DEV Community: DevLycan

Gemini vs GPT vs Claude vs Kimi: One Prompt, Four 3D Landing Pages

DevLycan — Sun, 09 Aug 2026 17:15:00 +0000

One Prompt, Four AI Coding Agents: What Changed in the Final Design?

I recently ran a practical experiment using Gemini 3.6 Flash, GPT-5.6 Sol, Claude Opus 5, and Kimi K3.

Each agent received the same main task: build a modern interactive 3D landing page for a fictional AI platform called NEXORA using HTML, CSS, Vanilla JavaScript, and Three.js. No design file, UI kit, or additional skill file was provided.

What made the experiment interesting was that the same prompt produced four noticeably different design directions.

Gemini focused on technical depth and product detail. GPT emphasized typography and brand personality. Claude produced a balanced and professional SaaS-style layout, while Kimi focused on clarity and straightforward structure.

This comparison is not intended to be a strict benchmark. The models, effort settings, and execution environments were not identical. Instead, the goal was to observe how different AI coding agents interpret the same requirements and make their own design decisions.

The article includes full-page results, strengths and limitations of each approach, and a Live Demo for every agent.

Read the full comparison:

Devlycan - Technology & Programming Insights

Devlycan - Technology, programming, AI, lifestyle, and future trends—simple insights for the new digital generation.

devlycan.com

Impeccable Skill: Building Web Pages Without AI Slop

DevLycan — Sun, 02 Aug 2026 18:00:00 +0000

AI coding agents can build polished interfaces surprisingly fast—but many of the results still look familiar: purple-blue gradients, equal-sized cards, rounded icons, and typography with little visual hierarchy.

This recurring style is often described as AI slop.

For this experiment, I used the same prompt to build the same web page twice:

First, with a standard AI coding agent.
Second, with the same agent using the Impeccable Skill.

The goal was not simply to compare which version looked better. I wanted to understand whether adding a dedicated design skill could change how the agent makes design decisions.

The difference was noticeable.

Without Impeccable, the page looked complete and functional, but it relied on common AI defaults. With Impeccable, the agent created a clearer hierarchy, used color with more restraint, introduced a more deliberate typography system, and avoided adding decorative content simply to fill the page.

I repeated the experiment across three interfaces:

An editorial magazine landing page.
A project-management dashboard.
A three-tier pricing page.

The article also includes:

A With vs. Without Impeccable comparison.
The prompts used in the experiments.
Installation with npx impeccable install.
How /impeccable init works.
The roles of PRODUCT.md and DESIGN.md.
Useful slash commands for typography, color, layout, motion, and responsive design.
Best practices and limitations worth knowing.

My main takeaway is that Impeccable does not magically turn an AI agent into a designer. Instead, it gives the agent stronger design context, vocabulary, and constraints.

That can be enough to move the result from generic decoration toward a more deliberate and reusable design system.

Read the full article:

Devlycan - Technology & Programming Insights

Devlycan - Technology, programming, AI, lifestyle, and future trends—simple insights for the new digital generation.

devlycan.com

What Is Claude Design? Building Visuals and Interactive Prototypes with AI

DevLycan — Sat, 25 Jul 2026 17:30:00 +0000

AI is moving beyond text and code into visual design and interactive prototyping.

In this article, I explore Claude Design, an AI-powered workspace that helps users turn ideas and natural-language prompts into visual drafts, landing pages, dashboards, presentations, and interactive prototypes.

Unlike traditional static mockup tools, Claude Design can generate runnable front-end experiences using HTML, CSS, and JavaScript, making the result closer to a real product than a simple image or wireframe.

The article covers:

• What Claude Design is and how it works.
• Key features and practical use cases.
• How it differs from a normal AI chat.
• Benefits and current limitations.
• How to write a clearer design prompt.
• Real prompt examples for dashboards and landing pages.
• How Claude Design can fit into a real product workflow.

Claude Design is not a replacement for designers or developers. However, it can make the early stages of ideation, communication, and prototyping faster and easier to refine.

Read the full article here:

Devlycan - Technology & Programming Insights

Devlycan - Technology, programming, AI, lifestyle, and future trends—simple insights for the new digital generation.

devlycan.com

What Is Claude Cowork? When AI Starts Working Like a Teammate.

DevLycan — Tue, 21 Jul 2026 18:00:00 +0000

If you've used Claude Code, you already know how much changes when AI stops suggesting and starts executing. In 2026, Anthropic brought that same agentic architecture to everyday knowledge work with Claude Cowork — no terminal required.

