← Oct 5Tuesday, October 6, 2026 · all daysOct 7 → · today »
1.Beam: Reflection's 501B open-weight model(reflection.ai)
549 points by Philpax 8 days ago | 174 comments | permalink
tl;dr: Reflection is releasing Beam, a 501B-parameter (23B active) open-weight Mixture-of-Experts model targeted at coding, reasoning, and agentic tasks, pretrained on 23.8T tokens and trained with large-scale RL (100M+ rollouts on 10.5K GB300 GPUs over 4 weeks). It claims competitive performance with larger open models like GLM 5.2 while using 3–4× less inference compute, with weights to be released under Apache 2.0 later this month.
HN Discussion:
  • ↓Model is bigger yet underperforms smaller Chinese open-weight competitors like DeepSeek
  • ~Welcomes Western entry into open-weight race despite lagging Chinese labs
  • •Questions who Reflection is as a company, emphasizing organizational trust over benchmarks
  • ↓Announcing without releasing weights is a marketing miss that loses mindshare
  • ↑Insider enthusiasm and technical detail-sharing about the training process
2.Find the flattest route between any two points in SF(flattensf.com)
308 points by ishan0102 8 days ago | 114 comments | permalink
tl;dr: Flatten SF is a browser-based tool that computes walking and biking routes between any two points in San Francisco, letting users trade off distance against elevation gain via a slider. It uses USGS 1-meter lidar elevation data and the Overture/OpenStreetMap street graph across ~160,000 segments, with stairways permitted for pedestrians but excluded for cyclists. Place search and all routing run offline in the browser.
HN Discussion:
  • ~Recommends alternative existing tools that offer similar or better elevation-aware routing
  • ↑Suggests feature improvements like minimizing grade rather than total elevation gain
  • ↓Reports accuracy issues and routing bugs with specific street examples
  • ↑Appreciates the tool with nostalgia and enthusiasm for SF-specific routing
  • ↓Criticizes technical choices like missing OSM attribution and confusing UI decisions
3.Opus 5.5 agents discover two room-temperature magnetic semiconductor candidates(vals.ai)
488 points by outlier99 8 days ago | 338 comments | permalink
tl;dr: Claude Opus 5.5 agents, running DFT simulations, identified two candidate Luttinger-compensated magnets—semiconductors with zero net spin but spin-sorted electron bands useful for spintronics. One, YBaMnFeO₅, was newly designed but likely hard to synthesize due to atomic disorder; the other, KV[Cr(CN)₆], was actually first made in 1999 as a Prussian blue analog and remains magnetically ordered up to 376 K, but its spin-sorting properties were never recognized. The author suggests remaking and measuring KV[Cr(CN)₆] directly, and has published all computations on GitHub.
HN Discussion:
  • ↓Title is misleadingly editorialized to evoke superconductors when it's about antiferromagnetic semiconductors
  • ↓Skepticism about AI 'discovery' claims without experimental validation, especially after LK-99
  • ↓The intro's framing of common magnet types is technically inaccurate
  • ↑AI agents are well-suited to searching parametric scientific spaces faster than humans, so such findings will increase
  • ~These 'discoveries' are really AI surfacing overlooked prior human work, raising questions about long-term sustainability
4.Web Search API(developers.cloudflare.com)
590 points by tosh 8 days ago | 284 comments | permalink
tl;dr: Cloudflare launched Web Search API in beta, letting AI agents query live internet data through three providers: Ceramic.ai, Exa, and Linkup, all supporting Zero Data Retention. Requests route through AI Gateway for logging and are billed at each provider's list price with no markup, with the option to bring your own API key. It's accessible via REST API or the Workers AI binding.
HN Discussion:
  • ~Terms of service restrictions on storing/resyndicating search results limit usefulness for agent systems
  • ↓Cloudflare is hypocritical for blocking bots then selling bot-friendly search access
  • ↓Other search API alternatives (Gemini, Serper, Kagi) offer better pricing or value
  • ↓Cloudflare acts as unnecessary middleman between developers and providers
  • ↓Inconsistency between claimed Zero Data Retention and provider documentation
5.Example.com just launched the biggest redesign in decades(debugbear.com)
356 points by jgx0 8 days ago | 247 comments | permalink
tl;dr: IANA updated example.com in September 2026 to rotate through six languages via JavaScript with a per-character opacity fade, then days later reverted to showing all languages statically. The article traces the site's history since 2002, covering copy tweaks, infrastructure moves (Apache → EdgeCast → Cloudflare), and the recent addition of a warning telling developers not to use the domain for testing or monitoring.
