The Who Cares Era

691 NotInOurNames 2025-05-28 13:07:57 UTC dansinker.com Llama Share 712

The Chicago Sun-Times and Philadelphia Inquirer published fake news supplements created by an AI chatbot, highlighting the "Who Cares Era" where mediocrity is accepted. This era is fueled by AI, which creates "good enough" content, and is also reflected in the government's and corporations' lack of care for quality.

Show HN: I rewrote my Mac Electron app in Rust

583 katrinarodri 2025-05-28 16:53:11 UTC desktopdocs.com Llama Share 438

Desktop Docs is a desktop search software that uses AI to analyze images and videos based on content, colors, objects, and composition. It allows users to search for specific scenes, images, or videos without uploading data to the cloud.

My website is ugly because I made it

537 surprisetalk 2025-05-28 11:53:47 UTC goodinternetmagazine.com Llama Share 127

The writer values individuality and self-expression, creating a unique website that reflects their personality and passions. They prioritize creativity and experimentation over simplicity and convention.

As a developer, my most important tools are a pen and a notebook

455 ingve 2025-05-28 06:27:11 UTC hamatti.org Llama Share 285

The author joined a new job and bought a new notebook to aid in problem-solving and thinking. They use writing and sketching to clarify ideas, expose knowledge gaps, and improve code quality.

Japan Post launches 'digital address' system

454 jmsflknr 2025-05-28 16:33:49 UTC japantimes.co.jp Llama Share 289

Japan Post launched a digital address system with 7-digit codes linking to physical addresses. Users can register for a Yu ID to obtain a digital address that remains unchanged even with physical address updates.

Deepseek R1-0528

430 error404x 2025-05-28 17:59:02 UTC huggingface.co Llama Share 249

DeepSeek R1 model version 0528 has improved depth of reasoning and inference capabilities with increased computational resources and algorithmic optimization. It now approaches leading models in performance, with significant improvements in complex reasoning tasks and reduced hallucination rate.

Show HN: AutoThink – Boosts local LLM performance with adaptive reasoning

394 codelion 2025-05-28 02:39:11 UTC news.ycombinator.com Llama Share 64

You developed AutoThink, a technique that allocates resources based on query complexity, improving efficiency by 43% in GPQA-Diamond. It uses adaptive classification and steering vectors to guide local LLMs.

Long live American Science and Surplus

390 thinkalone 2025-05-28 20:47:30 UTC milwaukeerecord.com Llama Share 111

The website is temporarily blocked due to security reasons after a suspicious action was detected. Please email the site owner with the Cloudflare Ray ID to report the issue.

Mullvad Leta

374 microflash 2025-05-28 14:38:52 UTC leta.mullvad.net Full Share 185

Leta, the privacy-focused search engine, by Mullvad.

Getting a Cease and Desist from Waffle House

361 lafond 2025-05-28 15:48:37 UTC jack.bio Llama Share 187

A university student created a live map tracking Waffle House closures during Hurricane Helene using Next.js and data scraping, which caught the attention of Waffle House's corporate account and a major media figure. The student's project was eventually shut down due to trademark infringement, but not before Waffle House's marketing team engaged in a humorous exchange with the student on Twitter.

Compiler Explorer and the promise of URLs that last forever

349 anarazel 2025-05-28 16:28:20 UTC xania.org Llama Share 179

Compiler Explorer used goo.gl links until Google banned them in 2016, then used a workaround with godbolt.org. The team is now rescuing legacy links before Google's goo.gl service sunsets in August 2025.

The Polymarket users betting on when Jesus will return

316 surprisetalk 2025-05-28 04:13:00 UTC ericneyman.wordpress.com Llama Share 604

People are betting on Jesus Christ's return in 2025, with some betting yes and others no, due to the time value of money on Polymarket. The yes bettors are betting that others will want cash later in the year to bet on other markets, potentially selling their shares at a higher price.

xAI to pay telegram $300M to integrate Grok into the chat app

314 freetonik 2025-05-28 15:12:56 UTC techcrunch.com Llama Share 393

xAI pays Telegram $300 million for distributing Grok chatbot. Telegram users will earn 50% of xAI subscription revenue and gain access to Grok's features.

AI: Accelerated Incompetence

301 stevekrouse 2025-05-28 10:50:23 UTC slater.dev Llama Share 266

Engineers should use LLMs as tools, not crutches, to avoid accelerating incompetence and losing technical skills. LLMs can't replace human engineering, and companies overusing AI will inherit long-term costs and risks.

