Claude Fable 5

1120 Philpax 2026-06-09 17:08:19 UTC anthropic.com Llama Share 933

Claude Fable 5 is a state-of-the-art AI model with exceptional performance in various areas, but it comes with risks and requires safeguards to prevent misuse. The model is released with conservative safeguards that trigger in less than 5% of sessions, and a trusted access program will be expanded to allow more users to access its capabilities.

GPT-2: Too Dangerous To Release (2019)

94 AbuAssar 2026-06-09 19:08:19 UTC naokishibuya.github.io Llama Share 19

OpenAI's GPT-2 was deemed too powerful to release due to concerns of misuse, but a smaller model was released for researchers. GPT-2 has 1.5 billion parameters and was trained on 40GB of web texts, achieving state-of-the-art results in various language tasks.

Ultrafast machine learning on FPGAs via Kolmogorov-Arnold Networks

17 ag2718 2026-06-09 19:22:19 UTC aarushgupta.io Llama Share 0

The user's Master's thesis involves designing hardware architectures for ultrafast inference and online learning using Kolmogorov-Arnold Networks (KANs). KANs are a natural fit for efficient, expressive LUT-NNs and can be trained on FPGAs for ultrafast inference and online learning.

Making Graphics Like it's 1993

623 sklopec 2026-06-09 11:08:19 UTC staniks.github.io Llama Share 102

The user is developing a side-project called Catlantean 3D, a traditional raycaster first-person shooter with a 256-color palette, aiming to recreate the feeling of early 90s games while using modern tools. The user has implemented various techniques to achieve this, including asset creation, lighting, and animation, and is planning to release the game on Steam in Q1 2027.

What it feels like to work with Mythos

79 swolpers 2026-06-09 18:08:19 UTC oneusefulthing.org Llama Share 62

The user tested the first public Mythos-class AI model, Claude 5 Fable, and found it to be a significant leap over previous models, capable of complex tasks with minimal human input. Fable's ability to delegate tasks to other models and generate impressive results with little human oversight raises concerns about the black box nature of AI and the potential loss of human control.

A giant star may have destroyed itself in one of the rarest explosions

100 wglb 2026-06-09 16:08:19 UTC phys.org Full Share 14

We’re checking your connection to prevent automated abuse

Microsoft's open source tools were hacked to steal passwords of AI developers

488 raffael_de 2026-06-09 08:08:19 UTC techcrunch.com Llama Share 169

Microsoft's Azure and AI development tools were hacked, allowing hackers to steal users' passwords and sensitive credentials. The company has temporarily removed some repositories, notified affected customers, and is continuing to investigate the breach.

Flat Datacenter Networks at Scale at Amazon

12 tanelpoder 2026-06-09 19:08:19 UTC perspectives.mvdirona.com Llama Share 0

Researchers developed RNG, a flat network design using random graphs, to improve data center efficiency and scalability. RNG uses 69% fewer routers, delivers 33% higher throughput, and cuts network power by 40% compared to traditional fat-tree networks.

Test-case reducers are underappreciated debugging tools

9 ltratt 2026-06-09 19:08:19 UTC tratt.net Prefix Share 0

Test-case reducers are less well known than they should be, and those who are aware of them don’t always realise the variety of ways we can use – perhaps even abuse! – them. In this post, I’m going to explore some of the things I’ve learnt while using these wonderful tools. I’ll start at the basics, because the idea is so simple that it can be hard to believe it works. I’ll then work my way ...

Apple decided not to roll out Siri in EU after denied request for exemption

260 flanged 2026-06-09 17:08:19 UTC reuters.com Full Share 433

Please enable JS and disable any ad blocker

The LD_DEBUG environment variable (2012)

27 tanelpoder 2026-06-09 18:08:19 UTC bnikolic.co.uk Llama Share 1

To diagnose shared library loading problems, set the LD_DEBUG environment variable to dump debug information. This can be used to resolve most loading problems quickly on Linux.

Biff.core: system composition for Clojure web apps

56 jacobobryant 2026-06-09 17:08:19 UTC biffweb.com Llama Share 15

The user has released the first library, biff.core, which provides system composition and interfaces for Biff projects. This library includes init functions that allow for clean extraction of logic and late binding, making it easier to manage project structure and dependencies.

Launch HN: Transload (YC P26) – Measuring freight items with CCTV

23 nils_spatial 2026-06-09 17:08:19 UTC news.ycombinator.com Llama Share 6

Julius, Jago, and Nils are building transload, a system that uses existing security cameras to automatically measure freight dimensions in LTL trucking terminals. Their system has two main steps: associating barcode scans with the correct objects in video footage and estimating the objects' dimensions in real-world units.

FCC wants to kill burner phones by forcing telecoms to get all customers' IDs

285 berlianta 2026-06-09 16:08:19 UTC 404media.co Llama Share 188

The FCC plans to require telecoms to store personal info like ID numbers and addresses for all phone customers, alarming privacy advocates. This change aims to combat scammers but could impact low-income people, domestic abuse survivors, and those seeking anonymity.

OpenCV 5 Is Here: The Biggest Leap in Years for Computer Vision

604 ternaus 2026-06-09 03:08:19 UTC opencv.org Llama Share 108

OpenCV 5 is a major release that modernizes the library with a new DNN engine, improved ONNX support, and hardware acceleration. It brings significant performance improvements, better language support, and a redesigned Hardware Acceleration Layer for faster image processing.

CEOs Who Think AI Replaces Their Employees Are Just Bad CEOs

50 speckx 2026-06-09 19:08:19 UTC techdirt.com Llama Share 7

CEOs are overhyping AI tools, forcing employees to use them without understanding their limitations. This "AI psychosis" leads to unrealistic expectations and potential layoffs, which may be an excuse for poor hiring decisions.

