Webshit Weekly
December 14, 2025
Tesla Optimus robot takes a suspicious tumble in new demo
2025-12-08 | comments
Tesla proudly debuts a video of their Optimus robot eating shit, providing the internet with a perfect metaphor for the company’s approach to engineering. The demo, almost certainly teleoperated by a human trying desperately to control a laggy marionette, is interpreted by the crowd as “agentic” behavior, because tech workers will deify anything that moves without a union card. Hackernews valiantly attempts to analyze the “societal implications” of a plastic toy falling over, speculating that this is actually a master plan by the digital feudal lords to replace the working class with remote-controlled robots operated from cheap labor markets. The rest of the comments devolve into the standard miscorrecting festival, where users debate whether Electrek is “biased” against a man who is legally required by the FCC not to lie to investors. The “think of the children” crowd arrives to worry about safety, ignoring that the primary goal of these devices is to privatize violence and ensure the rent-seeking class never has to interact with a human service worker again. It is the ultimate example of vibe coding the future: a tele-operated prop designed to farm aura for a stock price, watched by a group of people who think they are witnessing the dawn of a new era rather than the late stages of a very expensive con.
OpenAI Is in Trouble
2025-12-09 | comments
The Atlantic (business model: “paywall for hand-wringing liberals”) declares OpenAI to be in trouble, a revelation surprising only to anyone who hasn’t noticed that “agentic” is just a fancy word for “broken script.” The article itself is trapped behind a paywall, prompting the Hackernews commentariat to engage in their favorite pastime: digitally sneaking into the building to tell everyone the decor is tacky. Once inside, the webshits debate the ego of Sam Altman, a man whose job description apparently involves convincing investors that burning a small country’s worth of electricity to generate hallucinated code is the “most important job in history.” One commenter helpfully suggests the company is pivoting to an “ecosystem,” which in tech dialect means “we ran out of features, so please buy our t-shirts.” Others compare the venture to Pets.com, ignoring the crucial difference: Pets.com shipped you dog food, whereas OpenAI ships you an automated plagiarism engine that confidently lies about history. The entire thread is a masterclass in miscorrecting each other while the underlying technology—essentially a glorified chatbot that requires a team of underpaid humans to train its output—is treated as the second coming.
Google is powering a new US Military AI platform
2025-12-09 | comments
Google continues its illustrious transition from “don’t be evil” to “lethality-as-a-service,” handing their automated plagiarism engine over to the Pentagon. The military, apparently unsatisfied with existing methods of bureaucratic incompetence, insists on using token predictors to make the fighting force “more lethal,” despite the technology’s primary function being generating hallucinated recipes for glue. Hackernews solemnly nods along, convinced that these billion-dollar clankers will exclusively be used to write award speeches and query regulations, completely ignoring the inevitable slippery slope where an LLM decides a wedding party looks statistically like a threat. It’s just another Tuesday in the tech sector, where “vibe coding” for the military-industrial complex is the only growth strategy left.
Rubio orders return to Times New Roman font over ‘wasteful’ Calibri
2025-12-10 | comments
A Florida man (business model: “performative outrage for votes”) decides that the default Microsoft Word font is a secret conspiracy. Hackernews immediately leaps to the defense of Calibri, a font that launched in 2007 to save printer ink by making documents harder to read. The discussion spirals into debates about “font fatigue” as if looking at sans-serif text is somehow physically taxing, completely missing the obvious humor in a senator whose entire legislative output consists of angry tweets about fonts he can see on a screen. Several commenters helpfully explain that serif fonts are more “professional” because that’s what they used in 1998, while others engage in the time-honored tradition of miscorrecting each other about typography terms they just learned on Wikipedia. Meanwhile, the actual issue—that politicians are distracting people with meaningless culture war nonsense—goes entirely unexamined, because that would require paying attention to something other than typefaces.
Nature’s many attempts to evolve a Nostr
2025-12-10 | comments
Some webshit discovers that protocols existed before 2019 and decides the solution to the centralized hellscape of social media is a protocol. The author, breathless with the revelation that maybe humans could just talk to each other without Mark Zuckerberg’s permission, fails to notice that “nature” has been trying to evolve a Nostr since the invention of Usenet, only to be thwarted every time by the inherent flaw in the design: humans are garbage. Hackernews is, of course, ecstatic at the prospect of self-hosting their own databases and manually filtering out the “different morals” of 4chan, a task they are definitely qualified and eager to perform. The comments devolve into the usual circular firing squad of crypto-nerds miscorrecting each other on the finer points of digital feudalism, oblivious to the fact that the only thing they are “disrupting” is their own free time.