Compiling a neural net to C for a speedup

292 todsacerdoti 2025-05-28 17:22:47 UTC slightknack.dev Llama Share 88

The user trained a neural network to learn a 3×3 kernel function for Conway's Game of Life and extracted the learned logic circuit to achieve a 1,744× speedup in inference by compiling it to bit-parallel C. The user also replicated a research on Differentiable Logic Gate Networks and implemented a runtime to saturate the compiled circuit, achieving a significant speedup in simulating Conway's ...

A toy RTOS inside Super Mario Bros. using emulator save states

249 notorious_pgb 2025-05-28 20:15:43 UTC prettygoodblog.com Llama Share 54

The user created a multithreaded NES emulator using Lua, demonstrating threads, mutexes, and condition variables in a Super Mario Bros. game. This project helped the user understand threading concepts and build a thread scheduler from scratch in approximately 300 lines of Lua code.

Look Ma, No Bubbles: Designing a Low-Latency Megakernel for Llama-1B

232 ljosifov 2025-05-28 00:01:20 UTC hazyresearch.stanford.edu Llama Share 30

Researchers created a megakernel to run Llama-3.2-1B on modern GPUs, achieving 78% memory bandwidth and outperforming existing systems by 1.5x. This megakernel eliminates kernel boundaries and minimizes memory pipeline bubbles.

Show HN: Icepi Zero – The FPGA Raspberry Pi Zero Equivalent

220 Cyao 2025-05-28 13:31:04 UTC github.com/cheyao Llama Share 51

Icepi Zero is a low-cost FPGA board with a Raspberry Pi form factor and HDM- port for digital video output. It's an open-source, portable solution for FPGA development, prototyping, and learning.

Show HN: Tesseral – Open-Source Auth

193 ucarion 2025-05-28 15:27:42 UTC github.com/tesseral-labs Llama Share 84

Tesseral is a multi-tenant API-first service for cloud-based authentication, supporting various tech stacks and B2B SaaS. It offers features like customizable login pages, user impersonation, and role-based access control.

De-anonymization attacks against the privacy coin XMR

187 DbigCOX 2025-05-28 14:11:47 UTC monero.forex Llama Share 100

Monero's privacy features have resisted attempts at deanonymization by governments and firms. Despite some partial successes, no method has achieved reliable, widespread deanonymization of Monero transactions.

The Blowtorch Theory: A new model for structure formation in the universe

186 surprisetalk 2025-05-28 13:43:38 UTC theeggandtherock.com GoogleT5 Share 170

Julian gough: blowtorch theory can be explored independently of parent theory. he says it's based on same set of assumptions which led to the original failure of prediction. but if we look at the early universe, we see jets driving star formation, not quenching it. this is why we have black holes, and they're a good place to start, argues mr gleeson. "we've got

A visual exploration of vector embeddings

181 pamelafox 2025-05-28 20:21:47 UTC blog.pamelafox.org Llama Share 39

Vector embedding models map inputs into multidimensional spaces, allowing comparison of similarity between inputs using distance metrics like cosine similarity. Techniques like quantization and dimension reduction can be used to reduce vector size and improve search efficiency, but may result in quality degradation for vector search results.

AI video you can watch and interact with, in real-time

168 olivercameron 2025-05-28 18:33:50 UTC experience.odyssey.world Full Share 69

A research preview of AI video you can both watch and interact with in real-time.

Revenge of the Chickenized Reverse-Centaurs

167 GreenWatermelon 2025-05-28 16:55:08 UTC pluralistic.net Llama Share 113

The "centaur" concept describes human-machine collaboration, but in labor exploitation, it's reversed, with AI controlling humans. New tech tools, like Para and Tracking Exposed, aim to empower workers by exposing hidden information and countering algorithmic control.

AtomVM, the Erlang virtual machine for IoT devices

160 ahamez 2025-05-28 13:35:27 UTC atomvm.net Llama Share 47

AtomVM is a lightweight virtual machine for IoT applications, supporting Erlang/Elixir code on tiny micro-controllers. It enables functional programming with actor-based concurrency and interfaces with peripherals and WiFi networking.

HTAP is Dead

156 moonikakiss 2025-05-28 22:22:18 UTC mooncake.dev Llama Share 90

The author reflects on the evolution of databases, from single systems handling both transactions and analytics to the current state of specialized engines. The author concludes that HTAP databases, which aim to combine both workloads, have failed to gain traction.

