The Weekly API Cocktail – Issue #4
Each week we carefully handpick the best tech content. Yes, by «hand» we mean a part of an actual human. This week featuring AI Browser War, EU AI Act, productivity research, and (much)more.
What a week!
The AI Browser War • We're starting to see a new battle amongst giants and llm providers over reinventing browsers (bundled with ai, of course...). Perplexity launched Comet, OpenAI announced they'd do the same soon, Google is talking about bringing Gemini in Chrome (in fact, there are already some geminization available in chrome dev tools). I want to emphasize BrowserOS, an open source alternative that may be worth trying.
EU AI Act • The European Union has published a code of practice under the landmark AI Act to increase transparency in AI development. While currently voluntary until August 2026, compliance requires companies like Google, Meta, and OpenAI to publicly track model failures and address copyright concerns without using pirated materials for training (sic). Meanwhile, Cloudflare has announced that it will now block AI crawlers unless site operators explicitly allow them.
Idiocracy 2025 • We're feeling like super-heroes using AI for coding and engineering tasks, but should we? Researchers assessed how AI tools affect the efficiency and output of skilled developers. Also here. Or here.
#tech
The Rise of the Agent Manager: author explores how individuals handle (or struggle) managing multiple AI agents
Assumptions: software, despite being a fixed set of instructions at its core, evolves due to human assumptions about functionality that often change over time
Happy 20th birthday Django!: On July 13th, 2025, Django celebrates its 20th anniversary since the first public commit by Jacob Kaplan-Moss in 2005
#apis
API Architectural Styles of 2025: A nice summary of API options available nowadays; nothing innovative here but can be a good refresher.
#llms
How to Ensure Reliability in LLM Applications: An article by Eivind Kjosbakken discusses strategies for ensuring robustness in LLM implementations.
AI Agent Benchmarks Are Broken: The article argues that current AI agent benchmark tests lack rigor, often failing to capture the true capabilities of advanced systems.
Topic Model Labelling with LLMs: a Python-based method for labeling using the GPT4-o-mini tool
#tools
Some interesting tools I found this week... Some are useful, some are useless... your call!
Whispering: an open-source application designed for fast voice-to-text conversion on Linux systems
Arxiv Explained: short, jargon-free audio explanations covering key aspects of various papers on ArXiv
Dyan: design, test, and deploy REST APIs visually
JSON Schema Builder: an open-source GUI tool designed specifically for creating and managing JSON schemas
Betterlytics: an open-source cookieless web analytics tool
Local Lens: A 100% local development monitoring tool that captures both browser and server logs for LLM analysis.
That's all, folks!