Instead of prompting one command at a time, you define a goal. Claude plans the steps, runs multi-step work, and returns finished results for review.

Here's a practical breakdown:

AI Assistant vs AI Cowork — why responding to commands is fundamentally different from collaborating on a workflow.
Core capabilities — analyzing files, organizing data, and producing documents end to end.
Connected workflows — how it works with tools you already use.
Who benefits most — from developers to content creators and freelancers.
Using it responsibly — where human judgment still matters.

If you're exploring how agentic AI fits into real work, this is a good place to start. Read the full article below.

📖 Available in both English and Thai — read whichever you prefer:

🇬🇧 English:

Devlycan - Technology & Programming Insights

Devlycan - Technology, programming, AI, lifestyle, and future trends—simple insights for the new digital generation.

devlycan.com

🇹🇭 Thai:

Devlycan - Technology & Programming Insights

Devlycan - Technology, programming, AI, lifestyle, and future trends—simple insights for the new digital generation.

devlycan.com

What Is Claude Code? When AI Becomes a Serious Coding Assistant

DevLycan — Sat, 18 Jul 2026 18:30:00 +0000

Most of us have used Claude in a chat window. But Anthropic also ships
Claude Code — an AI coding agent that works from your terminal, reads
your actual project, and edits real files. It's a different way of working,
and I wanted to write down what I learned.

The core difference

When you ask an AI in a chat window, it only sees the code you paste in.
Claude Code reads the whole project structure through the terminal, so it
can tell you which files need to change before you touch anything —
database, controllers, routes, UI, validation, permissions.

What this article covers

What Claude Code can do — explain code, debug across the stack, refactor, add features, and support your Git workflow.
A real example — I ran it against a multi-tenant quotation platform had it analyze the whole structure Real output screenshots included.
Installation and slash commands — /init, /plan, /diff, /context, and advanced ones like /goal.
Who it's for, plus a recommended workflow: analyze → plan → edit → review diff → test → commit.
What to watch out for — commit before letting AI edit, review every diff, never paste API keys.

One thing worth knowing upfront: Claude Code needs a paid plan (Pro, Max,
Team, Enterprise, or Console API). The free Claude plan doesn't include it.

Full article (available in English and Thai):

Devlycan - Technology & Programming Insights

Devlycan - Technology, programming, AI, lifestyle, and future trends—simple insights for the new digital generation.

devlycan.com

What Is Claude AI? Why It Matters for Modern Work

DevLycan — Wed, 15 Jul 2026 18:30:00 +0000

Everyone is asking which AI assistant is "the best." That may be the wrong question. The better one is whether a given tool actually fits the way you work.

This guide takes a clear, no-hype look at Claude AI by Anthropic — what it does, where it fits, and what it costs.

What's covered:

What Claude is — how it handles reading, writing, summarizing, analysis, and coding
The model lineup — Haiku for speed, Sonnet for balanced everyday work, Opus for complex reasoning, and Fable 5 for long-horizon autonomous tasks.
Who benefits most — writers, developers, business teams, and marketers.
Real pricing and usage limits — Free, Pro, and Max, with honest notes on what "limited" actually means.
What to watch out for — handling confidential data, API keys, and verifying output before it reaches production.

For developers, the most relevant section is Claude Code: reading project structures, editing files, debugging, and fitting into an actual development workflow rather than sitting in a separate chat window.

The article does not argue that Claude is the best AI available. It argues that the value of any AI tool depends on how well it matches your workflow — and how deliberately you use it.

Originally published on Devlycan. A Thai version is also available here:

Devlycan - Technology & Programming Insights

Devlycan - Technology, programming, AI, lifestyle, and future trends—simple insights for the new digital generation.

devlycan.com

Stop Letting AI Guess Your Requirements — Try grill-with-docs

DevLycan — Sun, 12 Jul 2026 17:30:00 +0000

Ask an AI coding agent to "build an online appointment booking system," and it will start writing code in seconds—guessing everything you left out along the way.

Can customers book without an account? Can two people book the same staff member at the same time? What even separates a Customer from a Staff member from an Admin? When your prompt doesn't answer these, the agent decides for you—and some of those guesses will be wrong.

In this article, I walk through grill-with-docs, a skill from Matt Pocock's skills repository that interviews you one question at a time before any code gets written. It sharpens vague requirements into precise terms, writes the agreed vocabulary into a CONTEXT.md glossary, and produces a spec you can actually build from.

The article covers:

What grill-with-docs does and how it fits into an agentic workflow.
Installation and when to reach for it (and when not to).
A full step-by-step walkthrough with real prompts and screenshots.
Common mistakes to avoid and its practical limitations.