How we run untrusted customer code at scale

7 rguldener 2026-06-09 19:08:19 UTC nango.dev Prefix Share 0

Nango is a code-first platform for building product API integrations. Customers connect their apps to Salesforce, Google Calendar, Slack, and a few hundred other APIs. Much of the code behind those integrations is written by our customers and deployed to us. That code is untrusted, and can try to do anything: fetch an API, transform data, throw an exception, leak memory, or intentionally try ...

Let's Encrypt bans certificate usage in any US sanctioned territory [pdf]

197 piskov 2026-06-09 06:08:19 UTC letsencrypt.org Llama Share 149

You agree to use Let's Encrypt Certificates in compliance with applicable laws and this Agreement, and to immediately request revocation if your Certificate is compromised or inaccurate. ISRG disclaims all warranties and is not liable for any damages or losses related to the use of Let's Encrypt Certificates.

Apple's AI Can Now Change Your Passwords. What Could Possibly Go Wrong?

10 speckx 2026-06-09 19:08:19 UTC kylereddoch.me GoogleT5 Share 4

Apple's password monitoring feature compares saved credentials against a curated list of leaked passwords. samuel linn: agentic password-changing feature should preserve that instinct at every step - and avoid leaving systems halfway through an operation if the user isn't fully under the legitimate user' rsli: the more authority an agent receives, the less room there is for vague promises ...

Is Grep All You Need? How Agent Harnesses Reshape Agentic Search

88 Anon84 2026-06-09 14:08:19 UTC arxiv.org Llama Share 42

Researchers compared retrieval strategies in Large Language Model agents, finding grep generally yields higher accuracy than vector retrieval. However, accuracy still depends on agent architecture and tool-calling paradigm.

Show HN: Cost.dev (YC W21) – making agents cost-aware and cheaper to call

40 akh 2026-06-09 17:08:19 UTC cost.dev Llama Share 22

Infracost helps teams optimize cloud costs by applying FinOps policies in real-time during IaC development. It prevents 95% of new cost issues before deployment and saves up to $312/mo.

Emerge Career (YC S22) Is Hiring a Founding Growth Marketer

0 2026-06-09 12:08:19 UTC ycombinator.com Llama Share

We are looking for a growth marketer to lead student acquisition at Emerge Career, a company that provides second chances to low-income adults through workforce development. The ideal candidate has 5+ years of experience in growth marketing, expertise in at least one core channel, and experience building AI-powered marketing workflows.

Can LLMs Beat Classical Hyperparameter Optimization Algorithms?

73 galsapir 2026-06-09 15:08:19 UTC arxiv.org Llama Share 12

Researchers compared LLM-based and classical hyperparameter optimization methods on a small language model. A hybrid approach, Centaur, combining classical CMA-ES with an LLM achieved the best results, suggesting LLMs are most effective as a complement to classical optimizers.

The iPhone's Last Stand?

139 swolpers 2026-06-09 11:08:19 UTC stratechery.com Llama Share 178

Microsoft's Project Solara showcases a thin client approach where AI agents live in the cloud and interact with devices as portals. Apple's Siri, despite being behind the state of the art, is sufficient for consumers who don't want to work or care about productivity.

Show HN: Gravity – interactive solar-system simulator, from Newton to Einstein

109 qunabu 2026-06-09 12:08:19 UTC qunabu.github.io Full Share 24

Earth texture Solar System Scope · CC BY 4.0

Show HN: GentleOS – A pair of hobby OSes for vintage 32-bit and 16-bit PCs

60 luke8086 2026-06-09 12:08:19 UTC github.com/luke8086 Llama Share 83

GentleOS/32 is a retro OS for tinkering with old hardware, requiring i386 CPU, 4MB RAM, and VGA display. It supports standard PC devices and has a 16-bit spin-off, GentleOS/16, targeting 80186 devices.

Show HN: Learn from 30 historical figures, open source, nonprofit, self-hosted

27 micstradev 2026-06-09 16:08:19 UTC github.com/chipmates Llama Share 11

Agora Cosmica is a Living Library with 30 historical figures, offering free wisdom teachings and live AI conversations. It's a nonprofit, open-source platform with no tracking cookies or profiling, built on learning science and self-hosted speech tooling.

Unified Controllable and Faithful Text-to-CAD Generation with LLMs

48 PaulHoule 2026-06-09 14:08:19 UTC arxiv.org Llama Share 13

Researchers propose PR-CAD, a framework that unifies text-to-CAD generation and editing for controllable and faithful modeling. PR-CAD achieves state-of-the-art results and improves CAD modeling efficiency through a reinforcement learning-enhanced reasoning framework.

Show HN: Transit-format (JSON/MessagePack) reader/writer in C

3 delaguardo 2026-06-09 19:08:19 UTC github.com/dotfox Prefix Share 0

A fast, zero-copy Transit reader and writer written in C11 with SIMD acceleration. One codec-agnostic engine speaks all three Transit wire formats — JSON, JSON-Verbose, and MessagePack — and decodes straight into a single arena with borrowed (never copied) string payloads. Transit is a format and a set of libraries for conveying values between applications written in different languages. It ...

Using Optical Aberrations to Distinguish Real Astronomical Transients

30 solarist 2026-06-09 16:08:19 UTC arxiv.org Llama Share 2

Researchers found that transient images on Palomar sky surveys show coma aberration patterns, ruling out plate artifacts as an explanation. This supports hypotheses that transients have a physical origin rather than being instrumental effects.