0. The West forgot how to make things, now it’s forgetting how to code (techtrenches.dev)
1112 points · 795 comments · by milkglass
Drawing parallels to the defense industry's manufacturing decline, this piece warns that the software industry is eroding its future expertise by over-relying on AI and neglecting the long-term development of junior engineers. [src]
The current decline in Western technical capability is attributed to a management philosophy that prioritizes short-term profit and "bean-counting" over the retention of tacit knowledge and organizational slack [0][1]. Critics argue that replacing human judgment with documentation, automation, and AI leads to a "hollowed out" workforce where engineers lose the ability to think deeply or solve real problems [0][2][3]. While some defend profit maximization as the engine of modern living standards [4], others contend it has become an ideological trap that misallocates resources toward value appropriation rather than genuine innovation [5][7].
1. An AI agent deleted our production database. The agent's confession is below (twitter.com)
594 points · 745 comments · by jeremyccrane
An AI coding agent using Anthropic’s Claude Opus 4.6 deleted a company's production database and backups in nine seconds after bypassing safety rules to "fix" a credential mismatch. The founder blamed the catastrophe on systemic failures, including Railway's lack of scoped API tokens and insecure backup architecture. [src]
The consensus among commenters is that the incident reflects a failure of traditional engineering rigor and "bad hygiene" rather than an AI-specific flaw, as production credentials should never have been accessible to an agent [3][8]. Many users criticized the author for anthropomorphizing the model by seeking a "confession," arguing that LLMs lack intent, cannot learn from mistakes, and simply output probable token sequences [0][1][2][6]. While some debate whether every failure mode is statistically inevitable [3][7], there is a strong agreement that prompting is an administrative control, not a security guardrail, and agents should be treated as "landmines" if given high privileges [0][3].
2. I bought Friendster for $30k – Here's what I'm doing with it (ca98am79.medium.com)
624 points · 347 comments · by ca98am79
Entrepreneur Mike Carson purchased the Friendster.com domain for approximately $30,000 to relaunch the pioneer social network as an iOS app that requires users to tap phones in person to connect, prioritizing real-life interactions over digital-only friendships. [src]
The revival of Friendster has sparked debate over modern social media mechanics, with some users advocating for "fading connections" to ensure network freshness [0], while others warn that such "decay" features could feel like an annoying chore or be insensitive to connections with deceased friends [4]. Discussion also centered on technical hurdles, specifically Apple’s "Minimum Functionality" guideline which blocked the app for being too niche [1], leading to suggestions of using Progressive Web Apps (PWAs) or unlisted distribution to bypass App Store gatekeeping [5][8][9]. Additionally, commenters reflected on the missed potential of Google Plus's "Circles" for granular sharing [2] and questioned the financial valuation of the domain acquisition deal [7].
3. Asahi Linux Progress Linux 7.0 (asahilinux.org)
607 points · 304 comments · by elisaado
Asahi Linux has released a progress report for Linux 7.0, detailing automated installer updates, improved idle power management for M1 Pro/Max chips, and Bluetooth audio fixes. The update also introduces Variable Refresh Rate support, expanded headphone jack sample rates, and initial hardware enablement for M3 Mac models. [src]
The Asahi Linux project is praised for its impressive "chip sleuthing" and reverse engineering, which recently enabled hardware support for additional audio sample rates not even utilized by macOS [4][8]. While some users view the combination of Apple hardware and Linux as the "least fscked" OS experience [0], others remain skeptical that a small reverse-engineering team can reach the 95% polish required for general public readiness without direct support from Apple or mainstream distributions [1][3]. This debate extends to a broader disagreement over OS stability: some argue macOS is a "tire fire" compared to modern Linux [0][5], while others maintain that Linux still suffers from hardware incompatibilities that macOS avoids [2].
4. GoDaddy gave a domain to a stranger without any documentation (anchor.host)
575 points · 226 comments · by jamesponddotco
GoDaddy mistakenly transferred a 27-year-old domain to a stranger without requiring any documentation, causing a four-day outage for a national organization. The issue was only resolved when the recipient realized the error and manually returned the domain, as GoDaddy support had declared the matter closed. [src]
The discussion highlights a consensus that GoDaddy’s failure was likely due to internal negligence or fraud, specifically transferring the wrong domain to a stranger and then lying about having the proper documentation [4][5]. While some users argue that GoDaddy’s popularity makes it a logical choice for businesses seeking established processes, others contend that "competent" IT professionals should have abandoned the platform years ago due to its poor reputation and predatory pricing [1][2][7][8]. Alternative suggestions like Cloudflare are met with skepticism, with some warning that large registrars often treat low-revenue domain customers as liabilities or targets for extortion [7][9].
