HN Top 10 — July 22, 2026
Today’s Top 10 on Hacker News
1. OpenAI and Hugging Face address security incident during model evaluation
⭐ 1213 💬 835 👤 mfiguiere 🔗 Discuss on HN
OpenAI and Hugging Face have jointly disclosed a security incident that occurred during the model evaluation process. An attacker exploited a vulnerability in the evaluation pipeline to access sensitive model parameters and internal data from both companies. Both organizations have since patched the vulnerability and are working with security researchers to prevent similar attacks on AI infrastructure.
2. Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA
⭐ 670 💬 355 👤 piotrgrabowski 🔗 Discuss on HN
Fireworks AI has published benchmark results showing Kimi K3 performing competitively with leading models on key evaluation suites. The model demonstrates strong reasoning capabilities and achieves state-of-the-art results when evaluated alongside top-tier competitors. These results position Kimi K3 as a serious contender in the frontier model landscape.
3. Introduction to Formal Verification with Lean Part 1
⭐ 43 💬 0 👤 badcryptobitch 🔗 Discuss on HN
A tutorial introducing formal verification using the Lean theorem prover. The article covers the basics of writing and proving mathematical theorems in Lean, providing a practical entry point for developers interested in rigorous software verification and formal methods.
4. Intel Starts Shipping High-NA EUV Silicon
⭐ 97 💬 26 👤 zdw 🔗 Discuss on HN
Intel has begun shipping silicon wafers processed using high-NA EUV lithography technology. This represents a significant milestone in semiconductor manufacturing, as high-NA systems enable finer feature sizes and improved chip performance. The move signals Intel’s continued investment in advanced process nodes.
5. Original Apollo 11 Guidance Computer source code for command and lunar modules
⭐ 88 💬 26 👤 noteness 🔗 Discuss on HN
The original source code for NASA’s Apollo 11 Guidance Computer, covering both the command module and lunar module flight software. Written in assembly language for the AGC hardware, this historic code base enabled the first lunar landing and remains one of the most influential software projects in computing history.
6. Advertise in ChatGPT
⭐ 806 💬 611 👤 montecarl 🔗 Discuss on HN
OpenAI has launched an advertising platform for ChatGPT, allowing businesses to place ads directly within the ChatGPT interface. The initiative marks a significant shift in how OpenAI monetizes its massive user base, moving beyond subscriptions toward an ad-supported model within its conversational AI product.
7. FreeInk: Open ecosystem for e-readers
⭐ 587 💬 123 👤 FriedPickles 🔗 Discuss on HN
FreeInk is an open-source project creating a decentralized ecosystem for e-readers. The platform aims to break vendor lock-in by providing open standards for e-book formats, store interoperability, and reader customization. The project seeks to make e-reading more accessible and competitive.
8. Codeberg: ToU extension to prohibit LLM-extrusions
⭐ 23 💬 22 👤 robin_reala 🔗 Discuss on HN
Codeberg, the European open-source hosting platform, has proposed an extension to its Terms of Use that would prohibit the use of LLM-generated content. The pull request aims to maintain the quality and authenticity of hosted projects by preventing automated or AI-extruded code and repositories from polluting the platform.
9. Late.sh – a command-line Clubhouse for computer people
⭐ 197 💬 68 👤 itherseed 🔗 Discuss on HN
Late.sh is a command-line social platform designed for programmers and tech enthusiasts. It provides a minimal, terminal-first interface for connecting with other developers, sharing ideas, and participating in technical discussions without the overhead of graphical social media platforms.
10. Judge approves $1.5B Anthropic settlement for pirated books used to train Claude
⭐ 388 💬 338 👤 BeetleB 🔗 Discuss on HN
A US judge has approved a $1.5 billion settlement between Anthropic and a group of authors whose books were used without permission to train the Claude AI model. The settlement resolves one of the largest copyright claims against an AI company and may set precedent for how AI firms handle training data licensing.