How I use AI

This page is a collection of tips I often share with clients, colleagues, and friends to help them achieve better results with AI.

They come from my personal experience with agentic coding (aka “vide coding”), applied AI engineering, and using LLMs for knowledge work.

An introduction to Context Engineering

The quality of input dictactes the quality of output.

Garbage in = garbage out. This summarizes this entire guide: Elevate the quality of your inputs, to get the best outputs.

Context is essential:

• too little = generic output
• too much = lower prompt adherence (see context rot)

Agentic Coding

Read my context engineering tutorial for Cursor.

Minimize context pollution

You want to minimize irrelevant tokens in your context for two reasons:

• keep the model "focused" on relevant information
• save money by reducing token consumption

The best way I know to achieve this at scale for agentic coding is to use tokf.

Progressive disclosure

You want your LLM to have access to right context, but minimize context pollution. Your LLM to only should only have access to context relevant to its current task.

At the codebase level: progressive disclosure of architecture details

The best way to achieve this is to leverage AGENTS.md files. I suggest collocating them with the code as showcased here. In TypeScript, you can create nested folders just to have a dedicated AGENTS.md file for the relevant code. Think of AGENTS.md as “fish eye lens” on your folder: it has high-level view of what's inside the folder, but it should also have peripheric vision of what's related to it.

Read more on Fish Eye by Amelia Wattenberger.

At the tools level: progressive disclosure of instructions

Use only agent skills that you have read. Add them when a need is identified. Make sure their description is relevant. Customize the description if your LLM/harness doesn't pick it up.

Choosing the right tools

The field is moving fast. ChatGPT released around 2 years ago. New models release every couple months. You want to be able to leverage the best tools.

AI chat client

Don't restrict yourself to a single tool/LLM. Try everything. My personal favorite LLM client is LibreChat because it allows to me to have a single app, but be able to interact with any models.

It comes with some quality of life feature compromises, but the added benefits are invaluable. By using different models, you learn what they are good at and not. You're enable to save costs and be more productive.

Models recommendations

Large language models can be smart, fast, or cheap. Choose two.

TL;DR:

• OpenAI 5.x: has the best all-around model selection
• Claude 4.5/4.6: excellent models, terrible price
• Gemini 3: best price/cost compromise
• Kimi 2: most affordable (litte experience with 2.5)

Engineering planning & knowledge work

My recommendations for performance:

• gpt-5.2 (task planning)
• opus-4.5 (good at everything, just expensive)

Affordable, excellent value for price:

• gpt-5.1-high for engineering planing; medium is enough for most knowledge work
• gemini-3-flash best intelligence for value; excellent at small-scope task that requires tools

Cheap, best-value if you can afford extra prompting:

• kimi-2.5

Agentic coding

Opus 4.5+ and Codex 5.2+ are all very good models, but they are expensive. My personal preference is:

• Opus for "complicated" planning (I need the model to still do a lot of research before creating the plan)
• Codex for "long-running" tasks (I want the model to maintain clear goals over a long plan execution)

Cheap, best-value if you have super clear instructions:

• gpt-5.1-codex-mini-high

Get in touch

If you and your team need advice to help get more value out of AI for engineering and knowledge work, let's get in touch.