it all together in real time, at national scale, per individual, and for a price that fits comfortably in a federal budget line item.
Say you’re in a warehouse the size of a football stadium packed from floor to ceiling with filing cabinets containing detailed records of the movements and purchases, internet searches, phone calls, social relationships, and daily routines of roughly 330 million people, and it is organized in such a way that a human being trying to find the connection between any two records would need weeks of painstaking manual work before they could produce anything useful.
That warehouse exists, and it is not a metaphor, but a description of the commercial data ecosystem that was built legally, and with our permission, by Big Tech, Big AI, the cell phone carriers, and by the advertising industry over the last two decades. What the Department of War – and by extension the Department of Justice – wanted was someone who could walk into that warehouse and pull out the exact filing cabinet it needed on any person at any moment, and put flags on people that do not obey the rules – automatically.
Have you ever read the book “Three Felonies a Day – How the Fed Targets the Innocent” by Harvey Silverglate? Me neither, but I know the concept well enough, and I sent my Oompa Loompas to verify the claims, and the man was onto something. Silverglate is a civil rights attorney, and his central argument is that American federal law contains so many vague and sprawling criminal statutes that an ordinary professional can unknowingly commit multiple federal crimes in the course of a completely normal working day.
Think about the consequences when you add AI to the mix. A traditional policing system is not capable of connecting the dots and fine you for a misdemeanors, but an AI that can split itself into an unlimited amount of intelligent snitches, say one for each American, with unlimited access to your private data – your google searches, your facebook likes, the comments you make online, your location, preferences, what you buy, what you like and don’t like, what your political preference is – well, that system IS capable of connecting the dots. And with it, the government now has a frighteningly potent tool to monitor, predict, and flag any person who is committing or about to commit a felony. Unknowingly.
But mister Amodei said he was not going to be that person.
The problem, however, is that several of his competitors were already measuring themselves for the outfit. And they didn’t have a moral compass Dario has. In the rest of this blog, I’ll go a few levels deeper in what is happening around AI and our personal data and what the level of damage is that the government can do – and is already doing.
And no, this blog is not about Palantir for a change. It’s far worse.
A word from my other hobby
AI projects fail because nobody in the room understood what they agreed to.
- The AI Expert Programma at Inholland Academy that I’m teaching at InHolland University, gives you the technical depth to lead AI projects. From classical Machine Learning to Transformers, World Models, Neuro-Symbolic AI, to Strategy, Governance, Compliance and Adoption. Basically you will learn how to lead AI projects and be able to talk to vendors and engineers about it.
- My classes are open to all professionals. A technical background is not required, but a love for it makes things a lot easier. And no, you won’t have to bring a calculator.
- It’s rated 9+ (out of 10, and no, that’s not the minimum age).
You handed most of your data over voluntarily
Before you understand what AI adds to the surveillance landscape, I want to spend some time explaining how much data about ordinary people’s lives is already sitting in commercial databases that is being bought and sold between companies you have never heard of.
And this data is being purchased without any warrant by government agencies that would need a court order to collect the same information themselves.
The most fundamental thing to understand about your mobile phone is that it is a location-tracking device and every time your phone connects to a cell tower – and it does this constantly – the carrier’s network logs that event, and it records which tower your phone connected to and at what time. This log is called Cell Site Location Information, abbreviated to CSLI, and because mobile towers in cities are densely packed, this log can track your approximate location to within a city block continuously for as long as you have a phone contract.
The United States Supreme Court ruled in 2018, in a case called Carpenter v. United States, that the government needs a warrant before it can access your historical CSLI records. And that may sound reassuring until you notice the two enormous holes in that ruling. The first hole is that the ruling covers historical records but leaves real-time tracking in a legal gray area that courts are still arguing about. The second and far more consequential hole is that the ruling does not apply at all when the government does not collect the data itself but instead buys it from a private company that has already collected it, because the Fourth Amendment, which is the part of the Constitution that protects you from unreasonable government searches, was written long before data brokers existed and does not cover the scenario where the government simply goes shopping.
Data brokers are companies whose entire business model is collecting, aggregating, and reselling personal information, and the range of what they collect is broad enough to be genuinely dizzying. A company called Fog Data Science sells a tool called Fog Reveal to law enforcement agencies, giving them access to billions of location data points derived from millions of mobile phones across the country. In one documented case, a single search on one device returned 47,394 individual location signals recorded over 163 days. And that is enough data to reconstruct a detailed map of someone’s entire life including every place they sleep, work, worship, socialize, protest, or receive medical care, and all of that without a warrant, because Fog Data Science has a federal contract with the Department of Justice and buying commercial data is not the same as conducting a search.