Read the full walkthrough here 👇

Devlycan - Technology & Programming Insights

Devlycan - Technology, programming, AI, lifestyle, and future trends—simple insights for the new digital generation.

devlycan.com

Design Landing Pages with Google Stitch & Antigravity (MCP)

DevLycan — Thu, 09 Jul 2026 13:13:18 +0000

Designing a landing page with AI is becoming easier. However, turning that design into clean, structured, and usable frontend code is still an important challenge.

I recently updated an article that shares a practical workflow for using Google Stitch together with Antigravity through MCP to design landing pages and convert them into HTML/Tailwind CSS more accurately.

Key topics covered in the article:

• How to define a clear landing page concept before starting the design process
• How to generate UI designs with Google Stitch using prompts
• How to refine the design for real-world use, including CTA placement, layout flow, whitespace, and mobile UX
• How to create a DESIGN.md file to help AI understand the design system
• How to connect Google Stitch with Antigravity through MCP
• How to use the Stitch design as the single source of truth for code generation
• How to write a more structured prompt to reduce inconsistencies between design and code
• What to review before deployment, including responsiveness, semantic HTML, performance, and UX quality

The main idea of this workflow is not only to generate a visually appealing landing page, but also to preserve structure, spacing, visual hierarchy, responsiveness, and consistency with the original design.

The article also includes screenshots, step-by-step guidance, and prompt examples that can be adapted for real development workflows.

I hope this will be useful for developers, frontend engineers, and UX/UI designers who are exploring how AI can support the design-to-code process more effectively.

Read the full article here:

Devlycan - Technology & Programming Insights

Devlycan - Technology, programming, AI, lifestyle, and future trends—simple insights for the new digital generation.

devlycan.com

Build a Todo App with Antigravity 2.0 in 30 Minutes — A Cursor User's Perspective

DevLycan — Sat, 04 Jul 2026 14:33:18 +0000

My Hands-on Experience with Antigravity 2.0: Building a Todo App with Google Sheets as a Backend.

I recently spent some time testing Antigravity 2.0 through a small hands-on project.

Since Cursor has been my primary AI coding environment, I wanted to understand how Antigravity’s workflow feels in a real development scenario. To make the test practical, I built a simple Todo App step by step within approximately 30 minutes, using Google Sheets as the backend.

The goal was not to determine which tool is better. Instead, I wanted to observe how Antigravity approaches development compared to a workflow I am already familiar with.

A few key points stood out during the experiment:

Antigravity follows an agent-first workflow, where the AI plans before writing code.
Antigravity Desktop and Antigravity IDE serve different purposes, which may be confusing at first for users coming from other AI IDEs.
Artifacts are useful for reviewing the AI’s reasoning and implementation plan before moving deeper into the actual code.
Browser Subagent currently has some Windows-specific limitations, which are worth understanding before using it in a browser-based development workflow.

I documented the process based on actual usage, including screenshots, workflow observations, and an initial prompt example that others can adapt for their own experiments.

This article may be useful for developers, AI tool users, or anyone currently evaluating whether Antigravity 2.0 fits their development workflow—especially those coming from Cursor or other AI-assisted IDEs.

The article also includes real usage screenshots and a basic prompt example that can be adapted for your own projects.

Read the full article here:

Devlycan - Technology & Programming Insights

Devlycan - Technology, programming, AI, lifestyle, and future trends—simple insights for the new digital generation.

devlycan.com

The 2026 AI CLI Landscape: Claude Code, Gemini CLI (Antigravity CLI), and OpenClaw

DevLycan — Fri, 03 Jul 2026 00:56:38 +0000

Terminal-based AI agents have evolved considerably over the past few months, and several changes are significant enough that developers relying on these tools should be aware of them.

Most notably, Google has begun retiring Gemini CLI for individual users in favor of Antigravity CLI — a closed-source successor that has drawn some pushback from the community that built out Gemini CLI's open-source ecosystem. Meanwhile, Claude Code has moved to the Opus 4.8 and Fable 5 models with a 1M-token context window, and OpenClaw, the open-source "always-on" agent, has grown into one of the most-starred projects on GitHub — alongside a documented CVE worth knowing about before deployment.

