Hacker News
1Norway imposes near ban on AI in elementary school
Age-appropriate AI restrictions
Discussion centers on Norway's policy restricting AI use for ages 6-13, with broad support for the approach. Users argue young students need to develop fundamental reading, writing, and critical thinking skills before using AI tools that could shortcut learning.AI's impact on education quality
Teachers report widespread problems with AI in classrooms, from students using it to avoid learning to educators creating flawed AI-generated assignments. Many advocate for complete AI bans in educational settings due to negative outcomes.Technology restrictions in schools
Broader discussion of removing digital distractions from classrooms, including smartphone bans and returning to traditional teaching methods like pen-and-paper. Users debate effectiveness of these measures for improving learning outcomes.
2Google workspace threatening to block Firefox access
- Google Workspace is warning Firefox users they must use Chrome to access their services, though Google support claims this only applies to admin access and is just a recommendation, not a block. The user is concerned about being forced to switch browsers despite having a Business Plus account with no enterprise security features configured that would justify Chrome-only requirements.
Corporate IT policy vs user choice
Discussion centers on whether organizations should force Chrome usage through Google Workspace's Context-Aware Access features. IT professionals defend the practice citing security benefits and management capabilities, while others argue it's monopolistic and shouldn't be an option.Browser detection vs feature detection
Technical debate about websites checking for specific browsers rather than testing for features. Commenters argue this approach is flawed and advocate for feature detection over user-agent sniffing to avoid forcing browser compatibility workarounds.Google's security requirements justification
Users question whether the Chrome requirement is truly security-based, discussing features like Device-Bound Session Credentials (DBSC) and endpoint verification that may only work with Chrome, though the specific security policies aren't clearly documented.
3There are no instances in ATProto
- The post explains why people asking "where are the Bluesky instances?" are asking the wrong question. Unlike Mastodon which uses federated instances (separate hosting+app bundles that talk to each other), atproto separates hosting from apps entirely, like RSS and Google Reader. Users can swap hosting providers and try different apps independently, making "instances" an irrelevant concept in atproto's architecture.
Architectural differences between ATProto and ActivityPub
Discussion centers on whether ATProto truly solves decentralization differently than ActivityPub's "instances" model. Critics argue the author misrepresents ActivityPub to favor ATProto, while supporters claim ATProto's separation of data hosting, apps, and relays offers better scalability and user control than Mastodon's coupled architecture.Centralization concerns in practice
Users debate whether ATProto is actually decentralized given Bluesky corporation controls most user data, apps, and moderation. While the protocol enables decentralization theoretically, critics note the practical reality resembles centralization more than Mastodon's distributed instances.Google Reader analogy problems
The comparison to RSS/Google Reader is criticized as ominous since Google Reader's shutdown damaged communities despite RSS surviving. Critics argue this analogy doesn't inspire confidence in ATProto's resilience if Bluesky disappears, unlike Mastodon where individual instance shutdowns have less network-wide impact.
4GPT-5.5 hallucinates 3x more than MIT-licensed GLM-5.2
- Major AI labs are shifting away from endless scaling as larger models plateau in intelligence and show worse hallucination rates. DeepSeek V4 Pro with 1.6T parameters had 94% hallucination vs GLM-5.2's 28%, proving bigger isn't always better for real-world accuracy.
Model scaling limitations
Discussion centers on whether bigger models have plateaued in intelligence gains, with debate over claims that larger models don't improve performance proportionally and may actually increase hallucination rates when they don't know answers.Hallucination measurement and interpretation
Users debate how hallucination rates should be calculated and interpreted, noting that current metrics are conditional on models not knowing answers and don't reflect real-world usage patterns or absolute error rates.Training approaches to reduce hallucinations
Technical discussion on whether RLVR training can teach models to say "I don't know" instead of hallucinating, with mixed results from attempts and challenges in creating proper reward functions for uncertainty.
5Court Records Should Be Free
- The Electronic Frontier Foundation and other organizations support the Open Courts Act of 2026, which would replace the costly PACER system with a modern, free platform for accessing federal court records. PACER currently charges hefty fees that create barriers to public access of court documents, collecting over $150 million annually despite these being public records. The new legislation aims to eliminate these fees and improve public access to court filings.
Public access vs cost burden
Complex debate over who should pay for "free" PACER access. Arguments that making it free is regressive since lawyers (well-compensated professionals) are primary users, funded by general taxpayers. Counterarguments focus on public interest value for journalism, transparency, and equal access to law.Current workarounds and alternatives
Discussion of existing solutions like RECAP program and CourtListener that redistribute purchased PACER documents for free public access. Users share strategies for maximizing free quarterly limits and contributing to these liberation efforts.Fundamental principle of public law access
Core argument that court decisions constitute law itself, and citizens required to follow precedent should have free access to read it. Comparison to Hammurabi's Code and emphasis that charging for legal precedent creates barriers to justice.
6How many of the 170k English words do you know?