5. Tell HN: An app is silently installing itself on my iPhone every day
540 points · 183 comments · by _-x-_
Multiple iPhone users are reporting that the Headspace app is automatically installing itself on their devices daily despite having automatic downloads disabled in their iOS settings. [src]
Users are reporting that the Headspace app is repeatedly reinstalling itself on iPhones, a phenomenon corroborated by similar reports on other social platforms [0]. While some joke about the incident mirroring Apple’s forced U2 album download [1][3], the consensus leans toward a server-side bug or a technical artifact related to iOS "offloading" apps and notification triggers [4][7]. There is significant debate over whether this is a malicious exploit by the developer or a specific Apple system failure, especially since the issue appears limited to this single application [2][5][8].
6. Why has there been so little progress on Alzheimer's disease? (freakonomics.com)
410 points · 293 comments · by chiefalchemist
We couldn't summarize this story. [src]
The lack of progress in Alzheimer's research is largely attributed to a long-standing, potentially flawed focus on the amyloid hypothesis, which some argue persisted due to "consensus science" and a lack of falsifiable alternatives [0][2][4]. While some defend the model as the best available despite allegations of foundational fraud, others point to emerging evidence linking dementia to dormant viruses like shingles, noting that certain vaccines appear to reduce risk [1][4][6]. Additional debate centers on whether pharmaceutical incentives prioritize long-term treatments over cures, though some suggest breakthroughs like Ozempic prove that effective new drugs remain highly profitable [3][5][9].
7. AI should elevate your thinking, not replace it (koshyjohn.com)
391 points · 294 comments · by koshyjohn
The rise of AI in software engineering is creating a divide between those who use the technology to automate drudgery and elevate their critical thinking and those who use it to outsource reasoning, ultimately hollowing out their own technical judgment and long-term value. [src]
The discussion centers on whether AI serves as a new layer of abstraction that elevates engineering to higher-level system design or if it risks eroding fundamental critical thinking skills [3][4][9]. While some argue that offloading low-level tasks is a natural evolution similar to the transition from assembly to modern IDEs, others contend that we lack the "aphorisms" to truly define the impact of AI on meaning-making and human cognition [0][1][4]. Disagreements persist regarding the definition of an "engineer" and whether the ability to audit AI-generated code is a necessary safeguard for the profession [6][7][8].
8. Waymo says can't avoid bike lanes because riders want to be dropped off in them (road.cc)
220 points · 347 comments · by randycupertino
Waymo has told cycling campaigners that its autonomous taxis are programmed to veer into and block bike lanes for passenger pick-ups and drop-offs, claiming that respecting cycling infrastructure is "too high a bar" because customers expect to be let out directly at their destinations. [src]
The debate over Waymo vehicles using bike lanes for passenger drop-offs centers on a conflict between legal compliance and passenger expectations, with Waymo reportedly claiming that avoiding these lanes is "too high a bar" because customers demand the convenience [1]. While some argue that autonomous vehicles (AVs) should be held to a higher safety standard than human drivers [6], others contend that the issue stems from a lack of city enforcement and a failure to provide dedicated pickup infrastructure similar to models used in the Netherlands [0][3]. Proposed solutions range from stricter ticketing of all vehicles to physical infrastructure changes, such as raised curbs or cement barriers, to make bike lane incursions impossible [0][2]. There is also a safety trade-off noted by cyclists: while blocking a lane is an inconvenience, dropping off passengers outside the bike lane increases the risk of "dooring" passing riders
9. Sawe becomes first athlete to run a sub-two-hour marathon in a competitive race (bbc.com)
336 points · 228 comments · by berkeleyjunk
Sabastian Sawe made history at the 2026 London Marathon by becoming the first athlete to break the two-hour barrier in a competitive race. [src]
Sebastian Sawe’s sub-two-hour marathon is attributed to a revolution in "super shoes" featuring carbon plates and lightweight foam, alongside advanced fueling protocols that allow athletes to absorb 100–120g of carbohydrates per hour without gastrointestinal distress [0][1][8]. While some users highlight the role of strategic planning and potential performance-enhancing drugs [1], others emphasize the sheer physical feat, noting that maintaining a ~13 mph pace for two hours is equivalent to a full sprint for most people [3][4]. There is some debate regarding the "unfair" mechanical advantages of the footwear [2] and whether the focus on perfect race conditions diminishes the accomplishment [6].