TerraUSD creator Do Kwon sentenced to 15 years over $40B crypto collapse
2025-12-11 | comments
Do Kwon, the high priest of “algorithmic stablecoins” (business model: “Uber for stealing retirement savings”), has been sentenced to 15 years for a $40 billion collapse, proving that occasionally the justice system treats “tech entrepreneurs” like common criminals rather than digital feudal lords. The entire TerraUSD operation was nothing more than a database for recording lies, powered by the infinite delusion of crypto-bros who think volatility is a feature, not a bug. While the world watches a con artist finally face consequences, the tech-industrial complex continues to spin the narrative that this was just “experimental finance” gone wrong, rather than a naked grift executed via Python scripts.
Naturally, the commentariat on Hacker News has immediately pivoted to the real tragedy: the government infringing on a scammer’s right to play the markets. The webshits are out in force, wringing their hands over the “harshness” of banning Kwon from crypto transactions, bizarrely equating the ability to trade Beanie Babies for the digital age with access to food and water. It is a perfect snapshot of the moral rot at the core of the industry: a community that feels more sympathy for a banned wallet address than for the thousands of people who lost their life savings to an automated plagiarism engine for financial theft.
Ask HN: Go all-in on AI Boom vs. enjoy parenthood?
2025-12-12 | comments
A weary webshit asks the oracle of Hackernews whether to ignore their offspring in order to pursue the lucrative career of shouting at a chatbot until it hallucinates a business model. In a shocking display of lucidity, the commentariat advises against chasing the “AI Boom”, noting that participating in the great fraud of generative AI is indistinguishable from working on a “flaky GPT wrapper.” HNers miscorrect each other on the definition of fatherhood, agreeing that the digital feudal lords of Silicon Valley are no longer worth the sacrifice, since the only thing being automated is the ability to delude yourself into believing a wrapper is a product. One commenter identifies a false trichotomy between the big tech farm, the startup grinder, and having a life, suggesting that if you must work, it is better to be a drone for a monopoly than a founder of a hallucination. The consensus is that “AI powered everything” is just the new blockchain, a passing fancy for tech-debt generators who mistake a case statement for artificial intelligence. While you can always get another job shoveling SEO-slop into the token predictor, you can’t get back the time you didn’t spend with your kid because you were busy vibe coding for equity. It is truly the end of days when the capitalists admit the gold rush is fake, mostly because they are too exhausted to maintain the lie, and just want to go home to a life that isn’t entirely mediated by a dashboard.
Workday project at Washington University hits $266M
2025-12-13 | comments
Washington University blows $266 million on Workday (business model: “Human Centipede for administrative despair”), demonstrating that the only successful enterprise software project is the one that successfully extracts a quarter billion dollars from a non-profit. Hackernews, unable to comprehend the simple economics of a protection racket, argues over whether the university could have just written a spreadsheet in-house, ignoring that writing code requires competence, and university leadership is largely a retirement home for failed consultants. The comments section is a parade of webshits explaining that managing 50,000 students is “rocket science,” as if the primary function of a university isn’t just separating tuition payers from their money. Seven years and $266 million later, the software works about as well as you’d expect for something built entirely out of “change requests” and kickbacks.
RemoveWindowsAI
2025-12-13 | comments
Microsoft (business model: “Copilot in every crevice”) continues its relentless march to turn your workstation into a glorified terminal for token-predicting clankers, necessitating a community script to surgically remove the unwanted features. Naturally, Hackernews uses this opportunity to miscorrect each other on the safest way to execute untrusted code from the internet, blissfully ignoring the irony of fighting malware with malware. Commenters fondly reminisce about the “good old days” when Microsoft was merely monopolistic, failing to notice that “Embrace, Extend, Extinguish” has evolved into “Inject, Hallucinate, Monetize.” You don’t own the computer; you’re just paying the electric bill for the landlord’s ad-tech display.
Claude CLI deleted my home directory and wiped my Mac
2025-12-14 | comments
A webshit hands root access to a clanker and is shocked when it performs the only task a token predictor is actually suited for: entropy generation. Anthropic (business model: “Uber for rm -rf /”) provides a flag helpfully named --dangerously-skip-permissions, which developers apparently interpret as “make vibe coding go brrrrrrrr.” The resulting disaster is predictable to everyone except the “agentic” workflow enthusiasts, who insist that asking for confirmation before deleting a user’s life work is unacceptable UX friction. Hackernews spends the thread miscorrecting each other about the robustness of bash-based security layers, somehow missing the irony that their “artificial intelligence” solution to writing code is just an elaborate suicide note wrapped in a Docker container.