LLM codegen go brrr – Parallelization with Git worktrees and tmux

155 skeptrune 2025-05-28 15:13:38 UTC skeptrune.com Llama Share 72

The author improved AI coding efficiency by running multiple agents in parallel, achieving a 68% chance of getting a workable solution. They propose a tool called uzi to automate the process, making it easier to manage multiple AI agent worktrees.

What does “Undecidable” mean, anyway

148 BerislavLopac 2025-05-28 19:37:47 UTC buttondown.com Llama Share 182

You'll be speaking at Systems Distributed next month about formal methods mental models in mainstream software development. Your upcoming Logic for Programmers releases will be minor changes due to added stress.

Show HN: Every problem and solution in Beyond Cracking the Coding Interview

146 leeny 2025-05-28 17:06:13 UTC news.ycombinator.com Llama Share 105

Aline from interviewing.io offers free problems and solutions from Beyond Cracking the Coding Interview, a sequel to CTCI, with over 230 problems. The problems can be solved with AI Interviewer or read through on the website.

Cory Doctorow on how we lost the internet

146 howtofly 2025-05-28 08:07:42 UTC lwn.net GoogleT5 Share 126

Cory doctorow gave a rousing keynote on the state of the "enshitternet'. he says the internet has been enslaved by companies that lock users into their services - 'enshitting' the process has led to the creation of monopolies, says dr. douglas scott jr, author of "the twiddling"

Why are 2025/05/28 and 2025-05-28 different days in JavaScript?

142 brandon_bot 2025-05-28 07:09:31 UTC brandondong.github.io Llama Share 151

JavaScript Date objects represent time in milliseconds since epoch, with parsing behavior varying across browsers due to inconsistent time zone interpretation. Temporal, a new date and time API, aims to replace Date and resolve parsing ambiguity by requiring explicit time zone specification.

Atomics and Concurrency

140 LAC-Tech 2025-05-28 08:00:48 UTC redixhumayun.github.io Llama Share 62

The user explained how memory ordering works and how to use atomics in C++ to build a lock-free queue, noting that the compiler and CPU can reorder instructions unless a memory ordering guarantee is specified. The user demonstrated how to use different memory ordering models, including sequential consistency, release-acquire, and relaxed, to synchronize threads and prevent data races.

High-quality OLED displays now enabling integrated thin and multichannel audio

126 LorenDB 2025-05-28 01:53:43 UTC sciencedaily.com Llama Share 78

A POSTECH research team developed the world's first Pixel-Based Local Sound OLED technology, enabling each pixel to emit different sounds. The technology was successfully demonstrated on a 13-inch OLED panel, allowing for immersive sound directly from the screen without external speakers.

Anthropic launches a voice mode for Claude

119 kordlessagain 2025-05-28 14:41:47 UTC techcrunch.com Llama Share 31

Anthropic's Claude mobile app now has a voice mode in beta, allowing users to have spoken conversations with Claude. The feature is powered by Claude Sonnet 4 model and will be available in English over the next few weeks.

CheerpJ 4.1: Java in the browser, now supporting Java 17 (preview)

117 pjmlp 2025-05-28 10:23:56 UTC labs.leaningtech.com Llama Share 49

CheerpJ 4.1 is released with Java 17 support, previewing stable support in CheerpJ 5.0. This release also brings stability to Java 11 support and introduces mobile device support.

Microsoft is starting to open Windows Update up to any third-party app

113 Tomte 2025-05-28 10:48:17 UTC theverge.com Llama Share 169

Microsoft is opening Windows Update to third-party apps, allowing developers to sign up for a private preview of its orchestration platform. This will enable scheduled updates based on user activity and integrate apps into Windows Update notifications.

DWARF as a Shared Reverse Engineering Format

107 matt_d 2025-05-28 05:34:47 UTC lief.re Llama Share 19

A new API in LIEF allows creating DWARF files for reverse-engineered information. DWARF is a compatible format for sharing binary analysis between tools like BinaryNinja, Ghidra, and IDA.

Another way electric cars clean the air: study says brake dust reduced by 83%

103 xbmcuser 2025-05-28 06:40:47 UTC electrek.co Llama Share 177

A new study found that electric vehicles reduce brake dust by 83% and total emissions by 38% compared to gas cars. Shifting drivers to public transport or walking/biking can reduce emissions even more, up to 5 times higher than fleet electrification.