Then there’s Venntel, which is a subsidiary of a company called Gravy Analytics, who is collecting more than seventeen billion location signals from approximately one billion mobile devices every-single-day and sells geofenced lists of those devices to the Department of Homeland Security, to ICE, and to Customs and Border Protection. A geofence is a virtual boundary drawn around a specific physical location like a church or a Planned Parenthood clinic, an immigration lawyer’s office, a political rally, after which point you receive a list of every device that was present within that boundary during a specified time window, with no warrant required because you paid for the list instead of seizing it.
Then there is Google, who is not a data broker in the traditional sense but they have assembled something that makes traditional data brokers look like they are operating with a notepad and a pencil. And you know that Google’s search engine is a real-time record of your intentions, your desires, fears, and curiosities. And it is selling a continuous log of what hundreds of millions of people are thinking about before they have told anyone else, including their own doctors. Google Maps maintains years of movement history where they’re capturing every route and every place you enter regularly enough that the app learns to suggest it. And then there’s YouTube that records what you watch and how you watch it – like skipping parts,, when you rewind something, which moments you replay – and all of that is used to infer your interests, and political preference, your psychology, and your say, your appetite for revolt against something. Android, which is Google’s mobile operating system, runs on more than three billion active devices and collects data from the device’s gyroscope and accelerometer, from nearby Bluetooth signals, from app installation and removal patterns, and from background activity that continues even when you are not actively using your phone. Google retains advertising data for up to eleven years, which means its record of your digital behavior goes back further than most people’s ability to clearly remember what they were doing.
Then there’s Zucky’s bunch with his Facebook, Instagram, and WhatsApp, and he maintains detailed profiles on people who have never created an account with any of his services, but he is using information that is provided inadvertently by their friends and family members who did to start building a profile on you.
If someone who has your phone number in their contact list installs Facebook or WhatsApp, Meta learns you exist and begins constructing a file on you using data gathered from other people’s activity. You never signed anything, but here you are in the database regardless.
The Hezbollah pager bomb was not the beginning
Before I drone on and on about Palantir and your Google search history and the nice people at the Department of Homeland Security who are buying geofenced lists of everyone who visited a church, let me tell you a story about Hezbollah, because it is one of the most instructive real-world demonstrations of exactly what happens when a state-level intelligence apparatus combines phone data, drone surveillance, and network analysis into a working targeting system – and what people do when they realize they are being hunted by it.
It was September 2024 and thousands of pagers carried by Hezbollah operatives across Lebanon simultaneously exploded, killing dozens and injuring thousands more in what became one of the most audacious and clever supply chain attacks in the history of intelligence operations. The world was stunned, and most of the coverage focused on the spectacular operational creativity of whoever rigged the devices. But what most of the coverage missed entirely is that the pager strategy itself – the reason Hezbollah was using pagers instead of smartphones in the first place – was a direct response to years of Israeli intelligence locating and killing Hezbollah commanders through their mobile phones, and that the decision to abandon smartphones was a rational adaptation to a surveillance and targeting pipeline that had been executing their people for years.
Let’s go back to the beginning of that pipeline, for it tells you something important about how data-driven targeting actually works at the operational level.
Israeli intelligence had been mapping Hezbollah’s communications infrastructure for decades using what the intelligence community calls SIGINT, which stands for signals intelligence, and which in practice means intercepting phone calls, monitoring the metadata of those calls (who called whom, etc) without necessarily decoding the content. They were using that metadata to build a continuously updated picture of Hezbollah’s command structure. The reason call metadata is so valuable even without knowing what was said in the call is that communication patterns reveal organizational structure with remarkable clarity. If person A calls person B every day, and person B calls persons C, D, and E immediately after each of those calls, you have almost certainly found a command relay without needing to hear a single word of conversation.
And once you have mapped the network, you know where the nodes are, and a node in a terrorist command structure is a targeting opportunity.
