Post

https://twitter.com/itsPaulAi/status/1748388577418047572

More from @itsPaulAi

Paul Couvert

@itsPaulAi

Feb 22

Anthropic has released a full guide to building AI Agents

There's both a YT video and a detailed blog post with all the necessary tips.

Everything you need to know below

(And links to resources)

1. Understand the difference between workflows and agents

- Clearly distinguish between predefined workflows and autonomous agents.

- Choose the appropriate approach based on the complexity and variability of the task.

2. Design effective agent prompts

- Agent prompts should be open-ended and provide the LLM with access to tools.

- Provide clear instructions and descriptions of available tools.

→ Use Prompt chaining to decompose a task into a sequence of steps.

Read 12 tweets

Paul Couvert

@itsPaulAi

Feb 21

Copilot “Think Deeper” is quite underrated

You can automate your Excel data processing by using it for free.

With just one prompt I have:

→ Generated the macro
→ Sorted the data
→ Automatically calculated important values

Steps and template below:

1. Activate “Think Deeper"

Open Copilot (copilot .microsoft .com) and click on the Think Deeper icon in the text field.

(You can also use the Copilot in Windows 11).

2. Use this prompt template

Prompt:

You're an expert in data processing and writing Excel macros. Using the sample data below from an xlsx file, write a macro that sorts the rows in descending “Profit” order and calculates the average of the “Sales”, “Gross Sales” and “Profit” columns.

(PASTE THE FIRST 5/6 LINES OF YOUR FILE)

Read 7 tweets

Paul Couvert

@itsPaulAi

Jan 10

Microsoft has released its new open source model Phi-4

Incredibly powerful for its size and you can run it locally on your laptop!

Completely free and available on Mac, Windows, Linux 🧵

1. Download LM Studio

Go to and download the version for your OS. lmstudio.ai

2. Download Phi-4

Open LM Studio and in the search section type “Phi-4”.

Select a GGUF version for Windows/Linux or MLX for macOS.

Tip: if you have a potato laptop (like me), choose the Unsloth version.

Read 4 tweets

Paul Couvert

@itsPaulAi

Jan 2

Gemini can turn any book into an interactive mindmaps

You can do it for books of hundreds of pages thanks to the massive context window and with a single prompt.

Perfect for those who memorize better visually (demo and prompt below):

1. Upload your book

Open Gemini and upload your book in PDF format.

You can also use Google AI Studio by selecting the 1.5 Pro model.

2. Enter the prompt

Copy/paste the prompt from my own conversation:

Gemini will generate code that will allow you to view your mindmap. g.co/gemini/share/6…

Read 4 tweets

Paul Couvert

@itsPaulAi

Dec 22, 2024

Reasoning models are becoming the norm

So a friendly reminder that OpenAI has released a guide to writing prompts for them.

Few important points and example:

The most important thing is to use delimiters, for example XML tags that look like this:

<context>
Your text...
</context>

You should also nest tags, for example: <outer><inner></inner></outer>

A concrete example below:

Prompt example:

I'm building a simple Python calculator application. It's a command-line interface (CLI) application for now, focusing on basic arithmetic operations. I'm using standard Python libraries and haven't implemented any complex error handling or UI yet.

When I perform a simple addition, like '2 + 2', the calculator sometimes returns '5' instead of '4'. This doesn't happen consistently, but it's reproducible after a few calculations. It seems to be more frequent after performing a division operation.

Identify the cause of this incorrect addition result and ensure that basic arithmetic operations, specifically addition, always return the correct value. I need to find the bug in my logic that's causing this intermittent incorrect calculation, possibly after a division.

{{Your file 1 content}}

{{Your file 2 content}}

Read 5 tweets

Paul Couvert

@itsPaulAi

Dec 13, 2024

Google has just launched an AI feature capable of planning, reasoning and searching the web.

All this is done in a single prompt using Deep Research.

This is literally the first reasoning model that has access to the internet. Game-changer.

How to use it below 🧵

1. Select model

Open Gemini Advanced and select “Deep Research” from the list at the top.

Then type your prompt. Example:

“Research all the possibilities for running a small AI model (LLM/SLM) locally on a Chromebook, without an internet connection.”

2. Edit the plan

You can adapt the plan before launching the research.

Just ask Gemini to delete, add, or modify steps. E.g:

“Add a step after the 3rd one: “Find out if it is possible to install an Android app on a Chromebook and run AI models locally within that environment”.