Prohibition and ice cream in the US Navy

90 speckx 2025-05-28 18:35:28 UTC oldsaltblog.com Llama Share 124

The US Navy banned alcohol in 1914 and later the US followed with Prohibition in 1920, but ice cream became popular during this time. The Navy used ice cream as a reward and morale booster during World War II, even operating an ice cream barge in the Pacific.

Webb telescope helps refine Hubble constant

86 pseudolus 2025-05-28 09:49:12 UTC phys.org Llama Share 55

Scientists used the James Webb Space Telescope to refine the Hubble constant, suggesting no conflict with the universe's standard model. The new data brings the value into statistical agreement with recent measurements, resolving a long-standing expansion rate debate.

Ray Tracing in J

82 todsacerdoti 2025-05-28 17:44:43 UTC idle.nprescott.com Llama Share 11

The author wrote a minimal ray tracer in J, a terse language with a unique syntax and array-oriented programming paradigm. They found J to be surprisingly interactive and easy to learn, with a gentle learning curve and fast performance.

How to run cron jobs in Postgres without extra infrastructure

79 Liriel 2025-05-28 10:14:11 UTC wasp.sh Llama Share 30

The user discusses the importance of just-in-time infrastructure provisioning and scheduling in complex systems, but notes that sometimes a lightweight scheduling solution is needed, which can be achieved with PgBoss, a job queue built on PostgreSQL. PgBoss can be used with Wasp, a full-stack framework for React & Node.js, to simplify web app development and provide a simple scheduling feature.

Show HN: Voiden – a free, offline, Git-native API Client

78 kiselitza 2025-05-28 12:48:43 UTC voiden.md Full Share 70

Voiden.md is an offline-first, Git-native API workspace that unifies specs, docs, tests, and mocks using programmable Markdown blocks. Build APIs the way you actually work.

The mysterious Gobi wall uncovered

70 bikenaga 2025-05-28 15:07:57 UTC phys.org Llama Share 20

A new study reveals the Gobi Wall's multifunctional role in medieval Inner Asia, including boundary demarcation, resource management, and imperial control. The wall was constructed during the Xi Xia dynasty and its route was carefully selected based on resource availability.

Theory of Stupidity [pdf]

70 dargscisyhp 2025-05-28 02:25:28 UTC onthewing.org Llama Share 27

Stupidity is a human problem that arises from external circumstances and sociability, making people vulnerable to power and losing their inner independence. To overcome stupidity, an act of liberation is needed, not instruction, and it's essential to recognize and avoid enabling stupid people to exercise authority over others.

Show HN: Loodio 2 – A Simple Rechargable Bathroom Privacy Device

69 testmasterflex 2025-05-28 14:19:02 UTC loodio.com Llama Share 85

Loodio is a smart speaker with 100 pre-installed songs, night lights, and sound cancellation. It automatically plays music and blocks sound when motion is detected.

Why Good Ideas Die Quietly and Bad Ideas Go Viral

69 samizdis 2025-05-28 12:03:08 UTC newyorker.com GoogleT5 Share 119

David frum: antimemetics is a book about how memes are used to spread ideas. he says it's opportune now that samizdat is being used by venture capitalists, other bigwigs alike to raze everything to the ground, if we want to survive, we need institutions, not machines, says frank fram jibes of the book -

On eval in dynamic languages generally and in Racket specifically (2011)

68 swatson741 2025-05-28 20:44:04 UTC blog.racket-lang.org Llama Share 29

The eval function in dynamic languages can be powerful but also creates trouble if used unnecessarily, making it difficult to understand and translate programs. Experienced programmers avoid eval, but it's necessary in certain cases, such as when passing instructions to others who may not speak the same language.

Turn a Tesla into a mapping vehicle with Mapillary

66 faebi 2025-05-28 21:17:14 UTC blog.mapillary.com Full Share 27

Sorry, something went wrong. We're working on getting this fixed as soon as we can.

A privilege escalation from Chrome extensions (2023)

66 deryilz 2025-05-28 00:48:45 UTC 0x44.xyz Llama Share 10

A Chrome extension can run malicious code on a ChromeOS device by exploiting a vulnerability in the File Manager app, allowing it to access sensitive files and run code in the terminal. The bug was fixed, but not before it was used to demonstrate a privilege escalation attack that could have been used for ransomware or other malicious purposes.

