Christian Prokopp

Co-Founder | Data, Cloud, AI | Experience not Hype

[Christian Prokopp @prokopp]

Three new AI protocols you need to know 👀

AG-UI: Connects users to agents for interactive, real-time AI UIs.
A2A: Lets agents delegate and collaborate, no hardcoded skills.
MCP: Allows agents to find and use tools/APIs without hardcoding schemas.

More: https://t.co/7QoaKbc3w3

[Christian Prokopp @prokopp]

🚀 Giveaway 🚀 1,791,511,444 rows. 56.81 rows/sec. 200+ GB compressed. Amazon US, UK, DE product and bestseller data for 2023.

https://t.co/zwU6zTCv9s

[Christian Prokopp @prokopp]

Check out my latest article: Spring Clean Your Feed in 2 Mins https://t.co/Rqalj4BjI5 via @LinkedIn

[Christian Prokopp @prokopp]

💷 £450, £700, £900, ... potential savings?! Many numbers have been thrown around with the two National Insurance changes between January and April. It is not as complex as you might think. I untangled it in a succinct blog post on Tax Shrink.

☕️ Enjoy.

https://t.co/3oqJckZ05w

[Christian Prokopp @prokopp]

UK budget is out, 2% drop in NI. I made a calculator so you can check what it may mean to your income.

https://t.co/ZM44RkltFQ

[Christian Prokopp @prokopp]

Rumoured -1% NI in UK benefits incomes:
😞 Low income - nope
😕 High income - little
😐 £50k-ish - most
😣 Everyone loses more through inflation

🗜️ https://t.co/Y2j6PZbHR2

💸 Check impact on your personal income tax or optimise salary dividend strategy for Limited company.

[Christian Prokopp @prokopp]

My fun side project, optimising the salary dividend strategy for UK Owner Operators Ltds. They can save
from 22.11% to £11,130.59 🤯 based on three examples.

v0.1.5 dropped. Again 100% free, no ads either.

https://t.co/51wNx9bZvr

Send me ideas for the next release 🙏

[Christian Prokopp @prokopp]

🚀 Improve your writing using Copilot! Draft docs as markdown files & enjoy GPT4-powered autocompletion.

⁉️ Ask questions in-text, get instant answers.

🖌️ Use markdown for auto-styling

💡 Paste info for context-rich, prompt-driven autocompletion.

#WritingHack #Productivity

[Christian Prokopp @prokopp]

#ad 🌳 Sustainable and powerful. Discover the future of cloud computing with 4th Gen Intel® Xeon® Scalable processors. @IntelUK #CloudComputing #IntelXeon
https://t.co/TXFfm3Jucc

[Christian Prokopp @prokopp]

#ad. 💼 Redefine business agility with the 4th Gen Intel® Xeon® processor family. Creative cloud solutions start here. @IntelUK #CloudComputing #IntelXeon

https://t.co/TXFfm3Jucc

[Christian Prokopp @prokopp]

🤯 Realtime Stable Diffusion on a laptop is reality 💻
😮 Latest SDXL Turbo model paired with an M3 Max running on battery 🔋

You have to pause and admire the everyday wonders sometimes.

[Christian Prokopp @prokopp]

🎁 Sharing time. I found an easy way to fine-tune LLMs/ChatGPT
👷 Easily DIY a training dataset with my FAQ prompt
📻 Then fine-tune your model
💸 Best of all, it is not even that expensive

#LLM #ChatGPT #Dataset

https://t.co/KcLNE5N9dL

[Christian Prokopp @prokopp]

#ad. Step into the realm where sci-fi collides with tech reality, driven by the innovation of the 4th Gen Intel® Xeon® Scalable processor. @IntelUK #IntelXeon
https://t.co/pzEWtIZj8C

[Christian Prokopp @prokopp]

#ad. Dive into the evolution of cloud computing and discover how the 4th Gen Intel® Xeon® Scalable processors are revolutionizing AI and ML in the cloud. 🚀 @IntelUK #CloudComputing #IntelXeon
https://t.co/pzEWtIYLj4

[Christian Prokopp @prokopp]

🚪 ⚠️ Are OpenAI's Custom GPTs clever backdoors to content and data?

#OpenAI #ChatGPT #Data #Content

https://t.co/89GZcZRC0I

[Christian Prokopp @prokopp]

@pandabot9000 Indeed. In my case, the API/LLM IO creates more cost than the corpus size. But that depends on the use case.

[Christian Prokopp @prokopp]

This summer, I built a working RAG product prototype for a sizeable potential customer using LLM. I am shutting it down instead of rolling it out. Here is why, how I got there, and what it means for the future.

https://t.co/iZzwYXpXyY

[Christian Prokopp @prokopp]

== Data Machina #221 ==
LLM-RAG Apps Infeasible? Causal Topological Deep Learning. First look at GPT-4V. Visualising MatMul & Attention. AgentOps. Autogen & AI Agents. SapientML. Mistral 7B. MIC V-L Model. AnyMal Any Model.
by @ds_ldn https://t.co/niY9KwHOcE

[Christian Prokopp @prokopp]

@jsensarma Thanks! Hallucinations were unproblematic. Turning down the temperature & giving strict prompt instructions worked.
A bigger problem was semantically similar info but categorically different, e.g. multiple major doc/sw versions. LLM tends to merge them. Hard to make it choose.

[Christian Prokopp @prokopp]

.@UNICEFBD is doing an outstanding job bringing disadvantaged children back into education in Bangladesh. I am proud that Anjuman & Aziz Charitable Trust (AACT) is part of this story.
https://t.co/QZEGlkNbFH