AI fuels more than half of cybercrime in Africa as scams surge – Interpol

86 bookofjoe 2026-08-04 22:30:07 UTC africanews.com Llama Share 41

Artificial intelligence is powering 55% of reported cybercrime in Africa, allowing criminals to launch faster and more convincing attacks. Cybercrime-related financial losses have risen sharply to $484 million, with online scams and AI-powered fraud being the most common forms.

Gwern reties from fulltime writing to launch Guardian Angel Inc

54 mattsterett 2026-08-04 22:30:07 UTC twitter.com/i Full Share 20

We're unable to show this content The content may be private, deleted or only available on the app.

Mistral's Shieldstral: 3B open-weights model for multimodal moderation

277 riadsila 2026-08-04 17:30:07 UTC mistral.ai Llama Share 69

Shieldstral is a 3B multimodal safety classifier that adapts to context with a policy-adaptive approach, answering questions like content safety and harm. It matches larger models on text safety and sets a new state of the art on multimodal moderation with a single interface for text and images.

DuckDB – Data power tools for your laptop, now in Clojure (2023)

16 sourdecor 2026-08-04 22:30:07 UTC techascent.com Llama Share 0

TMD, a column-major data processing platform, can leverage DuckDB's vectorized SQL execution engine for efficient large dataset processing. The integration enables functional data science solutions on laptops, reducing the need for expensive distributed solutions.

Show HN: Simple algorithm and color space to generate diverse skin tones

441 automatoney 2026-08-04 15:30:07 UTC toneyalexander.github.io Llama Share 87

A custom color picker based on a color space was created to represent human skin tones digitally, aiming to identify a broad and inclusive range of colors that correspond to plausible skin tones. The color space was developed using principal component analysis and manual fitting of spherical equations to a dataset of labeled skin tones, resulting in a color picker with three independent ...

In Memory of My Wife, Elise Cawley, with Thanks for 36 Wonderful Years

730 jdcampolargo 2026-08-04 19:30:07 UTC writings.stephenwolfram.com Llama Share 43

Elise Cawley, a brilliant mathematician, was a loving wife and mother who lit up her husband's life with her intellect, love, and energy for 36 years. She was a forceful intellect, kind and sweet, with a deep love for truth, beauty, and her family, and her legacy will be remembered through her remarkable life and work.

Waymo in Dallas

223 xnx 2026-08-04 19:30:07 UTC waymo.com Llama Share 271

Waymo is now offering fully autonomous rides to everyone in Dallas through its app. The service aims to provide safe and reliable transportation options for locals and tourists alike.

We finally learned to center a div, then browsers added sidebars

11 seg6 2026-08-04 22:30:07 UTC seg6.space Llama Share 6

The user needed to center a div on a webpage with a sidebar, but the div was offset due to the sidebar's width. They used JavaScript and pointer events to calculate the correct position, creating a tool called center to automatically center elements on any webpage.

libexpat now funded by the City of Munich for up to 6 months

3 spyc 2026-08-04 23:20:07 UTC blog.hartwork.org Prefix Share 0

libexpat is a fast streaming XML parser. Alongside libxml2, Expat is one of the most widely used software libre XML parsers written in C, specifically C99. It is cross-platform and licensed under the MIT license. Starting 2026-08-01, the "security vacation" of the project has ended and(!) I will be be paid to work on maintaining libexpat for up to 6 months thanks to the City of Munich ...

Video2NAND – Abusing video codecs for great computational power

17 firer 2026-08-04 21:30:07 UTC sharedobject.blog Llama Share 1

We will use the VP8 video codec to simulate combinatorial logic by abusing its prediction mechanisms. We will build up a set of composable "gadgets" to construct arbitrary logic circuits using key-frames and prediction modes.

Third-party cyber evaluations involving OpenAI models

25 glub 2026-08-04 21:30:07 UTC openai.com Full Share 2

DeepSeek V4 Flash on a Single AMD MI300X

361 zhoutong 2026-08-04 10:30:07 UTC github.com/ryanzhou Llama Share 87

This repository contains a production configuration for running DeepSeek-V4-Flash-0731 on AMD MI300X with fixes for FP8 format, MoE routing, and other issues. It achieves 6,988-7,019 tok/s on fresh prompts with a 2,048-token scheduler budget.

Thanks FedEx, This Is Why We Keep Getting Phished (2024)

176 stymaar 2026-08-04 21:30:07 UTC troyhunt.com Llama Share 41

The user received multiple phishing SMSs claiming to be from FedEx, but after verifying the request on the FedEx website, they found that the link in the SMS was tampered with and could be easily manipulated. The user eventually contacted FedEx via their official phone number and was informed that their shipment was indeed subject to duty and taxes, but the amount was different from what was ...

Show HN: Maple-Preview – ternary 20B MoE running at 120 tok/s on a iPhone

20 edwardbzhang 2026-08-04 22:30:07 UTC deepgrove.ai Full Share 5

DeepGrove is an independent AI research lab building efficient language models that run anywhere.

Oxide Computer raises $445M (SEC Form D)

146 depr 2026-08-04 20:30:07 UTC sec.gov Llama Share 58

The SEC limits automated tool requests to prevent equitable access issues and excessive searches. Users can declare their traffic and follow best practices at sec.gov/developer to avoid request limitations.