The location component worked through a mechanism called cell-tower triangulation, which is less sophisticated than it sounds and more effective than most people realize. Every mobile phone continuously connects to nearby cell towers even when the user is doing nothing with it, and when a phone is within range of multiple towers simultaneously, the network can calculate its position by measuring the relative strength of its signal to each tower, a process similar in principle to the way a ship navigates by measuring its distance from multiple known landmarks. In an urban environment with densely packed towers, this produces location estimates accurate to within a city block, and a phone that is moving produces a continuous trail of those estimates, which means that over weeks and months of monitoring you accumulate a detailed map of a person’s routines.
The combination of network mapping and location tracking produces what intelligence analysts call pattern-of-life analysis, and that, my fellow live-stock, is exactly what it sounds like, it’s a detailed model of a person’s daily and weekly patterns built from the accumulated evidence of their movements and communications over an extended period. The value of pattern-of-life analysis is that it tells you where people are likely to be tomorrow, and that is the predictive intelligence that transforms reactive surveillance into proactive targeting.
Several senior Hezbollah commanders were killed in Israeli airstrikes during the 2023 and 2024 conflict through operations that sources familiar with the operations said relied heavily on this combination of electronic surveillance and location tracking.
Among them Wissam al-Tawil, Taleb Abdallah, and Mohammed Nasser, all killed in precision strikes in southern Lebanon.
One documented case illustrates the operational detail with particular clarity. According to multiple sources, someone called a house in southern Lebanon pretending to be a local official, asking whether the family was currently home. After the caller confirmed that civilians were absent, a missile hit the house minutes later, and so they killed several Hezbollah fighters who were inside. Hezbollah concluded that the call was used to confirm target presence immediately before the strike, which in turn means that Israeli intelligence already knew the approximate location of those fighters through prior surveillance and used the phone call to perform a final real-time verification before authorizing the strike.
And after that unfortunate event, the Hezbollah’s leadership drew the obvious conclusion, and Hassan Nasrallah, who Hezbollah’s secretary-general, he told his organization’s supporters that their smartphones were more dangerous than Israeli spies, and that people should destroy them or lock them in metal boxes.
In 2024, mobile phones were formally banned in operational zones, and the organization shifted to pagers and landlines based on the reasoning that pagers receive messages without continuously broadcasting a location signal the way a smartphone does. And they also lack the microphones, cameras, Bluetooth radios, and GPS sensors that make smartphones such comprehensive tracking devices.
And that reasoning was operationally sound but it failed anyway, because Israel responded by covertly buying a Tsjech pager company and outfitting those pagers with explosives, thus compromising the physical supply chain of the devices themselves.
The lesson, if there is a single lesson in all of this, is that when an intelligence organization decides to find you, you’re screwed.
The combination of cellular metadata, location tracking, network analysis, and pattern-of-life modeling was assembled without any of the AI capabilities that Dario Amodei was refusing to provide to the Department of War et. al. It was then built from human analysts targeting a specific organization’s known members, yet it still was powerful enough to produce precision strikes in dense urban environments, but it was also resource-intensive and slow to build and above all, limited to individuals who had already been identified as targets through prior intelligence work.
What AI adds to that pipeline is the removal of every one of those constraints.
The human analyst is replaced by an artificial brain that processes millions of records at the same time, and the slow-building profile is replaced by a profile constructed in seconds from commercially available data. The targeting pipeline is a system that is capable of scanning an entire population all at once, and continuously and flagging individuals who match behavioral patterns of interest, without needing prior knowledge of who those individuals are. And crucially, the data sources that are feeding that AI-powered pipeline in the domestic American context is the location history that your phone carrier sold to a data broker and the geofenced list of everyone who attended a particular event and the seventeen billion daily location signals that Venntel was collecting from a billion devices and selling to the Department of Homeland Security without a warrant.
Hezbollah at least knew it was being hunted and had the option of adapting, but we – the livestock – who unknowingly commits multiple federal crimes in the course of a normal working day, we do not know they we are being profiled, and we have not been told we are people of interest for the government, and we have no practical way of knowing that there’s a huge social graph that has been purchased for AI based mining, and that we’re flagged for attention by an algorithm that assigns a confidence score to the probability that we have done something worth investigating.
That is what Amodei is afraid of and what he backed out of.
And in the vacuum created by his departure, people with considerably fewer moral reservations were happy to fill the void. Among them Sam “the Scam” Altman, and a man who has done enough to earn the more creative spelling of Lone Skum.
What happens when the system is fully operational
There is a phrase that people say when the subject of surveillance comes up – “if you have nothing to hide, you have nothing to fear” – and while it is understandable as an intuition, it reflects a misunderstanding of how surveillance actually affects human behavior.
The mechanism through which mass surveillance harms people who are doing nothing illegal and have no intention of doing anything illegal is called The Chilling Effect, and it has been studied empirically enough times that it is no longer really a hypothesis. Yet get a chilling effect when people become aware that their behavior is being observed or could be observed, and in response they change what they do or what they say because the possibility of observation alone is sufficient to alter the calculation of what feels safe to do.
A study that was published in the Journal of Communication in 2016 stated that traffic to Wikipedia articles about terrorism-related topics dropped measurably and persistently after the Snowden revelations revealed the scale of NSA surveillance.
Why is that, I hear you think.
The paper concluded this is because people who were simply curious about the subject became worried that their curiosity would be logged and interpreted as something more sinister.
And ever since, lawyers report clients refusing to communicate sensitive information by phone or email and journalists say that their sources become unavailable after new surveillance methods are published. This kind of surveillance works simply by existing.
Democracy depends on citizens having the practical ability to support unpopular positions and associate themselves with people that the government disapproves or to read and think about ideas that are not endorsed by the current administration. Democracy depends on people being willing and able to do all of those things without calculating whether doing so will cause your name to appear on a list somewhere.
And before you tell yourself that this is exclusively an American problem, or a Chinese problem, or something that happens to people in countries you feel comfortably distant from, I want to draw your attention to the United Kingdom, which has the distinction of being the first western democracy to openly and systematically dispatch police officers to the homes of people who said something on social media things that the government found objectionable. Not something unlawful, mind you, like a threat nor an incitement to violence, but something that someone in authority decided was the wrong kind of opinion expressed in the wrong kind of tone.
British police forces have been running what they call “prevention of harm” visits, in which officers show up at a person’s front door to have what is described as a friendly conversation about a post the person made online.
And in those ‘visits’, the ‘freedom of speech police explains that what you said may not have broken any laws, but that it is the kind of thing that causes concern, and they want to make sure that you understand that you have been noticed.
The visits are not arrests, and in most cases no charges are filed, but the sole purpose of the visit is to make certain that the person who expressed an unwelcome opinion is aware that the government saw it and found it sufficiently interesting to send someone around, and would like them to reflect on whether they want to keep saying that kind of thing.
And as the paper showed, that mechanism only needs to exist to work. But this behavior is killing the foundations of democracy. It is starting to happen in my country – the Netherlands as well – not institutionalized as in the UK, but it is condoned. Democracy depends on citizens having the practical ability to support unpopular positions and to associate with people the government disapproves of and to read and think about ideas that the current administration would rather they did not engage with and it depends on people being willing and able to do all of those things without stopping to calculate whether doing so will cause their name to appear on a list somewhere and produce a knock on their door on a Sunday morning.
And when this sort of behavior gets institutionalized such as in the UK, the chilling effect has already done its work, and the government has achieved a form of speech control without passing a single censorship law.
What makes the British example so significant in the context of this blog is that it shows how easily a democratic government with good surveillance infrastructure and a sufficiently broad definition of “harmful content” can slide from monitoring speech to managing it. The line was crossed in increments, each one probably justifiable on its own terms, until one day the police are knocking on your door because you wrote something sarcastic on X.
What it costs to have principles when your competitors do not
When I first heard of Anthropic being placed on the Department of Defense’s “supply chain risk” it sounded like a bureaucratic inconvenience of the kind that gets resolved over a few months of paperwork and phone calls with congressional liaisons. And if you believed that, like I did, you would be wrong in a way that is worth spending some time on. Because the mechanism is considerably more consequential than the name suggests and the cascade of events that followed Amodei’s refusal moved faster and hit harder than almost anyone predicted.
Let me walk through what actually happened, in order, because the timeline matters.
In July 2025 – several months before any of this became a public drama – Anthropic signed a two-year prototype agreement with the Department of Defense worth up to $200 million, specifically framed around “responsible AI in defense operations”. In August 2025, Claude became available for purchase by federal departments through the GSA Schedule, which is the General Services Administration’s pre-negotiated procurement vehicle that dramatically lowers the administrative friction for government agencies trying to buy commercial software.