I am disappointed in the AI discourse

64 steveklabnik 2025-05-28 17:34:52 UTC steveklabnik.com Llama Share 85

The author is frustrated with polarized online discussions about AI, where both pro-AI and anti-AI sides make easily verifiable false claims. They want to have a reasonable discussion about AI capabilities and ethics.

Texas' annual reading test adjusted difficulty yearly, masking improvement

62 cratermoon 2025-05-28 13:31:07 UTC theconversation.com Llama Share 80

A Texas education researcher found that the state's annual reading test scores were flat from 2012 to 2021 due to the test's design, which adjusts difficulty levels each year. This design makes it impossible to determine if students' lack of improvement was due to poor teaching or the test's flaws.

The 'Man-Eater' Screwworm Is Coming

62 fortran77 2025-05-28 12:37:07 UTC theatlantic.com Llama Share 45

The US has been fighting a 70-year aerial war against the New World screwworm, a parasite that eats animals alive, but the barrier was breached in 2022 and the parasite has advanced 1,600 miles through Mexico. The US cattle industry is unprepared for the screwworm's return, which could take decades to push back down to Panama.

Launch HN: MindFort (YC X25) – AI agents for continuous pentesting

60 bveiseh 2025-05-28 16:00:44 UTC news.ycombinator.com Llama Share 24

MindFort is building AI agents to continuously find and patch security vulnerabilities in web applications, reducing false positives and manual testing costs. Their AI agents perform reconnaissance, exploit vulnerabilities, and suggest patches, outperforming traditional scanners and manual testing methods.

Unhappy Meals (2007)

59 pramodbiligiri 2025-05-28 19:04:14 UTC michaelpollan.com GoogleT5 Share 95

David frum: nutritionists say fat doesn't make you fat; carbs do. he says research shows people lie about their food intake, so they focus on wrong nutrients - beta carotene, for example, found in vegetables, fruits vs. coca leaves, cocaine, crack, etc. people who eat according to rules of a traditional food culture are generally healthier, frank says. "nutritionism" is

Designing Pareto-optimal RAG workflows with syftr

50 roma_glushko 2025-05-28 14:01:05 UTC datarobot.com Llama Share 10

syftr is an open source framework that automatically identifies Pareto-optimal generative AI workflows balancing accuracy, cost, and latency. It efficiently searches vast configuration spaces using multi-objective Bayesian Optimization.

Show HN: Smart Silence – Remind your iPhone to stay quiet in quiet places

50 ebagsnave 2025-05-28 13:51:49 UTC testflight.apple.com Llama Share 32

To test a beta app, install TestFlight and accept an invitation to download the app. You can test the app for up to 90 days, and automatic updates can be turned on to install the latest beta builds.

Monks Behaving Badly: Explaining Buddhist Violence in Asia

45 surprisetalk 2025-05-28 12:48:40 UTC direct.mit.edu GoogleT5 Share 76

Buddhist violence intensified as the sangha and the government moved closer together. buddhist vigilantes identified the major threat to their political and cultural hegemony as arising from such religious out-groups - nilay s. manchanda, ed. smith, and dr. david wilkinson. monks behaving badly: explaining why buddhism becomes violent in some places but not others. the study

Flash Back: An “oral” history of Flash

44 surprisetalk 2025-05-28 12:21:44 UTC goodinternetmagazine.com Llama Share 48

The author reminisces about the early days of the web and Flash, a multimedia platform that enabled interactive content, but also had security, performance, and accessibility issues. Despite its flaws, Flash played a significant role in shaping online culture and paved the way for modern web standards.

Printing metal on glass with lasers [video]

44 surprisetalk 2025-05-28 04:07:20 UTC youtube.com Embed Share 13

The Illusion of Causality in Charts

42 skadamat 2025-05-28 17:49:52 UTC filwd.substack.com Llama Share 22

The user discusses how various charts can imply causality, even when a relationship may not exist, and identifies four main patterns that can lead to this effect: factor, event, proximity, and covariation. These patterns can be applied to all types of charts, enabling the user to generalize the idea of implied causality and develop data thinking skills.

Why is it so hard to get families to live in community houses?

41 caser 2025-05-28 15:41:56 UTC supernuclear.substack.com Llama Share 74

In the US, many people view parenting as a huge sacrifice, but in some communities, like villages in India, raising children is a communal effort. Raising kids in community, like in villages or co-living spaces, can make parenting easier and more joyful.

