Vibe Coding

News

Gemini 2.5: Our most intelligent AI model

  • Google introduced Gemini 2.5, a state-of-the-art AI model that excels in reasoning and problem-solving. It leads in various benchmarks and is available for advanced users.

Improving Weather Forecasting with AI

  • Researchers at the Alan Turing Institute developed Aardvark, an AI-based weather forecasting model that offers faster and more efficient predictions compared to traditional methods.

Fast-Tracking Drug Discovery for Alzheimer’s with AI

  • The Oxford Drug Discovery Institute is using AI to accelerate drug discovery by analyzing vast amounts of scientific data, identifying potential targets for Alzheimer’s treatment.

Huawei Unveils Industry's First AI Core Network at MWC 2025

  • Huawei launched the first AI-native core network, enabling autonomous optimization and management. This marks a significant shift towards AI-driven infrastructure.

HONOR Plans $10B Investment in Open AI Ecosystem

  • HONOR announced the ALPHA PLAN, aiming to transform into a leading AI device ecosystem company by investing $10 billion in developing an open AI ecosystem.

NVIDIA Advances AI in Healthcare at GTC 2025

  • NVIDIA highlighted several AI advancements in healthcare, including the introduction of the DGX Spark AI supercomputer and the MONAI multimodal framework.

Pluralis Research Secures $7.6M for Decentralized AI Development

  • Pluralis Research raised funding to develop "Protocol Learning," a decentralized approach to AI model training that avoids centralized control.

Mira Network Launches Public Testnet for Decentralized AI Verification

  • Mira Network launched its public testnet, allowing developers to explore blockchain-based AI verification. The network processes 2 billion tokens daily.

Apple's AI Investment Plan

  • Apple announced a substantial investment plan in the U.S., including AI initiatives, underscoring its commitment to AI development.

Baidu's Ernie AI Model Updates

  • Baidu launched new versions of its Ernie AI model, enhancing its capabilities in natural language processing

AI For Good

India, which accounts for over a quarter of the world’s tuberculosis (TB) cases, is leveraging AI-driven technology in its effort to eliminate the disease by 2025. A key innovation in this initiative is DeepCXR, an AI tool developed by the Institute for Plasma Research, designed to detect early signs of TB in chest X-rays.

To enhance screening efforts, the Ministry of Health has integrated this AI software with handheld X-ray devices as part of a 100-day intensive campaign to combat TB. These portable devices improve accessibility and facilitate early detection, particularly in high-risk populations.

“AI tools are expected to be a game-changer in detecting presumptive TB patients and quick initiation of treatment,” said Anupriya Patel, Minister of State for Health and Family Welfare.

Despite advancements driven by AI and other technological interventions, experts have expressed skepticism about achieving TB elimination by 2025. As recently as August, medical professionals described this target as “unachievable,” citing ongoing challenges in reducing infections and mortality rates.

Prompt

Create a symbolic logo for "Intelekto Akimis" inspired by neural waves and brain synapses. Use repeating rhythmic wave patterns resembling an "M" or neuron network. Style: abstract, bold, black and white. Logo should feel dynamic, tech-savvy, and intelligent, reflecting movement, flow of information, and cognitive connection. Minimal details, strong shape, impactful from afar.

Tools I Use

Cudo Compute is a cloud-based service provider that offers high-performance computing, AI, and deep learning solutions.

Dubsado is great for contract writing and project management.

Folk is the number one AI powered CRM tool.

N8N is the most powerful automation tool

Gemini 2.5 Pro: Google's AI Breakthrough

Google Makes a Leap in AI Performance

Google has finally delivered on its AI potential with Gemini 2.5 Pro, representing perhaps the most significant jump in evaluation scores we've seen in recent AI development.

Benchmark Dominance

  • Gemini 2.5 Pro sits 40+ Elo points ahead on the popular ChatBotArena/LM Arena benchmark

  • This represents the second-biggest jump of a top model in LMSYS history, only behind GPT-4 Turbo overtaking Claude 1

  • Scored an impressive 18.8 on Humanity's Last Exam without search or tools

  • Currently topping independent evaluations including the Scale Leaderboard

Competitive Advantages

  • Maintains the longest context length in the industry

  • Demonstrates very strong multimodal performance, including audio capabilities

  • Google achieved this through "a significantly enhanced base model with improved post-training"

Reasoning Models

Gemini 2.5 Pro firmly establishes itself as a reasoning model, with Google stating they're "building these thinking capabilities directly into all of our models." This approach allows the model to handle more complex problems and support more capable, context-aware agents.

The industry is now seeing a spectrum of reasoning capabilities across different models:

  • Pure reasoning models (like DeepSeek R1)

  • Hybrid reasoners (like the new DeepSeek V3)

  • Models with reasoning toggles (Claude 3.7, Grok 3)

Gemini 2.5 Pro takes what the article calls a "brute force approach" — utilizing a large, intelligent model tuned to use significant reasoning tokens. The impressive performance suggests this approach is paying off, potentially setting a new direction for the industry.

While we await API pricing details to better understand its positioning against competitors like GPT-4.5, Gemini 2.5 Pro clearly represents Google's strongest showing yet in the competitive AI landscape.

