Qwen 3.8 27B

219 erdaltoprak 2026-08-14 15:22:37 UTC huggingface.co Llama Share 178

This repository provides FP8-quantized model weights and configuration files for the post-trained Qwen3.8-27B model, offering managed inference without infrastructure maintenance through the Qwen Cloud service. The model supports various tasks, including vision-language understanding, agent execution, and long-horizon agentic tasks.

Seven books I keep close because I love them

120 surprisetalk 2026-08-14 15:22:37 UTC blog.plover.com Llama Share 41

The author has a collection of books on their shelf, including Roget's Thesaurus, Sir Thomas Browne's works and the Decameron. They reflect on how these books inspire and inform their writing.

RustDesk now supports true unattended remote access on Wayland

43 rustdesk 2026-08-14 16:22:37 UTC rustdesk.com Llama Share 8

RustDesk now offers unattended access on Wayland without approval, supporting multi-monitor setups. A preview build is available for x86_64 Debian/Ubuntu-based systems to test and provide feedback.

Introducing Toast 1

83 mplappert 2026-08-14 15:22:37 UTC mixedbread.com Llama Share 23

Toast 1, a specialized search agent, offers frontier search quality at up to 10x cheaper and 12x faster than existing models. It can be used as a standalone model or as a subagent with other models.

Google is making private AI practical with homomorphic encryption

50 u1hcw9nx 2026-08-14 16:22:37 UTC blog.google Llama Share 28

Google's HEIR compiler toolchain enables secure private AI inference using homomorphic encryption, balancing privacy and security. The open-source tool converts pre-trained models to operate on encrypted inputs, making it a one-click solution for non-experts.

Why does Opus 5 feel worse to work with?

403 numeri 2026-08-14 10:22:37 UTC mun-logadan.github.io Llama Share 394

The author believes Opus 5 is a capable model but feels like a downgrade due to its tendency to make bold assumptions in the face of ambiguity, prioritizing benchmark performance over nuanced understanding. This approach may not be suitable for real-life applications where agents need to ask questions and seek clarification.

GLM-5.3: Frontier coding with emergent cyber capabilities

908 pella 2026-08-14 05:22:37 UTC z.ai Full Share 461

AI by Hand

19 sans_souse 2026-08-14 16:22:37 UTC byhand.ai Full Share 2

Math, Algorithms, Architectures, by hand. Click to read AI by Hand ✍️, by Prof. Tom Yeh, a Substack publication with tens of thousands of subscribers.

Every Fucking Website (2020)

520 doubletwoyou 2026-08-14 15:22:37 UTC lxe.github.io Llama Share 308

There's a COVID-19 issue and the website has terms to sign up for discounts, including a unique coupon code. Users must agree to these terms before getting 10% off.

In Australia, a home battery boom has helped cut wholesale power prices

206 speckx 2026-08-14 14:22:37 UTC e360.yale.edu Llama Share 160

Australia's home battery subsidy program has led to over 500,000 installations, slashing wholesale power prices by roughly half. The program is helping households use more solar energy and reducing grid costs nationwide.

France's top court blocks social media ban for under-15s

81 BlueBerry2001 2026-08-14 16:22:37 UTC reuters.com Full Share 24

Please enable JS and disable any ad blocker

RayforceDB – a pure C analytics database with a Lisp-like syntax

15 sourdecor 2026-08-14 16:22:37 UTC rayforcedb.com Llama Share 6

Rayforce combines columnar analytics, graph traversal, and recursive queries into a single embeddable execution pipeline for teams measuring latency in microseconds. It offers a unified analytics and graph pipeline with a managed cloud experience for seamless production workloads.

I turned my RSS feeds into an e-ink newspaper to stop reading on my phone

53 speckx 2026-08-14 15:22:37 UTC heyjonny.dev Llama Share 22

The author discovered a pocket-sized e-reader called the X4, which they customized with open source firmware Crosspoint to improve reading experience. They created their own "newspaper" using an app called Feedpaper, allowing them to read personal blogs on the device.

Ultraviolet Bird Photography

18 EndXA 2026-08-14 15:22:37 UTC uvbirds.com Llama Share 1

Birds see colors that humans can't, including ultraviolet light, which affects their plumage and behavior. This hidden dimension of color helps birds recognize species and navigate their environment.

Maximizing the value of your Claude Code sessions

10 twapi 2026-08-14 16:22:37 UTC claude.com Llama Share 0

Claude Code provides tips to optimize token usage and session efficiency, including setting models and effort levels before starting, using /context to view loaded context, compacting conversations when taking breaks.

Don't classify, hallucinate

161 softwaredoug 2026-08-14 11:22:37 UTC softwaredoug.com Llama Share 71

This text discusses using LLMs to classify products, specifically with the Wayfair WANDS e-commerce dataset. It proposes a method of asking an LLM to generate fake classifications and then finding the most similar real ones through dot product calculation.

