DeepSeek open-sources inference optimizations with 60–85% faster generation [pdf]

https://opengraph.githubassets.com/601aadeacea68fa320d2b01c2813f8ab2f130b25f6a8f24cbe5bd3a0b9abb5d4/deepseek-ai/DeepSpec
DeepSpec: a full-stack codebase for training and evaluating speculative decoding algorithms - DeepSpec/DSpark_paper.pdf at main · deepseek-ai/DeepSpec

Fintech Engineering Handbook

Fintech systems require careful handling of money to prevent errors and ensure trust. Key patterns include idempotency, deduplication, reconciliation, and event sourcing to maintain accurate and reliable financial records.

Previewing GPT‑5.6 Sol: a next-generation model

https://images.ctfassets.net/kftzwdyauwt9/3RkIKhLVsVWcJQ3czkTNMH/c63f9c43efd82ddf863f87d44edca201/Spend_Controls_Artcard.png?w=3840&q=90&fm=webp
A limited preview of the new gpt5.6 is available to u.s. government partners. the preview is designed to test the effectiveness of safeguards against misuse based on real-world scenarios - and to help users prepare for attacks if they encounter them en route to the tool sandboxed in their home pcs or home labs, for example, he says – and how they might be used...

Long Wave radio era set to end with switch-off

https://www.economist.com/cdn-cgi/image/width=1424,quality=80,format=auto/content-assets/images/20260627_BRP001.jpg
A Midlands town stood out on British wireless sets in the 1930s due to a powerful radio transmitter near Droitwich. The transmitter's masts still send broadcasts across Britain and Europe, including coded messages during WWII and cricket commentary.

Linux on Older Hardware: The Complete Revival Guide

https://www.fosslinux.com/wp-content/uploads/2026/06/linux-older-hardware-revival-guide-2026.png
Liam's guide helps revive old hardware with Linux by selecting the right distro, optimizing RAM, and upgrading to an SSD, reducing electronic waste and extending the machine's life by 2-3 years. The guide provides a tier system for distro selection based on RAM, CPU architecture, and storage, and offers techniques for browser optimization, zram compression, and service trimming to squeeze ...

Beer CSS – Build material design in record time

https://www.beercss.com/logo.png
The text is a dummy sample of the printing industry, used since the 1500s. It has remained unchanged despite the transition to electronic typesetting.

WordStar: A Writer's Word Processor (1996)

https://www.sfwriter.com/wordstar-under-vdosplus.png
Robert J. Sawyer, a science fiction author, uses WordStar for writing due to its intuitive interface and long-hand-page metaphor, which allows for easy editing and navigation. WordStar's unique control-key commands and customizable options make it a more productive tool for writers.

Why does kinetic energy increase quadratically, not linearly, with speed? (2011)

https://i.sstatic.net/TeOX5.png
The kinetic energy of an object is proportional to the square of its velocity because it is a consequence of Newton's second law and the definition of work. The quadratic dependence of kinetic energy on velocity is a result of the symmetry properties of space and time, specifically the homogeneity and isotropy of space and the homogeneity of time.

The US Army Issued Ocarinas to Soldiers in World War II

https://www.flutetunes.com/articles/my-flute-goes-to-war/simplified_fingering_chart.jpg
During World War II, the US government shipped thousands of flutes to soldiers, including ocarinas, which were chosen for their simplicity and durability. The ocarinas were made from plastic molds and had a simple fingering system, making them easy to play and learn.

Faster KNN search in Manticore: 2-pass HNSW, batched distances, and AVX-512

https://miro.medium.com/v2/resize:fit:700/1*2IcQ1F_alFncLEQftzEPWA.jpeg
Manticore's KNN search engine improved by up to 29% at high k with no API changes or index rebuild. Optimizations target distance computation and memory access patterns.

OpenTTD 16.0-Beta1

OpenTTD 16 beta is available with improved map generation and new features like custom control of land height and aging rate of cargo payments. The OpenTTD 16 Title Game Competition has been launched on TT-Forums for community members to design and vote on the new title game.

U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

https://img.semafor.com/50a98e4cab1aadfa340671d35d9027ed27e6899b-740x493.png?w=740&q=75&auto=format&h=493
The US government lifted its block on Anthropic's Claude Mythos 5 AI model, allowing release to over 100 US institutions. The decision follows intense talks between the government and Anthropic, which committed to working on protocols and standards for its models.

AI in mathematics is forcing big questions

https://spectrum.ieee.org/media-library/a-photo-shows-a-man-standing-in-front-of-the-projection-of-a-computer-screen-thats-filled-with-computer-code.jpg?id=67007150&width=980&quality=100
Mathematicians are concerned that AI may replace human mathematicians, but some see it as a tool to aid in problem-solving and collaboration.

MicroVMs: Run isolated sandboxes with full lifecycle control

https://d2908q01vomqb2.cloudfront.net/da4b9237bacccdf19c0760cab7aec4a8359010b0/2026/06/22/Screenshot-2026-06-22-at-10.49.45%E2%80%AFAM-1024x577.png
AWS Lambda MicroVMs provides isolated, stateful execution environments for user or AI-generated code, offering virtual machine level isolation, rapid launch and resume, and stateful execution. It's available in multiple regions, with up to 16 vCPUs, 32 GB of memory, and 32 GB of disk per MicroVM.

Fusion Programming Language

https://fusion-lang.org/fut-dark.png
Fork me on GitHub Fusion is a language for creating reusable components across multiple languages. It produces lightweight, human-readable code that fits each target language.

Hellishly Slow Level 13 Deflate Compression

Libdeflate level 13 is a slow but effective DEFLATE compression level that saves 0.134% on Silesia corpus. It runs 56.4x slower than level 12 but is suitable for data compressed once and distributed many times.

Jest/Vitest interactive course (runs in the browser)

This course covers Jest and Vitest fundamentals for testing frontend applications, including assertions, mocks, and async testing. It's free and compatible with both Jest and Vitest, preparing you for React frontend testing.

U.S. government will decide who gets to use GPT-5.6

IBM MCGA Gate Array Reverse Engineering

https://opengraph.githubassets.com/8cb7ca151dec67b220c6ad4f2b0df9782dcc568265cc14fa075ac944e014100b/schlae/IBM_MCGA
IBM's MCGA is a low-cost video chipset used in PS/2 models 25 and 30, consisting of memory and video formatter gate arrays. The formatter gate array decodes ISA addresses, manages RAMDAC, and generates pixel data in graphics and text modes.

Anatomy of a Failed (Nation-State?) Attack

https://grack.com/_thumbs/bdd1d20a96688a5dfc4ee0f14eed1d0a-600-600
The author was targeted by a fake-interview scam using a RAT called PinpinRAT, which was embedded in a TypeScript repository. The scam was sophisticated and used multiple fake websites, stolen history, and a patient timeline to avoid detection.

The gap between open weights LLMs and closed source LLMs

https://blog.doubleword.ai/doubleword-icon.png
The gap between open weights LLMs and closed source LLMs is shrinking, with some benchmarks showing a near catch-up by December 2026. However, a broader analysis of 18 benchmarks shows a more moderate and consistent gap of around 5 months.

Show HN: Hacker News on a train station-style flip board

A Hacker News departures board displays top 20 stories with rank, score, and title on a split-flap display. It fetches the front page on a schedule and allows users to pin their own text and toggle flap clicks.

Ultrasound imaging of the brain

https://alephneuro.com/blog/mindeye-1.jpg
Researchers have developed a new type of hardware using ultrasound to record brain activity without drilling into the skull. This technology captures detailed images of the brain's vascular system, potentially leading to breakthroughs in understanding and treating neurological conditions.

Om

https://daringfireball.net/misc/2026/06/om-john-2025-06-09.jpeg
The author and Om Malik were close friends for 20 years, often attending Apple keynotes together, and Om's presence improved any room with his kind words and humor. He transformed from a quick-trigger blogger to a thoughtful essayist after a heart attack in 2008, and his writing continued to earn respect until his passing.

Foreign funds help make housing unaffordable: research

https://news.mccombs.utexas.edu/wp-content/uploads/2026/06/iStock-1495524595-1200x800.jpg?v=1782388015
US housing prices rose 60% from 2019 to 2025 due to foreign investment and low supply elasticity. Cities can control supply sensitivity through zoning and permitting, promoting affordable housing.

We can still stop California's 3D printer surveillance scheme

https://www.eff.org/files/banner_library/3dprintca.jpg
California State Assembly passed legislation mandating 3D print surveillance software despite EFF's warnings about privacy risks. The bill now heads to the state senate, where it needs to be stopped to protect creators' tools and users' rights.

A C++ implementation of a fast hash map and hash set using hopscotch hashing

https://opengraph.githubassets.com/5f70ee753ff7fdc5ff7bc5c46ed2426af8b855d2154f512d99f3833286db0542/Tessil/hopscotch-map
The hopscotch-map library is a C++ implementation of a fast hash map and hash set using open-addressing and hopscotch hashing to resolve collisions. It offers better performances than std::unordered_map in most cases and is closely similar to google::dense_hash_map while using less memory and providing more functionalities.

SCC Technical Assistance Program

https://nerocam.com/SCC_Images/image001.jpg
A friend's Mazda MX-6 is beating his 2001 Sentra SE in a heads-up drag race. To improve the Sentra's performance, the car was stripped of its interior, wheels, and body kit, and modifications were made to the engine and exhaust system.

What Is a Nomogram and Why Would It Interest Me?

https://lefakkomies.github.io/pynomo-doc/_images/meyer_nomogram.png
A nomogram is a graphical calculator that uses a set of numbered scales to solve mathematical formulas. It was invented in 1880 and is useful for fast calculations of complicated formulas.

Show HN: Smart model routing directly in Claude, Codex and Cursor

https://opengraph.githubassets.com/48d7cb17b97528ec18ef091845290503eceeb51b9fa260365287883e8c090fbf/workweave/router
Weave Router is a drop-in proxy that picks the best model for every request from Anthropic, OpenAI, and Gemini. It routes requests per request and speaks everyone's API, with a cluster scorer derived from Avengers-Pro 2.