Naphtha Shortages Having a Growing Impact in Japan

https://www.nippon.com/en/ncommon/contents/japan-data/3004187/3004187.jpg
Calbee is switching to black-and-white packaging due to naphtha shortages caused by the Iran war. The shortage affects various industries, including food, manufacturing, and medical sectors.

SQLite is all you need for durable workflows

https://obeli.sk/blog/2026-05-29-sqlite-is-all-you-need/architecture.svg
The user agrees with DBOS that a separate orchestration tier is not necessary for durable execution, as workflow state can be kept in a local database like SQLite. A local SQLite database with Litestream backup to S3 is a simple and cost-effective way to achieve durability for many AI and experimentation workflows.

The dead economy theory

https://substackcdn.com/image/fetch/$s_!I8Nx!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F81a6d7ab-0c16-463d-b32c-3fc965eb4593_1500x911.png
The AI industry's focus on labor replacement threatens to eliminate human jobs, leading to economic instability and social unrest. The industry's prioritization of profit over people and lack of accountability may undermine democratic governance and exacerbate inequality.

Snowboard Kids 2 is 100% Decompiled

https://blog.chrislewis.au/snowboardkids2-decomp-report.svg
A C implementation of Snowboard Kids 2's functions has been completed, allowing for a readable and modifiable codebase. The project's success is attributed to the N64 decompilation community and coding agents, particularly Codex 5.5 xhigh.

Perry Compiles TypeScript directly to executables using SWC and LLVM

https://perryts.com/opengraph-image?edfa6fc43a1da671
Perry compiles TypeScript to native executables on various platforms. It produces standalone binaries with no runtime dependencies, enabling fast deployment and reproducible builds.

Math-to-Manim

https://camo.githubusercontent.com/57c7181a7ac2523dd28d1643f217afa063b4e407abee56077018ed5b63d21592/68747470733a2f2f6170692e737461722d686973746f72792e636f6d2f7376673f7265706f733d4861726c6579436f6f70732f4d6174682d546f2d4d616e696d26747970653d44617465
Math-To-Manim is a pipeline that turns math and physics prompts into Manim explainer videos and reusable artifacts. It uses a reasoning model to create an inspectable explanation path from question to understanding to animation.

Notes from the Mistral AI Now Summit

https://koenvangilst.nl/static/images/mistral-future-vision.jpeg
I attended the AI Now Summit by Mistral AI in Paris, learning about their full-stack AI approach, including compute, models, platforms, and consultancy. Mistral focuses on efficient, open, and custom models for European companies, emphasizing partnerships and on-prem deployment for sovereignty and return on investment.

MCP is dead?

https://cdn.prod.website-files.com/668aa55ff9fc77f31f6cc2f0/6a0f2926a73de3ec41126467_post%20image%20b.webp
The author argues that MCP consumes too much context and is slower than CLI approaches, but it's still useful for services without a strong CLI. They recommend using Skills to wrap existing CLIs for repeatable workflows and reserving MCP for services with no CLI or team-wide auth needs.
https://storage.googleapis.com/prusa3d-content-prod-14e8-wordpress-blog-prod/2026/05/7945ae91-easyprint-colormix-1024x576.jpg
Prusa developed a color mixing model called Prusa ColorMix to enable printing with dozens of colors on multi-material printers. The model is now integrated into PrusaSlicer and EasyPrint, allowing users to easily mix colors and predict the final result.

Shift will clean homes for free to train future robots

https://platform.theverge.com/wp-content/uploads/sites/2/2025/09/ROB_H_BLURPLE.jpg?quality=90&strip=all&crop=0%2C0%2C100%2C100&w=2400
Shift offers free home cleaning in exchange for footage to train robots, with a "magic hat" recording cleaners. The service is initially available in New York, with plans to expand to other cities.

WH proposes rules giving political appointees final approval on research grants

https://static.scientificamerican.com/dam/asset/dbb5e69a-08e8-4ebf-a0f3-45cc08846d1b/Vought.jpg?m=1779998711.831&w=600
The White House proposed regulations would give political appointees control over federal research grants and funding, replacing expert review committees. Critics say this would politicize science and silence research that is inconvenient to the administration.

It's hard to justify buying a Framework 12

https://www.jeffgeerling.com/blog/2026/its-hard-to-justify-framework-12/macbook-neo-and-framework-12.jpeg
Your nephew is choosing between a MacBook Neo and a Framework 12 laptop. The MacBook Neo offers better performance, display, and build quality at a lower price, making it the better value.

Show HN: Tiny-vLLM – high performance LLM inference engine in C++ and CUDA

https://opengraph.githubassets.com/e1c100870ce0e05c7d47f0e9e934eebc9ceee442c0d852cec0bef84a3b5af79c/jmaczan/tiny-vllm
We are building a high-performance LLM inference engine with C++ and CUDA, focusing on loading and running a pre-trained LLM model, specifically Llama 3.2 1B Instruct, and understanding the basics of floating-point numbers and their representation in binary format.

Ember.js 7.0

https://blog.emberjs.com/images/logos/e-icon.png
Ember v7.0 removes deprecated features and includes bug fixes, marking Ember 6.12 as LTS. The release follows Ember's Major Version Policy, ensuring backwards compatibility for existing code.