Newsletters I like

What Is Vibe Coding?

Vibe coding is an approach to writing code that prioritizes flow, readability, and personal style over rigid conventions. It’s not about following strict best practices or optimizing for performance at all costs—it’s about making your code feel good to write and satisfying to read.

Some programmers approach coding as a purely technical exercise, where efficiency and structure are the top priorities. Vibe coders, on the other hand, treat code like an extension of their thought process. They shape it in a way that feels intuitive, expressive, and sometimes even poetic. It’s about crafting code that resonates with the way you think rather than forcing yourself into standardized patterns.

This isn’t to say that vibe coding is about being sloppy or avoiding best practices entirely. Instead, it’s about bending the rules where it makes sense, using creative formatting, expressive variable names, and unconventional structures to enhance the way the code feels. Whether that means writing functions that read like natural language, using whitespace in artistic ways, or adding humorous comments to make future debugging sessions more enjoyable, vibe coding is about bringing personality into programming.

It’s also a response to the sometimes rigid nature of software development. While strict best practices make sense for large teams managing complex systems, many programmers work on projects that don’t require that level of rigidity. Whether you’re working on a personal script, an experimental project, or a creative coding piece, vibe coding offers a way to make the process feel more natural and enjoyable.

Core Principles

At this point there’s no strict definition of vibe coding, but there are a few guiding principles that define the approach:

Expressiveness

Vibe coding is a way of making code feel like an extension of your thinking. It’s about expressing ideas in a way that makes sense to you rather than conforming to arbitrary style guides. Instead of naming a variable tempVar1, a vibe coder might call it spicyNugget—not because it’s conventional, but because it makes the coding process more enjoyable.

Readability (On Your Terms)

Traditional readability is about making sure anyone can understand the code at a glance. Vibe readability is about making the code feel readable to the people who understand the vibe. It doesn’t mean writing cryptic or unnecessarily complex code—it means structuring code in a way that flows naturally, even if that means breaking from convention.

Aesthetic Flow

Code has a rhythm. Some people write it in a highly structured, uniform way, while others use indentation, spacing, and line breaks to create something that feels more organic. Vibe coding often embraces unconventional formatting, using extra spaces, blank lines, or even unconventional alignments to shape the way the code is visually experienced.

Minimalism vs. Maximalism

Some vibe coders love to write concise, poetic one-liners that express an idea in the fewest possible characters. Others enjoy writing long, narrative-style code blocks that tell a story. Both are valid approaches—what matters is that the code feels right for the context.

Where Vibe Coding Shows Up

Creative Coding

In fields like generative art, algorithmic music, and interactive design, the structure of the code often mirrors the aesthetic of the final output. Creative coders frequently use vibe coding to make their code feel as organic and expressive as the work it produces.

For example, in generative art, it’s common to see variables with poetic names like sunsetHues or melodySeed, reflecting the artistic intent of the code. The structure of loops and functions might also be arranged in ways that mimic the natural patterns they generate.

AI-Assisted Coding

Tools like GitHub Copilot, Jan AI, and AnythingLLM have introduced a new element to vibe coding. Since these tools generate code based on patterns in existing data, they can often produce solutions that feel surprising or unconventional.

Some vibe coders lean into this, using AI-generated code as inspiration for new ways to structure their own work. The randomness of AI suggestions can add a sense of spontaneity to the process, pushing programmers to experiment with different styles and approaches.

Personal Scripts

When you’re writing code just for yourself, you don’t need to worry about whether it follows industry standards. Vibe coding thrives in personal scripts—automations, small utilities, or experimental projects—because you can write them in a way that makes sense to you and no one else.

For example, someone writing a script to batch-edit images might structure their functions in a way that mimics the sequence of actions they would take manually, rather than following a traditional software engineering approach. This makes the script easier for them to understand and modify later.

Open-Source Projects

Some open-source communities embrace vibe coding as part of their culture. Instead of enforcing strict style guides, they allow contributors to express themselves through their code. This creates repositories where the codebase reflects the personalities and approaches of multiple developers, making the project feel more like a collaborative art piece than a rigid software product.

For instance, some niche open-source projects use fun and whimsical function names, creative indentation styles, or inline jokes to make the development process more engaging. While this might not fly in a corporate environment, it’s part of what makes these projects enjoyable to work on.

Vibe coding is about making code feel good to write. It’s a rejection of the idea that programming always has to be rigid, dry, and technical. Instead, it treats coding as a form of self-expression, where the way code is written is just as important as what it does.

While vibe coding isn’t suitable for every project—especially in large, structured teams—it has a place in creative coding, personal scripts, and experimental projects. It encourages programmers to write code that feels natural to them, shaping their workflows in ways that enhance creativity and enjoyment.

At its core, vibe coding is about embracing the human side of programming. It’s about making code an extension of your thought process, allowing intuition and aesthetics to play a role in how it’s structured. Whether that means playful variable names, unconventional formatting, or simply writing in a way that feels right, vibe coding gives programmers the freedom to make their code their own.

And if nothing else, it makes debugging a little more fun.