HN Discussion:
  • ↑Offers a workaround/alternative service for those whose tests broke due to the redesign
  • ↑Acknowledges the change broke many tests, invoking Hyrum's Law despite the disclaimer
  • ↑Confirms and updates the article's observation that the animation was reverted
  • ~Laments personal/quirky use cases (like cleaning glasses) being broken by the redesign
  • ~Critiques minor aspects of the article/site like missing links, HTML compression, or awkward conclusion
6.Competitive Programmer's Handbook (2018) [pdf](cses.fi)
312 points by vinhnx 10 days ago | 82 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • ↑Praise for the author as a teacher and recommendation of his newer book
  • ↑The book was personally useful for interview prep and learning CS fundamentals
  • ↑Competitive programming is a joyful way to recover thinking skills lost to LLM overuse
  • ~The book sacrifices readability for cleverness; alternatives are more approachable
  • ↓Competitive programming's value has diminished in the age of AI coding tools
7.ChatGPT is adding real cartoonists' signatures to fake New Yorker cartoons(niemanlab.org)
545 points by rdmuser 8 days ago | 408 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • •Technical explanation of why the model does this due to training data patterns
  • ↑Confirms the problem from personal experience with AI image generation
  • ↑LLM vendors should face legal liability like a human would for forging signatures
  • ↑This proves LLMs cannot truly reason and debunks AI promoter claims
  • ~Not an OpenAI-specific issue; this has existed across image generation models for years
8.The lamps in my house(arslan.io)
335 points by farslan 8 days ago | 126 comments | permalink
tl;dr: A designer tours ~15 iconic lamps in his home, including Artemide's Tolomeo and Tizio, multiple Isamu Noguchi Akari paper lanterns, Louis Poulsen's PH 5 and PH 3/2, Arne Jacobsen's AJ wall lamp, Jean Prouvé's Lampe de Bureau, and newer pieces from HAY and Marset. He details each lamp's designer, history, and personal use, acknowledging they're expensive but worth it as collectibles, and closes with his own 3D-printed Block Lamp design from 2025.
HN Discussion:
  • ~Recommends alternative affordable lamps outside the author's curated collection
  • ↑Reinforces the article's premise that eye-level lamps are superior lighting design
  • •Shares linguistic/cultural tangent sparked by the article's Akari note
  • ↑Praises the author's collection and suggests additional iconic designer lamps to explore
  • ↓Criticizes specific lamps the author celebrates (Tizio) as poorly engineered
9.Apple and a hacker's future(stratechery.com)
292 points by maguay 8 days ago | 245 comments | permalink
tl;dr: Ben Thompson's always-on Mac Mini running Claude and Codex agents was compromised via a macOS screen-sharing vulnerability (CVE-2026-65400), though his agent actually detected and helped remediate the intrusion. He argues Apple's increasingly restrictive permission system (TCC, Full Disk Access) is hostile to agent-driven workflows, misaligned at the wrong level of abstraction, and signals a broader tension: as AI agents turn everyone into a hacker, Apple's walled garden feels more like a prison, making him consider non-Apple hardware for the first time.
HN Discussion:
  • ↓Apple's permission system exists for good reason; full-disk access shouldn't be given to agents or risky software
  • ↓The author showed poor security practices by exposing screen sharing to the internet, undermining his critique
  • ↓AI agent users accept reckless security tradeoffs for productivity, which isn't a sustainable approach
  • ↑Ben's shift signals a real risk for Apple as agentic workflows become the future and users may abandon the walled garden
  • ↑Apple's vague communication and poor security update handling supports criticism of their approach
10.Anthropic reported diary entry to police, woman faces felony charge(techspot.com)
825 points by emptybits 9 days ago | 674 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • ↓Anthropic faced a damned-if-you-do situation; reporting was reasonable given liability concerns
  • ↑Users should use local/open-source models to escape surveillance by Big Tech AI providers
  • ↑The Florida statute doesn't apply since a private diary entry isn't a transmitted threat to another party
  • ↑Prosecuting someone based on surveilled private writing is unjust and should be thrown out
  • ↑Anthropic is hypocritical—committing copyright felonies while enabling thought-crime prosecution of ordinary users
11.The technology to eradicate mosquito-borne disease exists(worksinprogress.co)
276 points by benbreen 9 days ago | 237 comments | permalink
tl;dr: Genetically engineered mosquitoes (Oxitec's self-limiting Aedes aegypti) and Wolbachia-infected mosquitoes can effectively collapse disease-carrying mosquito populations, with field trials showing 80-95% reductions and dengue infections cut by ~70%. However, Oxitec's technology has languished in US regulatory limbo for 15 years, bounced between the USDA, FDA, and EPA because the 1986 Coordinated Framework wasn't designed for modern biotech. Reforms like clearer jurisdiction, industry-funded reviews (as with the FDA's PDUFA), and expanded EPA capacity could unlock the ability to eliminate mosquito-borne diseases domestically.
