The 29th International Obfuscated C Code Contest (IOCCC) 2025 Winners

83 matt_d 2026-06-07 06:00:54 UTC ioccc.org Llama Share 10

IOCCC29 had a high submission quality and similar volume to last year, possibly due to improved website design and social media presence. The contest's rules and guidelines were rewritten for this year, and the next contest, IOCCC30, is planned for late 2026.

I design with Claude more than Figma now

100 MrBuddyCasino 2026-06-07 06:00:54 UTC blog.janestreet.com Llama Share 64

The author was initially skeptical of LLMs but found them indispensable at Jane Street, especially for tasks they were not good at. They now use LLMs to create prototype features that exactly match their design ideas, freeing up time for iteration and refinement.

Valve P2P networking broken for more than 2 months

124 babuskov 2026-06-07 04:00:54 UTC github.com/valvesoftware Llama Share 57

Steam Networking has a major issue affecting Israeli players in P2P games since 13/03, causing high ping rates around 120ms. The issue is specific to Steam Networking and doesn't occur in non-Steam games or cross-play with non-PC platforms.

Field of clones: How horse replicas came to dominate polo

72 gscott 2026-06-07 04:00:54 UTC knowablemagazine.org Llama Share 37

Adolfo Cambiaso, a renowned polo player, pioneered horse cloning in 2006 to preserve his exceptional horses' genetics. He successfully cloned his flagship mare Cuartetera, leading to La Dolfina's unprecedented dominance in polo.

My Software North Star

40 kristoff_it 2026-06-07 05:00:54 UTC kristoff.it Llama Share 13

Software should be useful, correct, maintainable, and efficient to maximize end user utility. Developer experience is secondary, serving only to deliver software that is loved by users.

Tokenomics: Quantifying Where Tokens Are Used in Agentic Software Engineering

89 Anon84 2026-06-07 02:00:54 UTC arxiv.org Llama Share 27

Researchers analyzed token consumption in LLM-MA systems across software development stages. They found Code Review accounts for 59.4% of token consumption, with input tokens making up 53.9% of total consumption.

Harness engineering: Leveraging Codex in an agent-first world

180 pramodbiligiri 2026-06-07 00:00:54 UTC openai.com Llama Share 116

We used Codex to generate a million lines of code for a software product in 5 months with a team of 7 engineers, achieving an average throughput of 3.5 PRs per engineer per day. To maintain the codebase, we focused on designing environments, scaffolding, and feedback loops that enable agents to do reliable work, enforcing invariants and mechanical rules to keep the codebase legible and consistent.

Ntsc-rs – open-source video emulation of analog TV and VHS artifacts

330 gregsadetsky 2026-06-06 20:00:54 UTC ntsc.rs Llama Share 87

ntsc-rs is a free, open-source Rust effect that accurately emulates analog TV and VHS artifacts. It's multithreaded, SIMD-accelerated and available as a standalone, web app, and plugin for various video editing software.

Symbolica 2.0: Programmable Symbols for Python and Rust

67 mmastrac 2026-06-07 03:00:54 UTC symbolica.io Llama Share 6

Symbolica 2.0 is a high-performance symbolic computation framework with programmable symbols, improved ergonomics, and faster performance. It supports multiple numerical domains, including double-float arithmetic and JIT compilation.

Public Domain Image Archive

99 davidbarker 2026-06-07 01:00:54 UTC pdimagearchive.org Full Share 16

Explore our hand-picked collection of 11,082 out-of-copyright works, free for all to browse, download, and reuse. This is a living database with new images added every week.

How Liminalism Became the Defining Aesthetic of Our Time

53 zeech 2026-06-07 03:00:54 UTC hyperallergic.com Llama Share 33

Liminalism is an internet art movement that explores abandoned, empty spaces, evoking feelings of nostalgia, lostness, and uncertainty. This aesthetic is a reflection of our current moment, characterized by anonymity, alienation, and anxiety, and is often associated with the concept of the "Backrooms."

Biohub releases a world model of protein biology

55 gmays 2026-06-07 02:00:54 UTC biohub.org Llama Share 2

Biohub released a world model of protein biology, a scientific engine for prediction, design, and discovery that can map proteins and design new binders. The engine, ESMFold2, can predict protein structures and design binders in days, accelerating early therapeutic discovery.

Show HN: Oproxy – inspect and modify network traffic from the browser

39 sauravrao637 2026-06-07 03:00:54 UTC github.com/sauravrao637 Llama Share 6

oproxy is a local proxy for inspecting and modifying traffic, allowing developers to test and debug applications. It supports various features like rule sets, mock responses, and upstream proxy chaining.

Introducing Boron Buckyballs: Theory that B80 cages can’t be made is disproved

79 crescit_eundo 2026-06-07 00:00:54 UTC cen.acs.org Llama Share 19

Chemists have observed a boron buckminsterfullerene for the first time, providing evidence for an 80-atom cage structure. This discovery challenges density functional theory and may lead to new materials such as semiconductors or superconductors.

Human-Like Neural Nets by Catapulting

28 telotortium 2026-06-07 05:00:54 UTC gwern.net Llama Share 3

The author proposes a new scaling paradigm for deep learning, suggesting that human brains and neural networks (NNs) differ in their bias-variance tradeoff, with human brains minimizing bias and NNs minimizing variance. This difference could explain various anomalies in machine and biological intelligence, such as the need for large datasets and compute in NNs, the lack of transfer learning ...

