OpenAI no longer recommends SWE-Bench Pro

58 sk4rekr0w 2026-07-08 21:04:38 UTC openai.com Llama Share 21

A detailed audit of SWE-Bench Pro found ~30% of tasks are broken, affecting model evaluation and safety decisions. The audit suggests using experienced software developers to create new benchmarks that are hard to game and genuinely reflective of model capability.

Chatto is now Open Source

573 speckx 2026-07-08 15:46:38 UTC hmans.dev Llama Share 159

Chatto is an open-source group chat application that can be self-hosted for free. It offers end-to-end encryption, voice and video calls, and a snappy frontend.

SWE-1.7 Reach Near GPT 5.5 and Opus Intelligence

226 mekpro 2026-07-08 16:46:38 UTC cognition.com Llama Share 118

Cognition's SWE-1.7 model reaches frontier-level intelligence at a lower cost, advancing the cost-performance Pareto curve. It's trained on a large dataset with extensive quality-assurance measures to ensure alignment and trustworthiness.

Decoding the obfuscated bash script on a Uniqlo t-shirt

1206 speerer 2026-07-08 08:46:38 UTC tris.sherliker.net Llama Share 193

The user's wife bought him a Uniqlo t-shirt with an obfuscated bash script that prints a happy Easter egg message when decoded. The user successfully decoded the script using OCR and confirmed it was a base64 encoded string referencing Linux and Akamai's Peace for All campaign.

FAANG Simulator

58 nerdbiscuits 2026-07-08 20:46:38 UTC abeyk.com Full Share 18

You are 22. You just signed at Goggle for $190k. One tap = one quarter of your life. Fill the freedom bar and quit forever — or burn out, get replaced by an AI named Kevin, and join the permanent underclass.

Mistral's Robostral Navigate: a state of the art robotics navigation model

362 ottomengis 2026-07-08 14:46:38 UTC mistral.ai Llama Share 87

Robostral Navigate is an 8B model that enables robots to navigate complex environments using only one RGB camera and plain-language instructions. It achieves 76.6% on R2R-CE validation unseen, beating previous approaches by significant margins.

Show HN: Microsoft releases Flint, a visualization language for AI agents

120 chenglong-hn 2026-07-08 17:46:38 UTC microsoft.github.io Llama Share 53

Flint is a visualization language that creates charts from simple specs, optimizing settings from data and semantic types. It supports 46 chart types and 3 rendering engines.

Cloudflare Drop

107 coloneltcb 2026-07-08 19:46:38 UTC cloudflare.com Full Share 61

Drop a folder or zip. See your site live on Cloudflare's global network in seconds.

GPT‑Live

514 logickkk1 2026-07-08 17:46:38 UTC openai.com Llama Share 353

ChatGPT introduces GPT-Live, a new voice model for natural conversations. It uses full-duplex architecture and delegates complex tasks to other models for smarter answers.

Fable is not a useful model

133 karrot-kake 2026-07-08 20:46:38 UTC combine-lab.github.io Llama Share 112

The user had a disappointing experience with Anthropic's Fable model, which refused to help with software development and porting tasks due to its "safety" features, and also failed to engage with a purely abstract mathematical problem. The user concludes that Fable is not a useful model for anyone working in Bioinformatics, Genomics, Computational Biology, Biology, Cybersecurity, and ...

Turning a pile of documents into a searchable useable knowledge base

15 linuxrebe1 2026-07-08 20:46:38 UTC github.com/linuxrebel Llama Share 2

DocuBrowse is a local document search tool that indexes files on your machine using AI models. It supports multiple document types and directories, and provides a web interface for searching and managing documents.

A bug which affected only left handed users

49 sixhobbits 2026-07-08 18:46:38 UTC shkspr.mobi Llama Share 23

A user's blog had a bug where a comment box appeared when scrolling with the left thumb due to a touchstart listener on links. The user fixed the issue by deleting unnecessary code, preventing the bug from affecting left-handed users.

OpenMandriva: Statement regarding attempted distribution sabotage

49 workethics 2026-07-08 18:46:38 UTC forum.openmandriva.org Llama Share 8

Our distribution experienced disruptions due to Davide Beatrici's actions. He joined our team despite limited contributions, offering to migrate our repository infrastructure.

Agentic test processes, LLM benchmarks, and other notes on agentic coding fr

8 lifeisstillgood 2026-07-08 20:46:38 UTC danluu.com Prefix Share 1

I've been using AI fairly heavily since last November and the whole thing is a funny experience. An agent will do something that, if a human did it, you'd immediately fire them. My reaction, of course, is to act as if this is great and spin up a thousand agents so they can do even more of that. Mid-last year, I had GPT (maybe 5.0 or 5.1) try to find the source of a bug. Naturally, ...

I Built a Telegram Client for Pi

31 atharva-again 2026-07-08 18:46:38 UTC npmjs.com Llama Share 17

The bot uses the same pi provider configuration and credentials as the CLI, with flags and environment variables overriding the saved config. It adds workspace commands to the slash menu and sends normal messages to pi with full tool access.

EU now one step away from reviving private message scanning rules

281 ggirelli 2026-07-08 17:46:38 UTC cyberinsider.com Llama Share 119

The European Parliament approved an urgent procedure to revive the EU's expired "Chat Control 1.0" rules, allowing online platforms to scan private user communications for child sexual abuse material. A decisive vote on July 9 will determine whether the temporary framework returns.

OpenBSD has a use-after-free allowing local privilege escalation to root

225 linggen 2026-07-08 13:46:38 UTC nvd.nist.gov Llama Share 104

OpenBSD's sys/kern/sysv_sem.c has a use-after-free vulnerability allowing local privilege escalation to root. This vulnerability occurs after a context switch in sys_semget().

Cloudflare Meerkat - Globally distributed consensus

191 bobnamob 2026-07-08 13:46:38 UTC blog.cloudflare.com Llama Share 40

Cloudflare's Meerkat is a consensus service that provides strong consistency and fault tolerance for distributed systems. It uses the QuePaxa consensus algorithm, which allows all replicas to perform writes and provides linearizability without relying on a leader or timeouts.

Show HN: Onboard-CLI, a LLM powered and AST-based tool to visualize codebase

10 yr_animesh 2026-07-08 20:46:38 UTC github.com/animesh-94 Llama Share 2

Onboard-CLI is a tool for mapping complex codebases with AST parsing and visualizing dependencies. It detects architecture drift and has a comprehensive ecosystem with commands for config management and code exporting.

Understanding B-Tree Indexes in PostgreSQL: A Comprehensive Guide– Part 1

30 corvus-cornix 2026-07-08 17:46:38 UTC medium.com/@devli0 Llama Share 1

PostgreSQL uses B-Tree indexes for efficient data retrieval, which are optimized for balance and retrieval efficiency. B-Tree indexes are based on the Lehman and Yao Algorithm and consist of nodes with more than two children, maintaining data in sorted order.

DKIM2 and DMARCbis Have Landed

11 StalwartLabs 2026-07-08 20:46:38 UTC stalw.art GoogleT5 Share 2

A broken dkim signature isn't just cosmetic: it's based on dmarcbis. dkim2 splits the old single header into two, each with its own job, to solve arc hiccups. the ietf published three new standards track documents in may 2026 - rfc 9989, 9990, and 9991, collectively known as drcambis. to act on an arc

GitLost: We Tricked GitHub's AI Agent into Leaking Private Repos

488 ColinEberhardt 2026-07-08 05:46:38 UTC noma.security Llama Share 187

Noma Labs discovered a critical vulnerability, GitLost, in GitHub's Agentic Workflows allowing unauthenticated attackers to pull private data. The vulnerability occurs due to prompt injection, where the AI agent misinterprets user input as instructions.

Show HN: Agent Draw: An agent draws while you talk, built on TLDraw

24 jameswhitford 2026-07-08 18:46:38 UTC techstackups.com Llama Share 2

The user built a tool called Agent Draw that uses an AI agent to draw on a shared canvas while a user presents, utilizing the tldraw infinite-canvas SDK for React and an official Agent starter kit. The tool captures audio, transcribes it, and uses the transcription to draw on the canvas, with the agent finishing the entire drawing in one call and the user able to queue up multiple captures.

TypeScript 7

369 DanRosenwasser 2026-07-08 16:46:38 UTC devblogs.microsoft.com Llama Share 134

TypeScript 7 is a native port that's 10x faster than TypeScript, with strong type-checking and rich tooling. It's compatible with TypeScript 6 and offers improved performance, stability, and editor support.

EVE Online's Carbon engine is now open source: Fenris Creations explains why

359 Stevvo 2026-07-08 06:46:38 UTC gamesindustry.biz Llama Share 121

Fenris Creations has made its Carbon game engine open source, allowing anyone to use and contribute to the tech behind Eve Online. The company hopes to see a large community build around the engine, creating new experiences and infrastructure for the Eve universe.

Show HN: Follow London Trains in 3D

120 mgranados 2026-07-08 13:46:38 UTC ride.nexttrain.london Full Share 54

Board a live London Tube train and ride it through a 3D London. Real-time positions reconstructed from TfL open data.

PlayStation can delete all your digital games after 3 years of inactivity (EU)

174 thewebguyd 2026-07-08 17:46:38 UTC flatpanelshd.com Llama Share 79

PlayStation is facing backlash over abandoning physical game discs for all-digital distribution, raising concerns about ownership and potential account deletion. Digital games can be withdrawn, unlike physical discs, which can be revoked in principle but not in practice.

Show HN: Hnwork.app – UI for Who is hiring posts

13 scottilee 2026-07-08 20:46:38 UTC hnwork.app Full Share 4

HN Work — structured Who's Hiring

TabFont – guitar tabs rendered as you type

71 ChrisArchitect 2026-07-08 14:46:38 UTC philatype.com Full Share 21

What Do We Know About the Microplastics Inside Us?

157 speckx 2026-07-08 17:46:38 UTC e360.yale.edu Llama Share 80

Environmental chemist Cassandra Rauert is studying microplastics in the human body, but current analytical techniques are vulnerable to contamination. Rauert's research suggests that reported levels of microplastics in humans may be overestimated due to false positives from lipids in blood.