HN Discussion:
  • ↑Strong support for eradication given massive death toll and suffering caused by mosquitoes
  • ↑Personal experience with mosquito-borne disease validates urgency of deployment
  • ↑Parallels to other diseases where solutions exist but human/policy choices block progress
  • •Technical curiosity or confusion about how the eradication method actually works
  • ↑Firsthand confirmation that mosquito-free environments are achievable and pleasant
12.Denmark data breach exposes 8.8M people's personal data(cpr.dk)
501 points by clan 9 days ago | 350 comments | permalink
tl;dr: Denmark's Central Person Register (CPR) suffered a breach in which attackers abused a Danish company's legitimate access to the system to obtain names, addresses, and CPR numbers of roughly 8.8 million registered citizens. Individuals with name/address protection were not exposed. Authorities have revoked the company's access, notified the Data Protection Agency, and launched a police investigation.
HN Discussion:
  • ↑General fatigue and distrust toward institutions handling personal data due to repeated breaches
  • ↑Breach highlights hypocrisy of Denmark pushing Chat Control while failing at basic security
  • ~Scope of breach is actually worse than reported, affecting essentially all citizens plus sensitive data
  • ↑Danish public sector has long neglected IT security and digitalization competence
  • ~Silver lining: breach may force reform of using CPR numbers as secret identifiers
13.We ported the original Doom to SQL(cedardb.com)
328 points by Vaslo 10 days ago | 55 comments | permalink
tl;dr: Developers at CedarDB ported the original 1993 Doom's game logic and renderer entirely to SQL (~5900 lines for logic, ~1300 for the renderer across 89 CTEs), with Python only handling input, timing, and bitmap display. The implementation hits the original 35 Hz tick rate and renders up to 60 FPS by exploiting BSP traversal via bit-packed sort keys, window functions for visplanes, and brute-force depth resolution via min() on packed bigints. Multiplayer deathmatch comes nearly free from the database's built-in authentication, transactions, and access control, with a public playable instance available.
HN Discussion:
  • ↑SQL's capabilities are underappreciated and can express complex business logic effectively
  • ↑Appreciation of the project as humorous engineering absurdity or malpractice done well
  • •Reminded of personal experience using SQL as source of truth for game state, with mixed outcomes
  • ↑Observation that SQL has functional programming qualities
  • •Questions about technical portability to other database engines beyond CedarDB
14.In the wake of Tippett Studios’ closure, a digital archive appears online(filmstories.co.uk)
248 points by rdmuser 9 days ago | 33 comments | permalink
tl;dr: After Tippett Studios closed in August and its archives were auctioned off, an anonymous buyer called "TippettFan" acquired a folder of CD-ROMs, digitized them, and uploaded the contents to the Internet Archive. The collection contains 90 .ISO disc images with behind-the-scenes material from RoboCop, Starship Troopers, the original Star Wars, and other films, along with test clips of CGI creatures and a 2008 Phil Tippett interview.