User experience issues
Users complained about too many clicks per word (requiring 3 clicks instead of 1), poor mobile layout with submit buttons below the fold, lack of keyboard shortcuts, and overall clunky interface that made completing 100 words tedious and broke the flow of the quiz.Word selection and difficulty problems
Many criticized poor word stratification with easy words like "kinetic" labeled as expert while obscure joke words like "Hippopotomonstrosesquippedaliophobia" appeared. The test spent too much time on basic vocabulary before jumping to extremely obscure terms.Multiple choice format flaws
The quiz was easily gamed by always picking the longest answer option, recognizing patterns like correct/opposite answer pairs, and eliminating obviously wrong choices. Many wanted an "I don't know" option instead of forced guessing that inflated scores.
7Bobby Prince, composer for Doom, Wolfenstein 3D, and Duke Nukem 3D, has died
Technical details of Doom's sound
Bobby Prince explained he composed Doom's soundtrack on a Sound Blaster 1.0 with Yamaha OPL2 FM synth chip. He created custom drum sounds by playing instruments out of normal range, which later sound cards couldn't properly emulate. DOSBox provides the best modern emulation.Musical influences and references
Doom's music heavily referenced classic metal bands like Pantera, Slayer, and Judas Priest. Many players discovered these influences later, with some preferring the OPL2 chip versions over orchestral replacements that made games sound too much like movies.Tributes and lasting impact
Players shared memories of Prince's memorable compositions for Doom, Duke Nukem 3D, Wolfenstein 3D, and Commander Keen. His work was recently added to the Library of Congress registry, cementing his legacy as a legendary video game composer.
8CSSQuake
- cssQuake is a web-based recreation of the classic Quake game built using CSS and web technologies. The game features single player and multiplayer modes with traditional FPS controls including WASD movement, mouse look, clicking to fire, and spacebar to jump. Players can customize various gameplay options like crosshair, dynamic lighting, sound, particles, and enemy visibility, along with debug features for recording and capturing gameplay elements.
Performance comparison
Users noted that despite 30 years of hardware advancement, CSS Quake runs slower on modern M1 Macs than original Quake ran on 90s Pentium systems. Performance varies by browser - Chrome runs smoothly while Safari/WebKit has issues.Technical implementation
Discussion about how the project uses CSS for rendering while game logic is implemented in TypeScript/JavaScript. Users noted bugs with button interactions and questioned whether it's a full engine recreation or just a renderer.User experience issues
Users reported difficulty exiting the game (comparing it to vim), view clipping bugs, floating dead enemies, and cursor drift problems affecting gameplay across different browsers and platforms.
9AI Engineer Claims to Have Cracked Linear A
- Tom Di Mino, a self-taught AI engineer, claims to have deciphered Linear A, a Bronze Age Minoan script that has puzzled linguists for over a century. He believes it represents an extinct Semitic language related to biblical Hebrew. His breakthrough came in May 2024 while analyzing prayer inscriptions, identifying the verb "nawaya" meaning "to dwell." Using AI tools, he has proposed readings for 40 script signs and created a lexicon of 408 terms. His work is currently under review by experts at Rutgers and Cambridge universities.
Skepticism about verification
Commenters question the plausibility of the claim, noting Linear A's extremely small corpus (~7,500 characters) makes verification difficult. Many point out that with such fragmentary material, almost any translation could seem plausible, and proper peer review is needed.Technical methodology concerns
Discussion centers on the author's approach of matching Linear A symbols to Semitic roots, with critics demonstrating how this method could falsely "prove" English is Semitic through selective word matching and insufficient documentation of the full methodology.Semitic language hypothesis issues
Debate over whether Linear A could represent a Semitic language, given that such languages typically use consonantal writing systems rather than syllabaries with vowels, and the lack of expected Semitic loanwords in ancient Greek.
10Where to Find the Colors Your Screen Can't Show You
- The digital world lacks many colors that exist in nature, particularly vibrant cyans. Our screens use limited color ranges (like sRGB) that can't reproduce the full spectrum of colors our eyes can see. The most spectacular colors exist in natural environments where light is filtered through multiple layers - like sunlight passing through forest leaves or water with phytoplankton. These natural filters create extraordinarily saturated greens, cyans, and blues that no screen can display, making real-world color experiences far richer than anything digital technology can currently capture.
Color space limitations and display technology
Discussion centers on sRGB's inability to reproduce saturated colors, particularly orange/red/purple hues common in everyday objects. Users debate Display P3 vs wider gamuts, projector limitations, and the importance of 10-bit color depth for accurate reproduction.Real-world color perception vs digital reproduction
Artists and enthusiasts share experiences of how screens fail to capture vivid colors seen in nature, paintings, and specialized lighting like blue lasers. Discussion includes challenges in photography, printing, and the gap between human vision and display technology.Technical aspects of human color vision
Users explore cone cell stimulation, metamerism (different spectra appearing as same color), color space mathematics, and advanced topics like tetrachromacy and impossible colors. Includes discussion of CIE color models and their limitations.