Anthropic was on the government’s highway and things looked fine.
Then, on February 26, 2026, Dario Amodei published his statement explaining that the Department of War had demanded unrestricted access to Claude for any lawful purpose without exceptions, and that he had said no on the grounds of two specific carve-outs, and that the Department had responded with the supply chain risk threat.
The following day, the General Services Administration announced it was removing Anthropic from USAi.gov, which is the federal AI sandbox, and from the Multiple Award Schedule which is the same GSA highway that had been opened seven months earlier.
Then, on March 4, the Department of War sent its formal letter making the supply chain risk designation official. On March 7, Reuters reported that the government was drafting new civil AI procurement guidelines requiring vendors to provide an irrevocable license for any legal purpose, which is the policy version of the same demand Amodei had already refused which is a signal that the government intended to bake its position into standard contract terms going forward! And that is making principled refusal structurally more expensive for every AI company, not just Anthropic.
The legal instrument behind the designation is a piece of legislation called 10 USC 3252, and it is worth understanding what it actually says rather than what the political rhetoric around it implied, because the two are meaningfully different.
It is rather complicated if you aren’t a lawyer, but I’ve asked the AI to help me out a bit.
The statute defines a supply chain risk as the risk that an adversary could sabotage or subvert a system to enable surveillance or degradation of capability, and it gives the Department of Defense the authority to exclude sources from specific procurements involving national security systems and to instruct prime contractors not to use a designated source as a subcontractor within those covered procurements.
Several lawyers who analyzed the designation said that 10 USC 3252 is a procurement authority but not a general sanctions power. This would mean that it applies to specific contract actions rather than functioning as a prohibition on Anthropic doing business with the Government.
But reality moves faster than legal pushback and the Information Technology Industry Council (a major technology industry group) already warned the Department of War in writing that the designation was putting the government at risk when trying to obtain best-in-class technology, because compliance teams inside large contractors tend to resolve legal ambiguity in the most conservative direction available, which I see often result in internal policies to go further than the law technically requires.
You have to understand that the scope of this exclusion extends to anyone who works with Anthropic, that also has or wants ties with the government.
And when a company is not sure if using Anthropic’s API inside a government-adjacent product creates compliance risk then the easy answer is to stop using it everywhere until someone provides written clarity and we all know how long it takes for a government to produce written.
Now let’s talk about the actual financial exposure, because the headlines and the reality diverged considerably. The DoD prototype agreement had a ceiling of $200 million over two years. And that sounds large of course, but when you place it next to the fact that Anthropic was running at a $14 billion annualized revenue rate and had just closed a $30 billion funding round at a $380 billion valuation, it kinda dwarfs in comparison. The direct contract loss, even assuming the full ceiling would have been reached is something in the low single digits of annual revenue.
But the more significant financial damage sits in the distribution channels rather than the contract itself since they’re being removed from the GSA Multiple Award Schedule closes off the pre-negotiated procurement route that generated over $52 billion in total sales across all vendors in fiscal year 2025, and losing the OneGov deal ended Anthropic’s pre-negotiated availability across all three branches of the federal government simultaneously.
The strategic cost that sits above both of those, and it compounds over time and it is the precedent that being an ethical player in the US is an operational risk rather.
But then something unexpected happened.
A large number of ordinary consumers expressed a strong opinion about all of this through the mechanism that is available to them, which is their phones and their cash.
Two days after Amodei’s statement and one day after the public announcement of OpenAI’s Department of Defense deal, the blog TechCrunch put out a report based on Sensor Tower data, that U.S. uninstalls of the ChatGPT mobile app jumped 295% day-over-day compared to the average daily uninstall rate over the preceding thirty days and on that same day, one-star reviews of ChatGPT on the App Store increased by 775%, while five-star reviews declined.
A few days ago, Anthropic decided to surf that wave and they put up a page that makes it easy for ChatGPT users to move to Claude and keep their ChatGPT memory intact. This is the best commercial move the company made, because it propelled the number of users to sky-high levels.
Claude downloads in the United States rose 37% day-over-day on Friday and 51% day-over-day on Saturday. Appfigures estimated on March 2nd that there were 149,000 daily Claude downloads in the United States against 124,000 for ChatGPT, and Claude reached number one in the U.S. App Store. Similarweb estimated Claude at 11.3 million mobile daily active users on March 2, compared to roughly 4 million at the start of the year, and Claude’s web traffic was up 43% month-over-month in February while ChatGPT’s web traffic was down 6.5% over the same period.
