2. Morten Linderud resigning from Arch Linux development team (lists.archlinux.org)
3. AWS Secrets Manager Terraform: Least-Privilege Access (blog.victorsilva.com.uy)
Can't edit the IAM role? Lock down AWS Secrets Manager with Terraform's aws_secretsmanager_secret_policy for genuine least-privilege access control.
Learn how to enhance security in AWS Secrets Manager by using Terraform to implement a least-privilege access policy, even when you can’t modify the IAM role directly. This approach ensures tighter control over your secrets management.
4. How to Spot AI Writing (www.economist.com)
5. Show HN: Minimal LLM Post-Training Experiments on an 8GB GPU (SFT, DPO, GRPO) (github.com)
Minimal, readable LLM post-training experiments on one 8GB GPU. Measures forgetting, seed variance, and RL emergence. - pochenai/nano-llm-posttraining
7. I Fired My AI Assitant (chreke.com)
Why I decided to fire Claude Code.
A user shares their experience of letting go of an AI assistant, Claude Code, and the reasoning behind the decision. Discover the challenges and frustrations that led to this unusual move.
9. Get ready to flee, Americans in ten countries warned (www.dailymail.com)
The US Embassy has urgently advised US citizens across the Middle East to
The US Embassy has issued an urgent warning for American citizens in ten Middle Eastern countries, advising them to prepare for potential evacuation. This alert highlights escalating tensions in the region and the need for vigilance among expatriates.
10. Anime Professions (github.com)
Which real-world jobs does anime actually show? ~22k titles classified to ISCO-08. Live explorer + dataset. - elmiram/anime-professions
Discover how various real-world professions are depicted in approximately 22,000 anime titles, categorized according to the ISCO-08 classification. The dataset and live explorer provide insights into the representation of jobs in the anime industry.
11. Dead Software Walking: The ongoing evolution of relayd(8) and httpd(8) (rsadowski.de)
Rafael Sadowski
12. Signal Structure of the Starlink Ku-Band Downlink (2023) [pdf] (radionavlab.ae.utexas.edu)
14. Indexing the Data Lake for Online Point Queries (engineering.atspotify.com)
Companies like Spotify need vast quantities of data accessible at low latency for online services and, increasingly, AI Agents acting on behalf of users. Online services — such as portals and personalization features that use or display a user's listening history — need to look up and paginate through per-user data at interactive speeds. AI Agents answering questions like, "what was I listening to last summer?" need to retrieve a user's data quickly so they can reason over it — filtering, aggregating, or even running SQL locally — to build context for LLM prompts. Both patterns share the same underlying primitive: fast point queries by key over datasets that are often far too large to economically keep resident in a key-value (KV) store like Bigtable or DynamoDB.
Spotify and similar companies require efficient data retrieval methods to support personalized online services and AI Agents. This article discusses the challenges and solutions for managing vast data lakes to ensure quick access to user-specific information.
16. Show HN: Copy any website pixel perfect (github.com)
Contribute to HarKro753/claude-copy development by creating an account on GitHub.
18. Show HN: Wisp – a Linux shell with Lua scripting and structured pipelines (github.com)
A Lua-native Linux shell with real job control and structured (not text) pipelines - Hinikaa/wisp
19. Show HN: SmokeVPN – All-in-One WireGuard VPN Hub – Switch Exits in Realtime (smokevpn.com)
SmokeVPN - a logless, centralized VPN command center. One account, many exits, all your devices in one place.
20. Beltrunner: Game Design Postmortem (blog.gingerbeardman.com)
Last week I shared a look at an arcade-faithful version of Asteroids to demonstrate the totally new way of making games that I’ve been working on. It’s calle...
21. Fermi Paradox (en.wikipedia.org)
22. Toast IDE Gets Markdown Spell Checking (github.com)
a batteries-included, beautiful, in-terminal IDE. Contribute to paradise-runner/toast development by creating an account on GitHub.
23. Show HN: Aurora – AI Gateway built in Go (github.com)
The Fastest enterprise AI gateway — route LLM traffic across OpenAI, Anthropic, Gemini, Groq & 30+ providers via a single API. Self-hosted, no vendor lock-in. ( 55x faster than litellm ) - auro...
24. Show HN: Wienerdog – memory and self-improving skills for Claude Code/Codex (github.com)
Give your AI a memory, habits, and a daily routine — with nothing but files. One-line setup for Claude Code and Codex CLI. - wienerdog-ai/wienerdog
25. OpenAI's work on Git for large repositories (openai-git-upstream.openai.chatgpt.site)
OpenAI's work on Git for large repositories.
26. History remembers two Robert McNamaras (indevelopmentmag.com)
evidence and argument for a developing world
The article discusses the contrasting legacies of Robert McNamara, focusing on his influence and decisions regarding the developing world. It presents a nuanced view of his impact on global policies and development strategies.
27. When Internal Memory Fails: A No-Solder Wii U Recovery (smolnero.com)
It was just another hot Sunday at smol HQ. My wife and I were playing games with the family during our biweekly FaceTime game night, and as we were getting...
28. Show HN: Cockpit for you Claude Code agents in Rust (episko.dev)
Run dozens of Claude Code sessions at once. Every agent gets its own terminal, with live status, context, cost and permission prompts in one native app.
A developer created a tool to manage multiple terminal sessions while working with Claude, aiming to streamline project organization. The new solution has already improved workflow for their team, and they’re actively enhancing its features.
29. Assessment of open AI math results (twitter.com)
It's hard for an ordinary person to understand the complexity of these tasks. I'm no mathematician, and I don't see a difference between e.g., results 3 and 10.
So I had GPT-5.6 Sol Pro and Fable 5 Max classify these using @EpochAIResearch OpenMath's rubric:
— "Solid Result": A
So I had GPT-5.6 Sol Pro and Fable 5 Max classify these using @EpochAIResearch OpenMath's rubric:
— "Solid Result": A
An ordinary user evaluates the math capabilities of GPT-5.6 Sol Pro and Fable 5 Max, comparing their results against a specific rubric. The assessment highlights how these AI tools handle complex mathematical tasks, making it accessible for those without a math background.
30. Simon Willison on DeepSeek-V4-Flash-0731 (simonwillison.net)
The latest release in DeepSeek's V4 family, "with substantially enhanced agentic capabilities". It's 304 billion parameters - 167GB on Hugging Face - but it appears to punch well above its …