This post evaluates the performance of six AI models (GPT-5, OpenAI o3, Claude Opus 4.1, Claude Sonnet 4.5, Grok 4, and Gemini 2.5 Pro) in identifying and fixing three advanced security vulnerabilities: prototype pollution, an age...
About:
JP Posma is the author of the Kilo Code blog, a publication on Substack. The blog is associated with the website kilocode.ai.
Website:
Subscribe to RSS:
The post evaluates seven AI models, including Gemini 3 Pro Preview and Grok 4.1, on their ability to create a functional analytics dashboard using sample data. The models were tested on a semi-complicated UI task, with varying res...
The blog post compares two coding agents, Cursor's Plan Mode and Kilo Code's Architect Mode, focusing on their performance in designing a background email notification service. Kilo's Architect Mode scored higher (8.7/10) for addr...
Code Supernova is no longer available on Kilo Code, prompting a search for comparable free or affordable models. Benchmarks show that Grok Code Fast 1 performs similarly to Code Supernova while generating cleaner code. A hybrid ap...
Google has launched Gemini 3, its most advanced AI model, featuring enhanced reasoning, multimodal understanding, and improved agentic behavior. Available through the Kilo Code ecosystem, Gemini 3 outperforms previous models with ...
42Gen is developing an AI assistant focused on managing users' physical environments while prioritizing privacy. The team, led by co-founder David Samuelson, emphasizes on-device reasoning and anonymized queries to avoid sharing p...
The blog post analyzes Augment Code's recent switch from a message-based to a credit-based pricing model, comparing it with Kilo Code's token-based pricing through a series of development tasks. The author conducted tests on four ...
GLM-4.6, developed by Zhipu AI, has rapidly gained traction, recording 15.9 billion tokens in just 12 days, showcasing a significant shift in the AI market. This model operates without NVIDIA chips, utilizing domestic Chinese hard...
Google's newly released Antigravity IDE is revealed to be a proprietary fork of Windsurf, which Google licensed for nearly $2 billion. This post discusses the implications of this 'PORK' (Proprietary Fork) in the tech industry, hi...
Rogerio Chaves, CTO of LangWatch, discusses the challenges of making AI agents reliable and production-ready, noting that over 95% of enterprise agent projects fail to reach production. He introduces Better Agents CLI, a toolkit d...
The post discusses the rapid adoption of AI coding assistants among developers and the challenges it presents for engineering leaders. It emphasizes the need for a collaborative culture to manage AI usage effectively, suggesting p...
The post compares three AI coding models: Claude Haiku 4.5, GLM-4.6, and GPT-5 Mini, focusing on their performance in generating a job queue system in TypeScript with SQLite. The analysis includes metrics such as speed, cost, code...
The blog post discusses the growing use of AI tools in software development, highlighting a decline in developers' trust in AI due to accuracy issues and the concept of 'vibe coding.' It introduces spec-driven development as a sol...
The post discusses the challenges and strategies of integrating AI into software development, highlighting that many developers are skeptical about AI due to unmet requirements and control issues. It emphasizes the importance of l...
Warp has introduced a new pricing plan called 'Build,' which is essentially a rebranding of its previous plans. The new plan charges $20/month for 1,500 credits, but additional 'reload credits' are required for overages, which are...
The post announces the support for the latest version of MiniMax M2, highlighting its interleaved thinking and native tool calling features. It emphasizes that MiniMax M2 remains free for a limited time on Kilo Code and showcases ...
OpenAI has released four new models: GPT-5.1, GPT-5.1 Chat, GPT-5.1-Codex, and GPT-5.1-Codex-Mini, with varying context windows and optimizations for token usage. The models are available through Kilo Code and OpenAI’s API, with i...
This blog post discusses the latest updates in Kilo Code's CLI and Extension, highlighting features such as CLI checkpointing for rolling back conversations, a new Shell Mode for quick command access, and an Auto Cleanup feature f...
This blog post announces the introduction of Parallel Mode for multi-agent workflows in the CLI, allowing concurrent command execution. It highlights the release of MiniMax's new model, M2, which is temporarily free and boasts sig...
Kilo Code now supports Claude Haiku 4.5, Anthropic's latest AI model that matches the coding performance of the previous Sonnet 4 model while being significantly cheaper and faster. Haiku 4.5 is designed for rapid coding tasks suc...
The blog post discusses the frustrations developers face with Cursor's unpredictable pricing model, which includes confusing limits and unexpected costs. It highlights the negative impact on productivity and the unsustainable natu...
The post provides updates for Kilo Code, highlighting new features for Teams and Enterprise Plans, including a Usage Statistics Panel, Zero Data Retention setting for privacy, and a Revert All Changes Button. It also mentions impr...
This week's product roundup highlights the launch of Kilo Code CLI (v0.0.11), featuring new observability tools for Teams & Enterprise, expanded model support, and improved user experience. Key updates include the ability to track...
This blog post provides an overview of the latest updates in the Kilo Code ecosystem, highlighting the full support for OpenAI's newly released GPT-5.1 model family, which includes four new models. It details the enhancements in p...