Liquid AI reveals 8B-A1B MoE trained on 38T

https://cdn.prod.website-files.com/67cb8aa6e9184b6e44813189/6a17874a7db7a603d7eb3627_lfm2_5_8b_a1b_benchmarks.png
Liquid AI released LFM2.5-8B-A1B, an edge model for fast tool calling on consumer hardware. It expands context window, doubles vocabulary, and achieves tasks on entry-level laptops.

What Is a Dickover?

A dickover is a modal panel or popover that frustrates users by obscuring content to force unwanted interactions. They're common on websites and apps, often asking for subscriptions, cookie acceptance, or other unnecessary actions.

Bijou64: A variable-length integer encoding

https://www.inkandswitch.com/tangents/bijou64/leb128-noncanonical.png
bijou64 is a variable-length integer encoding that ensures canonicality, preventing multiple ways to represent a number, and runs faster than LEB128 in benchmarks. It uses a tag-based system and bit shuffling to achieve this, making it structurally safer and faster for certain use cases.

Citing 'severe' math deficits, UC faculty demand a return to SAT tests for STEM

https://ca-times.brightspotcdn.com/dims4/default/6d16572/2147483647/strip/true/crop/5472x3648+0+0/resize/1200x800!/quality/75/?url=https%3A%2F%2Fcalifornia-times-brightspot.s3.amazonaws.com%2F71%2F9a%2Fa7e16b5640b59762cc3e758ffc1a%2F1294188-me-uc-irvine-student-housing-001-ajs.jpg
UC faculty members are calling for standardized testing in STEM admissions due to concerns over students' readiness for college-level math. They want SAT or ACT exams to be required starting in 2027 to ensure students can handle college-level math.

You can just say it

Humans are valuable regardless of AI capabilities. Their value lies in their intent and creative process, not just the quality of their output.

On Rendering Diffs

https://pierre.computer/pierre-co-diffs-og.png
A good review surface needs more than text. virtualization helps with rendering problems, and moving highlighting off the main thread helped with parts of the processing problem. to emulate scrolling entirely, no native scrollable region is used, with simulated scrollbar and content updated via requestanimationframe as the user moves through the document. we don’t virtualize horizontal scroll ...

Is AI causing a repeat of frontend’s lost decade?

https://mastrojs.github.io/blog/2026-05-23-is-AI-causing-a-repeat-of-frontends-lost-decade/og.png
The introduction of AI and automation in programming is deskilling the profession, making it easier for businesses to hire generalists and reducing the bargaining power of workers, similar to how frameworks deskilled frontend development in the past. While AI can increase efficiency and productivity, it also risks producing lower quality work and requires developers to understand how to use ...

The mysterious Hy3 LLM is topping OpenRouter Model Rankings by a large margin

https://minimaxir.com/2026/05/openrouter-hy3/rankings_052526.png
The author is puzzled by the sudden popularity of Hy3 preview on OpenRouter, a service that provides access to various LLMs, despite its lower quality compared to other models. After analyzing the data, the author suspects that a single large app is using Hy3 as its data-processing backbone, but the true reason for its popularity remains unclear.

Free full BGP feed. IPv4 and IPv6 (2020)

The project now includes IPv6 feeds, but users are responsible for configuration and any issues. Users must configure local-as 65001 for IPv4 and not send IPv6 prefixes, and can optimize TCP stack settings.

GTA 6 Developers Unionize

https://rockstarintel.com/wp-content/uploads/2025/05/Lucia_Caminos_01-1024x576.jpg
Rockstar staff formed the Rockstar Game Workers Union to combat company actions, disputing firings as union busting. The union will face Rockstar in court, with a trial date set, and is seeking donations for their legal battle.

Britain has crushed immigration, and harmed itself

https://www.economist.com/cdn-cgi/image/width=1424,quality=80,format=auto/content-assets/images/20260530_BRD001.jpg
Britain previously had an open door policy for immigrants, allowing foreign workers to be hired at low wages. Politicians discussing football may hold lessons for the Labour Party in addressing Britain's vulnerabilities.

Show HN: TV Explorer. Adding advanced UI to free online TV

https://tvexplorer.live/project/images/meta-image.jpg
Don't just watch TV. Explore over 10,000 free channels from around the world.

The California state assembly has passed the 'Protect Our Games Act'

https://static.invenglobal.com/upload/image/2026/05/29/i1780023554243286.jpeg
The California State Assembly passed the 'Protect Our Games Act' requiring digital game publishers to maintain game accessibility after service ends. The bill aims to protect user rights and preserve games, with potential implications for the US gaming industry.

We should be more tired than the model

https://vickiboykis.com/images/logo.png
User feels disconnected from code when using agentic code generation, lacking internal processing. Adding friction to development, such as rewriting code, asking questions, and critiquing, helps solidify user's foundation and improves tool usage.

Letter from the Duke of Wellington to the British Foreign Office (1809)

Wellington, a British general, is writing from Central Spain in 1812, requesting clarification on his instructions from the British government as he leads an army against the French. He believes his mission is either to train British clerks in Spain or to drive the French out of Spain.

CAPTCHAs can still detect AI agents

https://research.roundtable.ai/captchas-detect-ai/captcha-results.png
Researchers designed CogCAPTCHA30 to test how humans and AI complete tasks, finding that while AI can perform tasks like humans, it does so with different processes. The Process Turing Test measures process equivalence, raising questions about the types of language models that are humanlike and the robustness of this discrimination process.