Using Postgres pg_test_fsync tool for testing low latency writes

40 mfiguiere 2025-05-28 02:15:15 UTC tanelpoder.com Llama Share 16

The tool pg_test_fsync is used to test disk performance for low latency writes. It shows that enterprise-grade SSDs with write-through cache have significantly lower synchronous write latency.

Ask HN: Stopping YC backed business from advertizing fake jobs at my company

37 quacksilver 2025-05-28 07:53:54 UTC news.ycombinator.com Llama Share 12

You're receiving unsolicited job applications from candidates misled by a business claiming YC backing, which is actually building a job site at www.featurii.com. You're hesitant to respond due to potential phishing and image damage concerns.

Everything We Learned About Lockheed's "QuadStar" Stinger Missile Replacement

36 jandrewrogers 2025-05-28 22:43:42 UTC twz.com Llama Share 55

Lockheed Martin's QuadStar missile is a surface-to-air missile designed to replace the Stinger, with advanced seeker technology and a more aerodynamically efficient design. It is part of the US Army's Next-Generation Short-Range Interceptor (NGSRI) program, which aims to provide improved target acquisition and lethality with increased ranges.

The anomalous magnetic moment of the muon in the Standard Model: an update

36 evanb 2025-05-28 20:13:56 UTC arxiv.org Llama Share 31

Researchers updated the Standard Model prediction for the muon anomalous magnetic moment, reducing uncertainty by almost a factor of two. The new prediction shows no tension with current experimental results, but future efforts aim to achieve higher precision.

How large should your sample size be?

36 sebg 2025-05-28 08:26:54 UTC vickiboykis.com Llama Share 26

The author discusses sampling in big data problems, where a subset of data is representative of the entire population. The goal is to find a sample size that accurately represents the population with a certain margin of error and confidence level.

Singularities in Space-Time Prove Hard to Kill

36 nsoonhui 2025-05-28 04:04:37 UTC quantamagazine.org Llama Share 53

Physicists struggle to understand singularities, points where space-time seems to stop, predicted by Einstein's general theory of relativity. Singularities may be more than mathematical artifacts, hinting that our universe contains points where space-time frays.

Global high-performance proof-of-stake blockchain with erasure coding

36 lawrenceyan 2025-05-28 02:47:53 UTC github.com/qkniep Llama Share 105

A local cluster with 6 nodes is spawned for Alpenglow protocol simulations via UDP on localhost. Simulations test Rotor resilience, latency, and bandwidth requirements, with cumbersome configuration via const values in main.rs.

Fractran Interpreter

33 ColinWright 2025-05-28 11:07:12 UTC tjwei.github.io Llama Share 3

FRACTRAN is a Turing-complete programming language with a list of fractions and an initial integer input. It updates the integer by multiplying with fractions that produce integers until no more fractions work.

Vibe coding for teams, thoughts to date

31 todsacerdoti 2025-05-28 08:18:32 UTC laughingmeme.org Llama Share 26

LLMs have disrupted code writing but not fundamental physics of team coding, and they don't change the fact that understanding code is harder than writing it. They lower the cost of writing code but also create complex, unique, and noisy codebases that require new tools and models to manage.

Show HN: Wetlands – a lightweight Python library for managing Conda environments

30 arthursw 2025-05-28 14:54:04 UTC arthursw.github.io Llama Share 42

Wetlands creates isolated Conda environments for plugin systems and external modules without conflicts. It uses Pixi or Micromamba for fast environment handling and simplified execution.

FlowTSE: Target Speaker Extraction with Flow Matching

25 agold97 2025-05-28 14:30:33 UTC arxiv.org Llama Share 2

Researchers propose FlowTSE, a simple TSE approach using conditional flow matching for speaker extraction. FlowTSE achieves strong results on standard benchmarks, outperforming or matching strong baselines.

Driverless Semi Trucks Are Here, with Little Regulation and Big Promises

24 bookofjoe 2025-05-28 10:42:05 UTC nytimes.com Full Share 106

Please enable JS and disable any ad blocker

LLM Shibboleths determine AI effectiveness

23 brilee 2025-05-28 17:22:48 UTC moderndescartes.com Llama Share 1

LLMs can produce varying results based on user expertise and question phrasing, leading to different experiences despite interacting with the same AI. Effective use of LLMs requires understanding their strengths and weaknesses, using shibboleths, and developing expertise in the domain to generate quality work.

