At first, we alternated between me trying out technical ideas and sending them to DOES team to give feedback and modify the approach, then rinse and repeat. Although they were new to AI, they could intuit how the prompts and settings would affect the outputs. Over time, our
PROMPT ENGINEERING
-
Optimizing AI Prompts: Specify Knowledge Sources Strategically
By
–
The best thing to do is to specify in the prompt the files it needs to be retrieve from his knowledge folder based on the request. You need to be as specific as possible when telling what information to use, and if that information is not enough, ask to perform a web search.
-
Custom GPT for Efficient Coding Task Automation
By
–
One thing that I've been doing often is, every time new documentation or a model is released, I create a very specific custom GPT, which I feed with all the important information to support these defined coding tasks. It has saved me so much time! Here's my Claude Engineer.
-
Multiple LLM Agents Debate to Improve AI Performance
By
–
You could improve AI performance through all sorts of clever techniques… or you could just have more LLM agents try to solve the problem and debate amongst themselves as to which is the right answer. It turns out that adding more agents seems to help all AIs with most problems.
-
Agentic Stock Analysis Tool with Langchain and Claude
By
–
Building an Agentic Stock Analysis Tool with Langchain, OpenBB, and Claude 3 Opus A great article that touches on: Custom Tool Creation
Deploying with LangServe
Prompting strategies Full OSS code! https://
sethhobson.com/2024/03/buildi
ng-an-agentic-stock-analysis-tool-with-langchain-openbb-and-claude-3-opus/
… -
Prompt Injection Security Vulnerability in AI Access Control Systems
By
–
They're entirely untrustworthy for anything like access control, because whoever gets control of a portion of the input tokens effectively controls the output https://
simonwillison.net/2023/Nov/27/pr
ompt-injection-explained/
… -
LLM Corporate Use Cases: Data Extraction Over Safety Trade-offs
By
–
I wonder about that: if you're building public-facing chatbots then sure, you want everything as safe as possible – but from what I've seen the most valuable corporate use-case for LLMs is structured data extraction, where any denials are just a huge nuisance
-
Anton Shifts from Fine-tuning to Prompting Strategy
By
–
Anton going from fine-tuning to prompting is the biggest plot twist in Tech Twitter history
-
Haiku Model Becomes Default Choice in Application Calls
By
–
Nearly every call with this uses Haiku by default 🙂