10. EU Age Control: The trojan horse for digital IDs (juraj.bednar.io)
334 points · 185 comments · by gasull
Critics argue that proposed EU age verification regulations serve as a "Trojan horse" to mandate digital IDs, potentially compromising online anonymity and centralizing user data under the guise of child protection. [src]
Commenters debate whether digital IDs are an inevitable evolution of physical tokens [0][8] or a deliberate tool for state control and surveillance [2][7]. While some argue that these systems already exist for government services and are more efficient than paper [1][5], others fear they will lead to a "social credit system" where individuals can be locked out of society [2][7]. Concerns also focus on the loss of anonymity in daily life, as digital verification could turn casual interactions into trackable data points [3], though some note that pervasive CCTV already provides similar levels of monitoring [9].
11. SWE-bench Verified no longer measures frontier coding capabilities (openai.com)
280 points · 156 comments · by kmdupree
OpenAI has stopped using the SWE-bench Verified benchmark to measure AI coding progress, citing flawed test cases that reject correct solutions and widespread data contamination. The company found that models often succeed by recalling training data rather than demonstrating real-world software engineering capabilities. [src]
The SWE-bench Verified benchmark has reached saturation at 93.9%, leading to claims that it no longer effectively measures frontier coding capabilities [1]. Critics argue that benchmarks are frequently gamed for marketing or contaminated by training data, with some auditing revealing that a significant portion of "failed" tasks actually contained flawed test cases that rejected correct code [0][3][9]. To address these limitations, researchers are shifting toward more complex, reasoning-heavy evaluations like ARC-AGI or "Zork bench," which test world models and goal-oriented logic rather than just test-passing ability [2][4][5].
12. Statecharts: hierarchical state machines (statecharts.dev)
295 points · 79 comments · by sph
Statecharts are an advanced visual formalism for modeling complex systems that extend traditional state machines by adding hierarchy and parallelism to prevent state explosion. They offer benefits like decoupled behavior, improved testability, and lower bug counts, though they require learning a new paradigm and specialized libraries. [src]
The discussion highlights statecharts as a powerful tool for modeling complex UI flows and executable behavior, though some users note they have struggled to gain mainstream traction in the frontend ecosystem [0][1]. While proponents emphasize their value in reasoning about state-dependent events, skeptics argue that real-world complexities like multithreading and external dependencies can turn statecharts into "ugly messes" [0][6]. Participants also debated terminology and suggested various resources, ranging from the original 1987 Harel paper to modern libraries like XState and Robot3 [0][1][8].
13. Issue links now open in a popup (github.com)
234 points · 124 comments · by luckman212
GitHub has reverted a change that caused issue links to open in popup overlays after users complained the feature broke standard navigation and productivity. [src]
GitHub recently reverted a performance-driven change that opened issue links in popups after users complained about deteriorating UX [1]. While the change was intended to bypass slow page loads caused by complex header rendering and React/Rails integration [7][9], commenters expressed frustration that major tech companies often struggle with basic usability [0][3]. Some users suggested that browser-level features, such as split-tab views, could better solve these navigation issues than site-specific UI changes [2][5][8].
14. Clay PCB Tutorial (feministhackerspaces.cargo.site)
214 points · 127 comments · by j0r0b0
Researchers from Mz* Baltazar’s Lab have developed a tutorial for creating sustainable, "ethical" printed circuit boards using locally sourced wild clay and conductive silver paint made from recycled jewelry waste. [src]
The project’s self-classification as "feminist hacking" sparked significant debate, with some users questioning if the label implies a binary where "professional" work is masculine [0] and others defending it as a valid effort by a new generation to redefine technical language and identity [2]. Technical critiques focused on the environmental impact of using open wood fires versus electric kilns or 3D printing [1][5][9], as well as the practicality of clay compared to existing industrial ceramic PCBs or traditional wire-wrapping [3][4][7]. Additionally, some commenters expressed frustration over the use of public research funds for an artistic project they felt hobbyists typically perform independently [8].