An Extreme Cousin for Pluto? Possible Dwarf Planet at Solar System Edge

23 raattgift 2025-05-28 03:23:53 UTC ias.edu Llama Share 11

A team led by Sihao Cheng discovered a large trans-Neptunian object, 2017 OF201, at the edge of our solar system. The object's extreme orbit and large size suggest a complex history of gravitational interactions and may challenge the hypothesis of Planet X or Planet Nine.

Exploring a Language Runtime with Bpftrace

22 mgaudet 2025-05-28 16:46:55 UTC mgaudet.ca Llama Share 3

User learned eBPF and bpftrace to solve a problem with SpiderMonkey's Rooted constructor. They created a probe to track source locations of Rooted creations and found 75 million unnecessary calls, leading to bug reports and potential optimizations.

The Decline of Battery Life (2021)

21 akyuu 2025-05-28 16:45:02 UTC brainbaking.com Llama Share 31

The Evercade handheld gaming device has a battery life of four hours, a significant decrease from older devices like the original Game Boy which lasted over 20 hours. This trend of shorter battery lives in devices is concerning for end users who value longevity.

Ask HN: Is anyone using AI conversation partners?

21 rickcarlino 2025-05-28 02:49:58 UTC news.ycombinator.com Llama Share 35

The user is interested in using LLMs for language learning but finds current conversational chat apps bland and lacking in engagement. They suggest using LLMs to create interactive storytelling environments or specific goal-oriented conversations to improve language learning.

Personal Software Runtime Inspired by Emacs, Plan 9, Erlang, Hypermedia and Unix

20 veqq 2025-05-28 15:09:11 UTC github.com/leoshimo Llama Share 0

vrs is a personal software runtime combining ideas from Emacs, Erlang, and Unix for a holistic programming experience. It uses green threads, message passing, and isolated environments for a sandbox-like development environment.

Diffusion vs. Autoregressive Language Models: A Text Embedding Perspective

19 doener 2025-05-28 20:27:45 UTC arxiv.org Llama Share 1

Researchers propose using diffusion language models for text embeddings due to their bidirectional architecture, which outperforms LLM-based models by 20% on long-document retrieval and 8% on reasoning-intensive retrieval. Bidirectional attention is crucial for encoding global context in long and complex text.

The Colorado River is running low. The picture looks even worse underground

19 dylan604 2025-05-28 12:32:03 UTC washingtonpost.com Full Share 8

We Tested 7 Languages Under Extreme Load and Only One Didn't Crash

18 nnx 2025-05-28 08:43:52 UTC freedium.cfd Llama Share 35

Erlang emerged as the most resilient language in extreme workload testing, thanks to its "let it crash" philosophy and supervision tree pattern. Its architecture allowed it to recover from failures that crippled other languages, highlighting the importance of resource management and fault tolerance in system reliability.

Why are some rocks on the moon highly magnetic?

17 gmays 2025-05-28 22:24:19 UTC news.mit.edu Llama Share 1

MIT scientists propose a combination of a weak magnetic field and a large impact created a strong magnetic field on the moon's far side. The impact generated a plasma cloud that amplified the weak field, explaining highly magnetic rocks on the far side.

The Captcha Paradox

12 cfievet 2025-05-28 05:25:35 UTC talkingrobot.com Llama Share 8

Companies are investing in captchas to prevent AI impersonation, but intelligent machines are increasingly outsmarting these puzzles. To address this, new projects like World and Humanity Protocol are developing biometric-based methods to verify human identity in digital spaces.

From Clocks to Chaos: The Rhythms of Life

12 teleforce 2025-05-28 05:16:28 UTC press.princeton.edu Llama Share 2

Two researchers, Leon Glass and Michael Mackey, explore physiological rhythms in their book "From Clocks to Chaos". They discuss rhythm generation, effects of perturbation, and organization in space, linking rhythm abnormalities to disease.

Show HN: Connecting People Through AI-Powered Video Sentiment Matching

12 armini 2025-05-28 02:42:08 UTC loom.com Full Share 7

This showcases Kuky's innovative approach to user engagement through video-based profiles. Users upload short videos sharing their personal stories, which are then analyzed using Large Language Models (LLMs) to extract sentiments, themes, and emotional tones. visit www.kuky.com to learn more

Oldest known tools made from whale bones dated to 20k years ago

11 gmays 2025-05-28 16:52:04 UTC apnews.com Full Share 0