Moving beyond fork() + exec()

290 jwilk 2026-06-06 15:00:54 UTC lwn.net Llama Share 287

A recent Linux kernel proposal to add "spawn templates" to optimize the fork() and exec() pattern has been rejected, but it may point to a new process-creation primitive in the future. The discussion highlights the need for a better process-creation mechanism, with some suggesting a native implementation of posix_spawn() and others proposing alternative approaches using existing system calls.

Arithmetic Without Numbers – How LLMs Do Math

13 old_sound 2026-06-07 06:00:54 UTC alvaro-videla.com Llama Share 4

A language model can perform arithmetic operations without explicit programming, but the question is whether it recalls a pattern, runs an algorithm, or merely produces a plausible next token. The project Rune found that a language model's internal state can supply operation and operand structure for arithmetic prompts, which can drive a calculator route under an opaque-prompt boundary.

Speculative KV coding: losslessly compressing KV cache by up to ~4×

5 kkm 2026-06-07 07:00:54 UTC fergusfinn.com Prefix Share 0

· 14 min read · Cover: Andreas Cellarius, Theoria Lunae, from Harmonia Macrocosmica (1660), engraved by Johannes van Loon. The size of LLM context grows by the day. KV caching is what makes running those long contexts affordable: it trades compute for memory so the model doesn’t re-prefill work it has already done. But as agentic workflows push contexts ever longer, storing and moving the ...

Meta confirms 1000s of Instagram accounts were hacked by abusing its AI chatbot

560 speckx 2026-06-06 19:00:54 UTC this.weekinsecurity.com Llama Share 199

Meta notified 20,225 people whose Instagram accounts were hijacked through a flaw in its AI chatbot, allowing hackers to access account information and posts. The breach occurred from April 17 to this week, with Meta disabling the chatbot and instructing users to reset their passwords.

Nvidia is proposing a beast of a CPU system for Windows PCs

270 tosh 2026-06-06 13:00:54 UTC twitter.com/lemire Full Share 459

Something went wrong, but don’t fret — let’s give it another shot. Some privacy related extensions may cause issues on x.com. Please disable them and try again.

Zeroserve: A zero-config web server you can script with eBPF

223 losfair 2026-06-06 15:00:54 UTC su3.io Llama Share 54

zeroserve is a fast, zero-config HTTPS server that serves websites from a single tarball and supports eBPF scripting for middleware and proxying. It beats nginx in serving small files and has a faster proxying path.

Sem: New primitive for code understanding – not LSPs, but entities on top of Git

113 rohanucla 2026-06-06 21:00:54 UTC ataraxy-labs.github.io Llama Share 40

Rohan made changes to the code, adding and modifying functions, and deleting another. The changes affect 42 entities, including the loginRoute and authMiddleware, due to dependencies on db.findUser and rateLimiter.

Google to pay SpaceX $920M a month for compute capacity at xAI data centers

231 toephu2 2026-06-06 08:00:54 UTC cnbc.com Llama Share 807

SpaceX has signed a deal with Google to provide AI compute capacity for $920 million a month until June 2029. The agreement allows Google to end the deal if SpaceX fails to deliver the committed amount of GPUs by September 30, 2026.

Show HN: TakoVM – Isolated model and tool execution used by enterprises

19 sakuraiben 2026-06-07 03:00:54 UTC github.com/las7 Llama Share 7

tako-vm is a tool for running AI-generated code in isolated Docker containers with optional gVisor sandboxing. It includes job queues, retries, and execution history, and can be installed and run with various commands.

Ask HN: What was your "oh shit" moment with GenAI?

603 andrehacker 2026-06-06 08:00:54 UTC news.ycombinator.com Llama Share 1003

The user initially dismissed AI models like DALL-E and ChatGPT as parlor tricks but had a series of "Oh shit" moments as they realized the potential of these models to solve real problems and change their approach to development. These moments included using Claude code to generate a Zillow description, building a web tool to sort media files, and using AI to translate code and track down ...

Show HN: Free animated icon library for Vue

18 evolabs 2026-06-07 04:00:54 UTC respeak-io.github.io Full Share 5

The largest animated icon library for Vue 3. 535 Lucide icons, drop-in and tree-shakeable, with hover, tap, and in-view triggers and a live gallery.

Motorola effectively bricked its entire line of WiFi routers without explanation

129 thisislife2 2026-06-06 15:00:54 UTC mashable.com Llama Share 57

Motorola's MotoSync+ app has been down for nearly a month, preventing users from setting up and using their WiFi routers. The issue has left many customers without access to their routers, with some unable to even factory reset their devices.

The circus freaks of open source

60 keyle 2026-06-07 05:00:54 UTC drewdevault.com Llama Share 16

The user discusses two examples of individuals, Terry A. Davis and Kent Overstreet, who created notable projects but struggled with mental health issues and were subsequently harassed and ridiculed online. The user argues that it's essential to approach these situations with compassion and respect for individuals' privacy, rather than sensationalizing their struggles or participating in ...

Pokemon Emerald Ported to WebAssembly (100k FPS)

311 tripplyons 2026-06-06 12:00:54 UTC pokeemerald.com Full Share 87

Keyboard: arrows, Z=A, X=B, Enter=Start, Shift=Select.

Unicode Fonts and Tools for X11

37 kristianp 2026-06-06 23:00:54 UTC cl.cam.ac.uk Llama Share 7

X11 fonts have been extended to include ISO 10646-1 Unicode characters, with new fonts and revised versions of existing fonts available. However, X11 was not designed to handle complex scripts like Arabic, Syriac, and Indic, requiring special libraries like Pango to support these scripts.