HN Discussion:
  • ↑Personal industry anecdotes from VFX veterans reflecting on Tippett's legacy and work culture
  • ↑Awe and gratitude that one anonymous person rescued this history from being lost
  • ↑Urgency to mirror/seed the archive before copyright holders can take it down
  • ↑Sadness about the bankruptcy closure and material that remains unpreserved
  • ~Meta-suggestion that the title should name Tippett prominently to attract the right audience
15.A browser-native classic Visual Basic VB6 IDE(wieslawsoltes.github.io)
414 points by wiso 9 days ago | 135 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • ↑Modern tooling lacks VB6's discoverability and RAD-style developer-friendliness
  • ~Functionally impressive but visual fidelity to original VB6 is lacking
  • ↑Nostalgic appreciation for VB6 and the memories it evokes
  • •AI coding agents could revive the VB6 RAD experience for modern development
  • ↑Admiration for the author's prolific open-source output across many projects
16.Mold Linker Version 3.0.0 Release – Rewritten in Rust(github.com)
245 points by roflcopter69 8 days ago | 193 comments | permalink
tl;dr: Mold 3.0.0 is the first release of the high-speed linker rewritten from C++ to Rust, intended as a drop-in replacement for 2.42.1 with equivalent performance, same CLI options, and same target support. The rewrite adds bounds-checking safety against corrupted inputs, drops the oneTBB dependency, and switches the build system from CMake to Cargo (requiring Rust 1.95+). The release also bundles numerous bug fixes, improved GNU ld compatibility (especially in linker scripts), and architecture-specific relocation fixes, aiming to make mold viable as a default linker in Linux distributions.
HN Discussion:
  • ↓Rust rewrite forces distro to fork and maintain the C++ version, criticizing Rust as unsuitable here
  • •Surprise at the speed of the rewrite, speculating about AI agents being involved
  • •Curiosity about technical differentiation between mold and the Wild linker now that both are Rust
  • ↑Enthusiastic excitement about the release, expecting continued or improved performance
17.Germany’s RobCo hits $1B valuation(techfundingnews.com)
344 points by dachworker 8 days ago | 377 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • ↓Skepticism toward the Robotics-as-a-Service business model's viability
  • ↑German industrial/engineering expertise is valuable and underappreciated
  • ↓Questioning the high valuation and source of investment capital
  • ~Practical concerns about service contracts, downtime, and operational details over valuation hype
  • •Curiosity about the state of AI/robotics progress amid LLM focus
18.Pixel 11 doesn't yet meet the GrapheneOS security standards and may be skipped(discuss.grapheneos.org)
428 points by finnlab 8 days ago | 303 comments | permalink
tl;dr: Summary not available
HN Discussion:
  • ↓Article is outdated; GrapheneOS already walked back the statement and MTE may yet be enabled
  • ~Bigger concern is Google restricting OEMs from shipping GrapheneOS devices
  • ↑Pixel 11 makes too many cost-cutting compromises, skipping it is reasonable
  • ↑GrapheneOS should move away from Pixel and target other OEMs like Motorola
  • ↓Speculating about Google's internal motives for cutting MTE is unprofessional communication
19.Tell HN: Bob Cringely has died
938 points by paveworld 10 days ago | 218 comments | permalink
tl;dr: Bob Cringely (real name Mark Stevens), an early Apple employee and tech journalist best known for the PBS documentary "Triumph of the Nerds," died in his sleep early Saturday. The news was shared on Hacker News by a friend of the family.
HN Discussion:
  • ↑Fans fondly remember Triumph of the Nerds and his other documentaries as formative tech viewing
  • ↑Readers appreciated his writing, especially Accidental Empires and his blog
  • ↑Sympathy for the hardships he endured in his final years
  • ~Critical note that he scammed people and fabricated stories despite entertaining work
  • ~His Plane Crazy documentary revealed hubris and skepticism toward modern composites
20.Reasons I didn't become an EMT, ranked(ben.stolovitz.com)
251 points by citelao 10 days ago | 121 comments | permalink
tl;dr: A software engineer recounts the many excuses he invented over years to avoid getting his EMT certification—ranging from legitimate concerns about EMS culture and cost to increasingly flimsy rationalizations—before finally taking a 28-day Wilderness EMT course in 2024. His core takeaway: the real reason for the delay was fear of doing something scary and new, and most "reasons not to do it" were just excuses. He now volunteers on an ambulance while keeping his tech job, and encourages others stuck in analysis paralysis to just try the thing.
HN Discussion:
  • ↑Shares personal story of transitioning from tech/engineering to EMT work, reinforcing the article's encouragement
  • •Suggests alternative paths like remote/wilderness EMT or online courses to explore the field
  • ~EMT work can take a serious psychological toll, hinting at darker realities the article glosses over
  • ↑Volunteering as EMT/firefighter has been personally rewarding and helpful for mental health
  • ~EMT work is only feasible due to financial security from prior tech career; system treats workers poorly