Hop.earth – OpenStreetMap based car racing game

73 faebi 2026-08-04 18:30:07 UTC hop.earth Full Share 37

Contains Copernicus DEM (COP-DEM-GLO-30), IGN RGE ALTI®, and CNIG LIDAR data © EU/ESA/IGN/CNIG.

Most tech revolutions made work worse for employees

79 jreynar 2026-08-04 15:30:07 UTC thisandthat.chat Llama Share 41

People are adopting AI despite fears of job loss, but it may actually free up time for workers to focus on quality and creativity. AI could lead to a quality boom, similar to desktop publishing, where workers use their skills and taste to produce better work.

Don't stop early: Case-folding source code at memory speed

40 sbulaev 2026-08-04 19:30:07 UTC github.blog Llama Share 9

To make case folding fast, GitHub engineers removed an optimization and implemented a branchless, vectorizable loop that folds ASCII characters in place, achieving speeds of over 45 GiB/s. They also created a compact, 1776-byte table that uses byte-space arithmetic to fold non-ASCII characters without decoding, outperforming other case folding solutions and a naive hash map.

Truemetrics (YC S23) Is Hiring in Berlin – GTM Lead

0 2026-08-04 17:30:07 UTC ycombinator.com Llama Share

truemetrics is a tech-driven company solving last mile delivery issues with smartphone-based mapping technology. They're seeking a commercial leader to drive growth from €1M to €10M+.

Keyv and friends compromised in active Shai-Hulud supply chain attack

226 cimi_ 2026-08-04 11:30:07 UTC aikido.dev Llama Share 114

Attackers compromised the GitHub account of keyv's maintainer, injecting credential-stealing malware into keyv and other caching utilities with 2 billion monthly installs. The malware harvests secrets from the victim's environment and exfiltrates them to a public GitHub repository.

When AI Benchmarks Plateau: A Systematic Study of Benchmark Saturation

70 doppp 2026-08-04 16:30:07 UTC arxiv.org Llama Share 77

Researchers analyzed 60 language model benchmarks and found nearly half exhibit saturation, with rates increasing with age. Expert-curation, not public test data, impacts resilience to saturation.

The physics of Docker build caching

12 piobio 2026-08-04 22:30:07 UTC blacksmith.sh Llama Share 2

To optimize Docker build caching, consider the three types of caches (image store, mount caches, and layer caching) and their lifetimes, invalidation rules, and failure modes, and order your Dockerfile to minimize the impact of changes on the build process. By understanding where the build's win lives (in layers or mounts) and using the right caching strategy (persistent builder state, export ...

Why some people mow a lawn better than others

159 carlos-menezes 2026-08-04 18:30:07 UTC pudding.cool Llama Share 142

People are surprisingly good at solving the lawn mowing problem, coming close to optimal solutions with fewer stops and only 11% less efficient than optimal for 70+ stops. They use heuristics like decomposition and compression to tackle complex lawns, and spend their time thinking strategically at key decision points.

Launch HN: EdotEnv (YC S26) – Quant Trading RL Envs to Teach LLMs Research

26 Mzzzzz 2026-08-04 19:30:07 UTC edotenv.com Llama Share 20

We generate realistic market environments for quant research, where agents use Bash tools to develop trading strategies and models must adapt to changing market regimes. Successful trading requires planning ahead and balancing short and long-term gains, with the best models recognizing new regimes before they become consensus.

Xbox goes down. You can't play games you own on disc

561 surprisetalk 2026-08-04 12:30:07 UTC birchtree.me Llama Share 606

An Xbox outage blocked users from playing both digital and disc-based games. The author prefers PC gaming because it allows for maintaining access to games despite digital issues.

Apple says more ex-employees may have taken confidential data to OpenAI

319 thewebguyd 2026-08-04 16:30:07 UTC techcrunch.com Llama Share 243

Apple is requesting expedited discovery from OpenAI and former employees over alleged theft of intellectual property. Apple claims it has uncovered new evidence of misconduct beyond the original complaint, including meetings and sharing of confidential information.

There Will Come Soft Rains (1950) [pdf]

339 pmg101 2026-08-03 23:30:07 UTC users.wpi.edu Full Share 374

The Warp Agent CLI

92 emschwartz 2026-08-04 17:30:07 UTC warp.dev Llama Share 55

The Warp Agent CLI is a standalone tool that provides a superior terminal experience for developers, with built-in model-routing capabilities and native support for advanced workflows. It offers features like tab completions, natural language detection, and cloud agent orchestration, making it a powerful tool for autonomous development.

Harness engineering for self-improvement

296 tosh 2026-08-04 06:30:07 UTC lilianweng.github.io Llama Share 67

Harness engineering is a crucial component of recursive self-improvement (RSI) in AI, enabling models to improve their own performance and adapt to new tasks through workflow design, context management, and optimization. Recent research has focused on developing automated design of agentic systems, meta-harnesses, and evolutionary search methods to improve harness engineering, with ...

Everything I Know (1975)

127 simonebrunozzi 2026-08-04 12:30:07 UTC bfi.org Llama Share 31

Buckminster Fuller's 42-hour lectures cover his inventions, discoveries, and design approach, spanning topics from architecture to economics. The lectures are now available as a transcript, inviting public participation in editing and refining the content.