I've just published an updated, fact-checked comparison covering:

What actually changed with Gemini CLI's retirement, and what it means if you have scripts or CI/CD pipelines depending on it
Claude Code's current model lineup, context window, and new Dynamic Workflows feature
OpenClaw's architecture, extensibility via ClawHub, and the security considerations that come with deep system access
A full feature-comparison table (cost, context window, open-source status, setup complexity)
A practical case study walking through how all three tools can work together on a real project

Would be curious to hear which of these you're using day-to-day, and whether the Gemini → Antigravity transition has affected your workflow.

Full article here:

Devlycan - Technology & Programming Insights

Devlycan - Technology, programming, AI, lifestyle, and future trends—simple insights for the new digital generation.

devlycan.com

AI IDE Comparison 2026: Antigravity vs Cursor vs Kiro vs Devin Desktop

DevLycan — Mon, 29 Jun 2026 13:24:16 +0000

The AI IDE Landscape Has Shifted: A 2026 Comparison of Antigravity, Cursor, Kiro, and Devin Desktop.

The first half of 2026 has reshaped the AI IDE space more than any previous period. Major acquisitions, full product rebrands, and a category-wide shift from request-based to credit-based pricing have changed the calculus for developers choosing their tools.

I recently updated my comparison article on Devlycan to reflect these changes, and I'd like to share a brief summary of what's new.

What Changed in the First Half of 2026

March 19, 2026 — Windsurf retired its credit system in favor of daily and weekly quotas. Pro pricing moved from $15 to $20, and a new $200 Max tier was introduced.

May 7, 2026 — AWS Kiro reached general availability, officially replacing Amazon Q Developer as the spec-driven IDE in the AWS ecosystem.

May 19, 2026 — Google launched Antigravity 2.0 at I/O 2026 with a standalone desktop app, CLI, and SDK. The AI Ultra tier was reduced to $100/month.

June 2, 2026 — Cognition rebranded Windsurf to "Devin Desktop" via an over-the-air update. The Cascade agent was replaced by Devin Local, with Cascade reaching end-of-life on July 1.

June 17, 2026 — AWS introduced a Kiro Pro Max tier at $100/month and released a native iOS app at AWS Summit NYC.

Key Takeaways from the Comparison

Entry-level pricing has converged at $20/month across all four tools, but what each plan delivers varies significantly.
Every IDE now uses a credit or quota system, making real cost dependent on which AI model you choose to run.
Proprietary models have become a differentiator — Cursor's Composer 2.5, Devin Desktop's SWE-1.6, and Google's Gemini 3.5 each optimize for different workloads.
Open standards are quietly converging — MCP, ACP, and AGENTS.md adoption is making agents increasingly portable across editors.

Who Should Use Which Tool

Cursor remains the most balanced choice for everyday coding productivity.
Antigravity suits teams exploring multi-agent automation within the Google/Gemini ecosystem.
Kiro fits AWS-centric teams that value spec-driven workflows and compliance traceability.
Devin Desktop excels at large codebases, agent-heavy workflows, and regulated industries requiring HIPAA, FedRAMP, or ITAR compliance.

Read the Full Comparison

The full article covers detailed pricing tables, usage limit breakdowns, supported AI models per tool, pros and cons, performance comparisons, and a complete FAQ section.

Read the full article on Devlycan

Devlycan - Technology & Programming Insights

Devlycan - Technology, programming, AI, lifestyle, and future trends—simple insights for the new digital generation.

devlycan.com

Disclaimer: AI IDE pricing changes frequently. The information reflects publicly available data as of late June 2026. Please verify with each vendor's official pricing page before subscribing.

AI Models & Tools Compared

DevLycan — Thu, 25 Jun 2026 13:21:03 +0000

Selecting an appropriate AI model involves more than simply choosing the most powerful option.

Today, there are many AI models and tools available, including ChatGPT, Claude, Gemini, DeepSeek, Mistral, Lovable, Qwen, Kimi, Grok, and GLM. Each one has different strengths, limitations, and ideal use cases.

If you have ever wondered why one model performs brilliantly for coding but struggles with long documents, or why another excels at summarizing PDFs but feels slow or expensive, this guide will help you connect those dots and make smarter choices.

I wrote a practical comparison guide that looks at these tools from a real-world workflow perspective, including:

Writing and document-related tasks.
Coding and software development.
Multimodal workflows such as images, PDFs, and videos.
App prototyping and internal tools.
Token cost and model selection.
Big model vs small model.
Model routing for better cost, speed, and quality.

The goal is not to decide which AI model is the best overall, but to understand which model or tool fits the task you are working on.

You can read the full article here:

Devlycan - Technology & Programming Insights

Devlycan - Technology, programming, AI, lifestyle, and future trends—simple insights for the new digital generation.

devlycan.com