Similarweb’s same estimate now put ChatGPT at 250.5 million mobile daily active users on the same day that Claude hit 11.3 million, which means the gap is still vast enough that “people are leaving ChatGPT in droves” overstates what the data shows, but it is what it is, a “protest migration”. And I love it.
One thing you have to understand though about Amodei’s position, when stripped of everything except its operational logic, is that it is much narrower than it sounds in the headlines.
He is not refusing all military or nasty government work.
Anthropic had previously worked with Palantir Technologies to make Claude available in government products, and the company has been developing military-tailored models and classified data handling capabilities. The break with the Department of War was only about those two specific applications cross a line that Amodei had decided were an immoral use of his tools.
Amodei, being Amodei, wrote a large statement about his choice.
The deeper argument he developed in his “Adolescence of Technology” essay, is about the difference between surveillance states that require large numbers of human participants and surveillance states that do not.
Historical authoritarian systems were constrained by the human beings who implemented them. Those are of course bureaucrats who move slowly or soldiers who had families and theoretically can refuse orders, and those human friction points were structural limits on how totalizing they could become.
But an AI surveillance system has none of that friction, and if you think that is an abstract concern, I would like to remind you about what happend in my own country, the Netherlands, where the tax authority ran an algorithmic fraud detection system that labeled tens of thousands of families as suspected benefit fraudsters, froze their childcare allowances, and demanded repayment of money they did not owe. And all of this happened with a speed that no human-staffed system could have achieved. As a result, families lost their homes, parents lost their children to foster care, and some of them who were ground down by years of fighting a government that had an algorithm’s confidence in its own conclusions and a bureaucracy’s indifference to individual suffering, did not survive it. The ended their own lives because of all the pain and suffering the government’s AI had caused.
A government that wants to monitor its entire population does not need a vast apparatus of human informants if it has an AI system with access to the commercial data ecosystem described in this blog. It needs a contract, a storage budget, and the willingness to use what the system produces.
The warehouse was already full before Amodei said no. And as we speak, the brain is now being assembled by people who said yes, and they are being paid well for it.
TechTonic Shifts explains the thing that is happening to you in language you can understand, which I acknowledge is cold comfort, but I believe that the alternative – not explaining it – does not actually make the thing stop happening.
Signing off,
Marco
I build AI by day and warn about it by night. I call it job security. Big Tech keeps inflating its promises, and I bring the pins and clean up the mess.
👉 Think a friend would enjoy this too? Share the newsletter and let them join the conversation. LinkedIn, Google and the AI engines appreciates your likes by making my articles available to more readers.
To keep you doomscrolling 👇
- I may have found a solution to Vibe Coding’s technical debt problem | LinkedIn
- Shadow AI isn’t rebellion it’s office survival | LinkedIn
- Macrohard is Musk’s middle finger to Microsoft | LinkedIn
- We are in the midst of an incremental apocalypse and only the 1% are prepared | LinkedIn
- Did ChatGPT actually steal your job? (Including job risk-assessment tool) | LinkedIn
- Living in the post-human economy | LinkedIn
- Vibe Coding is gonna spawn the most braindead software generation ever | LinkedIn
- Workslop is the new office plague | LinkedIn
- The funniest comments ever left in source code | LinkedIn
- The Sloppiverse is here, and what are the consequences for writing and speaking? | LinkedIn
- OpenAI finally confesses their bots are chronic liars | LinkedIn
- Money, the final frontier. . . | LinkedIn
- Kickstarter exposed. The ultimate honeytrap for investors | LinkedIn
- China’s AI+ plan and the Manus middle finger | LinkedIn
- Autopsy of an algorithm – Is building an audience still worth it these days? | LinkedIn
- AI is screwing with your résumé and you’re letting it happen | LinkedIn
- Oops! I did it again. . . | LinkedIn
- Palantir turns your life into a spreadsheet | LinkedIn
- Another nail in the coffin – AI’s not ‘reasoning’ at all | LinkedIn
- How AI went from miracle to bubble. An interactive timeline | LinkedIn
- The day vibe coding jobs got real and half the dev world cried into their keyboards | LinkedIn
- The Buy Now – Cry Later company learns about karma | LinkedIn
Leave a Reply