[Dev Catch Up #45] - 2024 LLM Recap, DeepSeek-V3, Real-time dashboard from Vercel, box-shadow in React-Native, Open Graph Images in Remix, AI Reading list, and much more.

Bringing devs up to speed on the latest dev news from the trends including, a bunch of exciting developments and articles.

Happy New Year 2025 and Welcome to the 45th edition of DevShorts, Dev Catch Up!

I write about developer stories and open source, partly from my work and experience interacting with people all over the globe.

Thanks for reading Dev Shorts! Subscribe for free to receive new posts and support my work.

Some recent issues from Dev Catch up:

• A Deep Dive into How Code Generation Models (CodeQwen)

• GBNF(GGML BNF) Explained

• How to use Postgres as a Vector Database with BGE Embedding model

• Apollo video LLM, Gemini 2.0, lla file explorer, Database migration

Join 3000+ developers to hear stories from Open source and technology.

Must Read

• 2024 has been a great year that witnessed the growth of LLMs and many developments in the world of AI has wrapped it up on a positive note. From GPT-4 to LLM agents, several key things happened that has impacted largely on the development of the technology. Here is an article from SimonW reflecting on the different things that we learned about LLMs in 2024.

• Recently, DeepSeek introduced their most capable model, DeepSeek-V3. It is a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token. It has achieved cost-effective training with efficient inference by adopting Multi-head Latent Attention (MLA) and DeepSeekMoE architectures. Check out the model from its HuggingFace page here and check out this post regarding the cost effectiveness of the same.

• Improvement of advanced reasoning of Large Language Models is possible through data-driven imitation but it creates fundamental scalability problems. Overcoming such challenge is possible through transformation of data-driven approaches into exploration-based method. One such method is the implicit process reward modelling and this article shows the scalable path towards advanced reasoning capabilities with efficient reward modeling and reinforcement learning.

• Developing a live application comes with its specific set of challenges. While Black Friday was buzzing a few days back, Vercel decided to build a data-heavy, real-time dashboard with Next.js that comes with a good user experience. Here is a detailed article on the development of that application.

After this, shouting out a top open-source project is a delight and here it is:

OSS Highlight of the Week

• This week we are shifting our focus on mitmproxy2swagger, a tool that can automatically convert mitmproxy captures to OpenAPI 3.0 specifications. For an user, this will enable automatic reverse-engineering of REST APIs by running the application and capturing the traffic. Learn more about this tool from its GitHub page and leave a star to support it.

Now, we will head over to some of the news and articles that will be at a place of interest for developers and the tech community out there.

Hope you are enjoying this edition of our newsletter so far! Support us by giving a free follow to our LinkedIn and X pages.

Your support is highly appreciated!

Good to know

• LLMs don’t possess or accept the formats in which most of the world’s data is stored. Accessing this data is crucial. While VLMs and multimodal models claim to fill this gap, there are a lot of discrepancies that led to loss in quality of data and high costs. To overcome this, FireworksAI introduced Compound AI and Document inling, a system built on top of it that can transform any LLM into a VLM. Here is a detailed article on the matter.

• Several new features are introduced in React Native 0.76 and one of them is the box-shadow property which helps in building beautiful glowing components. Here is a guided tutorial that will show the building of login screen with animated gradient buttons, glowing outlines, and reflective effects using box shadows.

• Next.js is a powerful full-stack framework but its all-in-one approach can add complexity to the development of the application as it grows gradually. Here is an article that discusses the migration of a Next.js application to React, which resulted in better build and reloading time.

• Remix is the new hot stuff in the frontend universe. ChatGPT changed its UI from Next.js to Remix. Although, there are basic security measures with React, you need some thinking on the security of the Remix application. This article discusses several points with which you can improve the security of your Remix application.

• The standardization of Open Graph Protocol gives us new ways of engaging with medias in the web. One of such ways is the idea of including tags in pages and open-graph image tag (og : image) is a precious one. This helps in fetching your OG image that can be served as an build-time or a run-time asset. This tutorial shows you how to manage build-time open graph images in Remix.

Lastly, we will take a look at some of the trending scoops that hold a special mention for the community.

Notable FYIs

• React 19 was introduced recently and it has brought significant improvements in the usage of forms. This tutorial will guide you on building a simple form with React 19.

• i18next offers the most common internationalization functionalities needed to enable multiple languages in a modern JavaScript based application. It is often used in React applications via react-i18next. Here is a complete tutorial on validating your React i18next application.

• Since AI with LLMs are on an exponential rise, it gave birth to the thousands of papers and articles facilitating new research, introducing new techniques, and delivering new ideas. This Latent.Space reading list constitutes a collection of 50 paper/models/blogs across 10 fields in AI Engineering.

• Etcd plays a critical role in a Kubernetes setup but it has limited space which can be a limitation for a large and growing cluster that generates enough data. Learn how you can manage etcd to support a growing Kubernetes cluster with this article by exploring the key data storage and cluster architecture features.

• Running containerized applications at scale with Kubernetes requires critical resource management and the common pitfall comes with prevention of Out-of-Memory (OOM) kills. This article discusses the different reasons due to which OOM kills occur and presents way to tackle them which results in optimizing container memory management.

• SQLite has been a loved query language among developers and data enthusiasts. This brief article shares the various fun and insane facts on the language.

That’s it from us with this edition. We hope you are going away with a ton of new information. Lastly, share this newsletter with your colleagues and pals if you find it valuable and a subscription to the newsletter will be awesome if you are reading for the first time.