Must Read
- #ai
- #llm
- #research
- #xai
Kimi K2.7: 1,700 Coding Tasks Boost SWE Agent Performance
Surge AI shares insights from post-training their Kimi K2.7 agent on 1,700 coding tasks, leading to significant performance gains on major benchmarks.
Feed for the day of:
Must Read
- #frameworks
- #solidjs
- #release
- #dx
SolidJS 2.0.0-rc.8: ESM-Only, SSR XSS Fixes & Enhanced DX
SolidJS moves closer to 2.0 with a release candidate featuring ESM-only runtime, critical SSR XSS patches, and significant developer experience improvements.
Must Read
- #python
- #tooling
- #testing
- #dx
`wrapture`: A New Python Library for Monkey Patching & Observability
Discover `wrapture`, Graham Dumpleton's new Python library that combines monkey patching with advanced testing and observability features, including zero-code tracing.
Worth Reading
- #ai
- #dx
- #research
Adapting to AI: An Engineer's Perspective on Change
Simon Willison discusses the initial disheartening feeling when AI agents perform complex coding tasks, and how to reframe one's role in a rapidly evolving software landscape.
Worth Reading
- #ai
- #openai
- #research
- #performance
OpenAI Scales Storage to 1 Billion ChatGPT Users
OpenAI details the evolution of Habitat, their globally distributed storage platform, now serving over a billion ChatGPT users and 22M requests per second.
Worth Reading
- #ai
- #llm
- #dx
- #tooling
OpenRouter Pitfalls: What Developers Need to Know
Simon Willison explores the hidden complexities and potential issues when relying on OpenRouter's automatic model routing, offering solutions for developers.
Worth Reading
- #python
- #dx
- #release
Python 3.15 Soft-Deprecates `re.match()` for Clarity
Python 3.15 introduces `re.prefixmatch()` as a clearer alternative to the confusing `re.match()`, which is now soft-deprecated, improving regex usability.
Worth Reading
- #css
- #web-platform
- #ux
- #design
Shrink-Wrap Text with CSS `position-anchor`
Learn how the new CSS `position-anchor` property can be leveraged to create more compact text layouts, especially when combined with `text-wrap: balance` or `pretty`.
- #ai
- #security
- #release
- #tooling
AI-Assisted Security Audit for Datasette Patches
Datasette releases security patches 1.0a39 and 0.65.4. The audit process notably leveraged frontier AI models like Claude and GPT, highlighting a new approach to finding subtle bugs.
- #ai
- #llm
- #xai
- #research
Boris Cherny: AI-Generated Code Needs Higher Standards
Boris Cherny of Anthropic argues that production code written by AI, like Claude, requires a higher bar and extensive guardrails compared to human-written code.
- #ai
- #security
- #open-source
Hugging Face's Message to AI Security Agents
Hugging Face's `security.txt` contains a witty message for AI agents looking for vulnerabilities, directing them to a public CyberGym benchmark instead.
- #ai
- #llm
- #research
- #open-source
Essential Reading: Open-Source AI & Open Models
A comprehensive reading list to help developers get up to speed on open models and their implications in the AI landscape.
- #ux
- #accessibility
- #browser
User Permissions: The Importance of Recovery
This article highlights that users often don't know how to recover denied permissions, emphasizing the need for better user education and design around settings.
Feed for the day of: