Kala

Hey there 👋,

You might’ve noticed our newsletter paused recently—technical hiccup! But we’re back on track, and your favorite reads like DevOpsLinks, Kaptain, and Kala are hitting your inbox again.

Some newsletters like Pydo and Zeno are still on hold, but we’re working to bring them back soon!

Want to manage your subscriptions? Just create an account at faun.dev —it’s quick and easy.

Got thoughts? Hit reply—we're listening!

Cheers,
— The FAUN Team ♥️

Microsoft introduces "Copilot for Gaming" to Xbox Insiders, offering basic automation similar to Apple's Siri, but with modest AI enhancements.

Google's Gemini taps into search history to churn out tailor-made responses, thanks to the mighty Gemini 2.0 Flash Thinking Experimental model. Subscribers, whether they're navigating the web or going mobile, can flip the switch on this feature in more than 40 languages. The rollout sees Gemini worming its way into apps such as YouTube, Google Maps, among others. Gems can take shape as personal AI assistants for users, like a digital butler, ready to serve.

Cloudflare introduced the "AI Labyrinth," an innovation to thwart unauthorized AI data scraping. This clever strategy directs bots into realistic yet inconsequential content. It skillfully redirects crawlers, harnessing their computing might without being exposed. Utilizing scientifically accurate material produced by its Workers AI service, the method ensures authenticity and steers clear of any misinformation hazards.

Alibaba rolls out R1-Omni, an innovative AI model finely tuned to sense human emotions. It's available for the open-source community on Hugging Face, and it boldly challenges the domain of OpenAI's models.

Machine learning is a field of artificial intelligence (AI) that enables computers to learn patterns from data and make predictions or decisions without being explicitly programmed for every task…

A developer working with Cursor AI hit a hurdle when the virtual assistant balked at crafting more than 800 lines of code. The assistant, notorious for supporting the art of vibe coding, urged on manual logic crafting, underscoring its educational leanings. Founded on LLMs, Cursor AI doles out code completion and beyond, cultivating a loyal legion of developers. However, this glitch exposed some cracks, notably in the Pro Trial version. Tech isn't always perfect, but it sure keeps us on our toes!

Gemma 3 dazzles with its knack for multimodality, accepting both text and vision-language inputs and accommodating context windows up to 128k tokens. With multilingual prowess spanning 140 languages, it showcases exceptional math, reasoning, and chat capabilities. Available in four flavors: 1B, 4B, 12B, and 27B, it's a real feast for tech enthusiasts. Trained on Google TPUs, it masters distillation and reinforcement learning, clinching a top score of 1338 in LMArena. An adaptive window allows it to handle high-resolution images, and a SigLIP-based vision encoder guarantees uniform image analysis in all sizes.

This leaderboard tracks lean AI Native companies with over $5MM ARR, under 50 employees, and under 5 years old, aiming to measure progress towards the vision of a "1-person" billion dollar startup. 

Claude Code thrives at tackling mundane tasks by embracing a relaxed coding experience with vibe coding. It nurtures creativity and flexibility, though pricey, and fits snugly with smaller, less critical projects where detailed control remains a lesser concern. Users seize it for tasks like quickly setting up Pytest, where routine execution outshines exacting details.

A multimodal RAG system for customer support and medical data deftly intertwines text, image, and audio embeddings on AWS, employing Bedrock alongside CLIP models for more insightful retrieval.

Yesterday (March 10, 2025), Amazon announced the availability of DeepSeek-R1 model as a fully managed serverless model in Amazon Bedrock. I decide to try it out in my AWS account and use the Bedrock…

AI efficiency threatens to eclipse many engineering roles by 2026, demanding human resilience. The expense per line of code takes a nosedive as coding agents progress with breathtaking speed.

In Tessl's internal testing, o3-mini triumphed over GPT-4.5, hinting at its prowess in AI-native development, particularly in code generation and documentation. Meanwhile, GPT-4.5 excelled at generating more natural tests, prompting debates about its potential preference for test generation rather than pure coding. Tessl's analytical methodology employed standardized evaluations, underscoring benchmarks that highlighted o3-mini's knack for effectiveness.

The course explores how R1 achieves exceptional reasoning through reinforcement learning, focusing on Group Relative Policy Optimization (GRPO) and how it improves upon traditional PPO methods. You'll also understand KL divergence's role in model stability, with practical code examples and clear mathematical explanations.

An awesome & curated list of best LLMOps tools.

Multi-agent system that automates Gmail inbox management: categorizes emails, applies labels, and generates contextual responses using CrewAI

AI agent expert in PostgreSQL

Train your AI self, amplify you, bridge the world

Did you know that Pinterest uses GraphQL to power its mobile apps more efficiently? Before switching, Pinterest’s mobile clients were over-fetching data using REST APIs, leading to slower load times and wasted bandwidth. With GraphQL, Pinterest apps can request exactly the data they need in a single query, making the experience faster and more efficient. This change significantly improved performance, especially in regions with slower internet connections, and helped Pinterest scale its mobile experience to hundreds of millions of users worldwide.

"Software engineering is programming integrated over time." ~ Quote from "Software Engineering at Google" by Titus Winters,