"Solving a largely imaginary user goal"

29 euthymiclabs 2026-08-14 16:22:37 UTC unsung.aresluna.org Llama Share 33

The author agrees with Lea Verou's proposal for a "smart" dual-state dark mode toggle, which simplifies the user interface by hiding complexity. This approach abstracts away underlying system complexities and focuses on user goals, reducing cognitive load and improving overall usability.

DeepSeek peak/off-peak pricing update

198 fagnerbrack 2026-08-14 10:22:37 UTC api-docs.deepseek.com Llama Share 108

V4-Pro and V4-Flash models have varying reasoning efforts for tasks, with Expert Mode available on app/web. API pricing is updated with peak and off-peak rates, offering more flexible workload scheduling options.

Cursor is now a part of SpaceX

50 billy_jones_75 2026-08-14 13:22:37 UTC cursor.com Llama Share 26

Cursor has been acquired by SpaceX, gaining access to the largest fleet of GPUs. This acquisition will enable Cursor to build stronger models at lower cost, making it easier for users to solve harder problems.

Count Binface receives over a quarter of votes in Clacton by-election

10 tcp_handshaker 2026-08-14 16:47:37 UTC bbc.com Llama Share 2

Count Binface, a comedian and satirist, received nearly 10,000 votes in the Clacton by-election. He has previously run for election as Lord Buckethead and won viral fame with his novelty candidacy.

Self-hosted web push Cloudflare Worker, works on iOS

31 saiday 2026-08-14 15:22:37 UTC kukuroo.cc Llama Share 10

Notifications should be sent directly to a user's device via Cloudflare Workers, bypassing middlemen like apps and servers. This approach allows for secure, encrypted alerts without requiring an app or dashboard access.

The TEMU-Fication of Software, Digital Goods and Services

73 surprisetalk 2026-08-14 12:22:37 UTC xn--gckvb8fzb.com Llama Share 50

The author speculates that the future of software, books, music, and movies will have a two-tier market system where most content is cheaply generated by AI but high-quality human-made work becomes a luxury segment. The bulk of the market will be driven by ultra-cheap AI-generated content while a smaller niche will value and pay for human-made work.

A Contract-Grade Verifier for LLM-Generated GPU Kernels

3 Jimmc414 2026-08-14 16:59:37 UTC arxiv.org Prefix Share 0

Systems that generate GPU kernels with language models report high correctness rates. Those rates come from a single loose test: run the kernel on a few random inputs at one fixed shape and accept it if the output is close to a reference. A kernel can pass that test and still be silently wrong. It can return an ordinary number where the true answer is a NaN or an infinity, differ from run to ...

AI Model Atlas – visualizing populations of ML models as interconnected 3D graph

22 bj-rn 2026-08-14 15:22:37 UTC run.cosmograph.app Full Share 4

Show HN: Graft – Claude Code hooks that cut grep tokens by 42%

26 shrishdwi 2026-08-14 16:22:37 UTC github.com/nanonets Llama Share 25

Graft is a tool that improves the efficiency and correctness of code completion by building a graph of the codebase, allowing agents to skip unnecessary exploration. Graft has been verified through SWE-bench and shown to be cheaper, faster, and more correct than traditional methods.

Gödel, Escher, Elisp: The Beauty of Macros

31 JNRowe 2026-08-14 14:22:37 UTC chiply.dev Llama Share 9

Emacs Lisp macros allow users to leverage the blur between program and data, enabling flexible and expressive code. Macros can be used to extend the language, adding new defining forms that communicate in the problem domain's own diction.

Discrete Fourier Transform by Hand

5 Bluestein 2026-08-14 16:54:37 UTC byhand.ai LocalLLM Share 1

DFT simplifies to matrix multiplication, revealing its structure as a series of cosine wave projections. Neural networks like U-Net learn similar transformations, while DFT remains a fixed, manually designed method for signal analysis.

HashAgent – Share an AI agent as a URL, runs locally via WebGPU

30 masonhsu 2026-08-14 13:22:37 UTC hashagent.pages.dev Full Share 3

Create and share private AI agents as URLs. Local WebGPU inference, built-in web search, no account or tracking.

Show HN: LuaCAD – Parametric CAD Scripted in Lua

3 adius 2026-08-14 16:45:37 UTC luacad.ad-si.com Prefix Share 0

Scriptable CAD with Lua. Write parametric 2D and 3D models in Lua and export them to 3MF, STL, OBJ, PLY, OFF, AMF, or SCAD, or render them straight to a PNG. LuaCAD embeds Lua 5.4 in a Rust engine that evaluates CSG operations directly (via Manifold) or generates OpenSCAD code for external rendering. Operator overloading for natural CSG syntax (a + b, a - b, a * b) Both crates vendor their ...

The Qdrant Output Connector

4 softwaredoug 2026-08-14 16:37:37 UTC opencrawling.org Prefix Share 0

As AI applications expand across enterprise ecosystems, vector storage demands extreme query throughput, low memory footprint, and strict security guarantees. While vector databases excel at similarity search, enterprise content repositories (like SharePoint, Alfresco, and Iceberg) require document-level Access Control Lists (ACLs) to ensure users only retrieve information they have explicit ...