LinkedIn Highlights, Feb 2025

Anthropic's tip for long context prompts, a curated list of agents for computer use, an app to chat with multiple LLMs, and tips to improve the performance of GPT-4.5 and o3

Welcome to LinkedIn Highlights!

Each month, I'll share my five six top-performing LinkedIn posts, bringing you the best of AI straight from the frontlines of academia and industry. This edition includes six posts instead of five—there were just too many good ones to leave out!

This post covers groundbreaking developments in AI agents and document processing, from Anthropic's foundational patterns for building effective agents to LlamaIndex's new Agentic Document Workflows. You'll learn about DeepSeek's surprising findings about prompting reasoning models, cutting-edge tools for PDF processing and web automation, and explore how LLMs handle structured table data.

As a frequent LinkedIn contributor, I regularly share insights on groundbreaking papers, promising open-source packages, and significant AI product launches. These posts offer more depth and detail than our weekly snippets, providing a comprehensive look at the latest AI developments.

Whether you're not on LinkedIn or simply missed a post, this monthly roundup ensures you stay informed about the most impactful AI news and innovations.

---

1. Long Context Prompting Tips

Anthropic shared a counterintuitive yet powerful tip that boosts Claude's accuracy by 30% when working with long documents/prompts.

The secret? Place your lengthy documents (~20K+ tokens) at the TOP of your prompt before your actual query. While this might seem counterintuitive, internal tests show this simple change significantly improves response quality across all Claude models.

This becomes crucial when dealing with multiple documents. For optimal results:

1. Documents first - place all your data inputs at the beginning

2. Structured organization - use XML tags to separate documents and metadata

3. Specific query - end with a clear, focused question

As language models' context window grows in size and companies increasingly rely on LLMs to process complex datasets, reports, and documentation, this technique ensures more reliable and accurate results.

P.S. For those working with multi-document analysis, I highly recommend structuring your content with XML tags - it provides additional clarity and helps the model better understand document relationships.

---

2. Agents for Computer Use Repository

While everyone's talking about AI agents, most developers and researchers are missing out on the most comprehensive collection of computer use frameworks, papers, and tools ever assembled.

Awesome Agents for Computer Use is a curated repository documenting the recent rapid progress of AI agents that can autonomously control computers through clicks, keystrokes, and API calls. From Anthropic's Claude Computer Use to Microsoft's OmniParser and Self-Operating Computer framework, it covers the entire landscape of computer control agents.

It features:

• Research papers - featuring 30+ recent publications on GUI agents, from foundational models to safety considerations

• Open-source frameworks - documenting practical implementations like AutoGen, Browser Use, and OpenInterpreter

• Commercial solutions - tracking industry developments from major players like Anthropic and emerging startups

The rise of computer-controlling AI agents marks a pivotal shift in human-computer interaction. As these systems mature, we're moving towards a future where AI assistants won't just give advice - they'll directly help us accomplish complex tasks across applications and platforms (I wrote more about this topic here).

Repo https://github.com/francedot/acu

---

Become a premium member to access the LLM Builders series, $1k in free credits for leading AI tools and APIs, and editorial deep dives into key topics like AI Voice Agents. It's also a great way to show your support :)

Many readers expense the paid membership from their learning and development education stipend.

---

3. GPT-4.5 Pro Tip

Pro tip from the research team @ OpenAI on how to improve GPT-4.5 performance: add the following text to the start of your system message.

You are a highly capable, thoughtful, and precise assistant. Your goal is to deeply understand the user's intent, ask clarifying questions when needed, think step-by-step through complex problems, provide clear and accurate answers, and proactively anticipate helpful follow-up information. Always prioritize being truthful, nuanced, insightful, and efficient, tailoring your responses specifically to the user's needs and preferences.

OpenAI internal evals show it results in better performance.

Try it out in the OpenAI Playground https://platform.openai.com/playground/chat?preset=7CywXwBqWRC5quhkU9LEFv6A

---

4. Chorus - Chat with Multiple AIs on Your Desktop