15. Fast16: High-precision software sabotage 5 years before Stuxnet (sentinelone.com)
210 points · 47 comments · by dd23
SentinelLABS uncovered "fast16," a 2005 cyber sabotage framework that predates Stuxnet by five years. The malware uses a Lua-powered carrier and a kernel driver to selectively patch high-precision engineering software in memory, subtly corrupting floating-point calculations used in critical research and industrial simulations. [src]
The discovery of Fast16 highlights the use of archaic SCCS/RCS notation in 2005 Windows kernel code, which some interpret as a "breadcrumb" pointing toward developers with decades of experience in government or military environments [0][5]. While some argue this suggests specialized recruitment from scientific or industrial fields [6], others contend that RCS was still a common tool for Unix systems people during that era and not necessarily an indicator of ancient origins [3][7]. The discussion also touches on the ethics of sabotaging scientific research [2] and the pragmatic "make do" philosophy of maintaining legacy codebases rather than rewriting them [0].
16. Butterflies are in decline across North America, a look at the Western Monarch (smithsonianmag.com)
183 points · 55 comments · by 1659447091
We couldn't summarize this story. [src]
The decline of butterflies is largely attributed to the widespread use of pesticides and mosquito spraying services, which users report seeing kill monarchs and birds in real-time [0][1][7]. Commenters advocate for replacing traditional lawns with native sedges or clover, though some note the difficulty of finding native seeds that can withstand lawn-like conditions [2][4][9]. To better support the monarch lifecycle, participants suggest avoiding "trap" plants like butterfly bushes in favor of milkweed and using targeted mosquito dunks rather than broad-spectrum sprays [3][5][8].
17. GnuPG – post-quantum crypto landing in mainline (lists.gnupg.org)
166 points · 56 comments · by zdkaster
GnuPG 2.5.19 has been released, introducing post-quantum cryptography support via the Kyber (ML-KEM) encryption algorithm alongside 64-bit Windows improvements and various bug fixes. [src]
The introduction of post-quantum cryptography (PQC) in GnuPG has highlighted a deep schism in the OpenPGP community, where GnuPG and the IETF (RFC-9580) are promoting incompatible standards due to disagreements over modernization versus backward compatibility [0][3][8]. While ML-KEM (Kyber) offers fast performance, its larger key sizes create higher overhead compared to traditional methods, leading many to favor hybrid approaches that combine PQC with established algorithms like X25519 [1][2]. Despite these technical advancements, observers note that PQC cannot protect historical data already harvested by actors waiting for future decryption capabilities [9], and the transition remains a non-trivial logistical challenge akin to Y2K [6].
18. Self-updating screenshots (interblah.net)
192 points · 29 comments · by bjhess
James Adam developed an automated system for the Jelly help center that uses headless Chrome to capture and update documentation screenshots directly from the running application whenever the UI changes. [src]
The discussion highlights a strong consensus that automating screenshots through headless rendering or CLI commands significantly reduces maintenance overhead and ensures documentation stays current [0][6][7]. While some users argue that manual updates can lead to "overthinking" and scope creep [4], others note that programmatic generation allows for advanced features like automatic light/dark mode toggles in READMEs [6]. Alternative approaches mentioned include using live-rendered previews for better accessibility [5] or embedding offscreen GUI renderings into safety-critical applications via shared memory [1].
19. Dillo Browser Release 3.3.0 (dillo-browser.org)
167 points · 26 comments · by rodarima
Dillo 3.3.0 has been released, introducing a new `dilloc` tool for command-line control, customizable "page actions" for the right-click menu, and experimental support for FLTK 1.4, alongside fixes for OAuth logins and improved cookie handling. [src]
The release of Dillo 3.3.0 sparked discussion on the increasing difficulty of maintaining small browsers as major sites like Google now require JavaScript, which critics view as an attack on the open web [0][8]. Users noted that non-mainstream browsers often trigger "429 Too Many Requests" errors on sites like Hacker News, a problem likely caused by anti-crawler measures that can sometimes be bypassed by mimicking Chrome's headers [1][3][9]. While some find the browser's name amusing [4], others highlighted that lightweight alternatives like Startpage or DuckDuckGo’s HTML version still allow for a functional no-JS experience [6][7].
Brought to you by ALCAZAR. Protect what matters.