Radek Voltr retweeted
In lieu of flowers, memorial contributions may be made to:
Johns Hopkins Alzheimer’s Disease Research Center Fund
alzresearch.org/giving/
Alzheimer’s Foundation of America Caregiver Resources
alzfdn.org/donate/
Photo Credit: Tommy Garcia
Radek Voltr retweeted
To commemorate my father’s 95th birthday, I want to remind Spock fans that dad’s parents were Ukrainian immigrants which shaped my father in many ways. Four years of Putin’s war has brought death and destruction to our people, our homeland. LLAP. #standwithukraine #ukraine
Radek Voltr retweeted
Tohle je Andrej Babiš. 👉 Spolupracoval s StB pod krycím jménem Bureš. Během své kariéry agenta předal 15 udání. Podle Vrchního soudu v Praze spáchal dva trestné činy. Na podzim by měl Městský soud v Praze pravomocně rozhodnout o způsobu a výši trestu.
👉 Andrej Babiš je populista - názory mění jako na obrtlíku, podle toho co chtějí lidé slyšet. Za tím účelem si neustále měří, co má lidem říkat. Andrej Babiš slibuje nesplnitelné - snížit daně a zároveň zvýšit výdaje státu a ještě ušetřit. Že to není možné? Lidé mu to přesto věří.
👉 Andrej Babiš je oligarcha ve střetu zájmů. Jak zmapoval investigativní novinář @JaroslavKmenta, před rokem 1989 pracoval v podniku zahraničního obchodu Petrimex v Maroku, měl diplomatický pas a vozil si v kufříku ukládat peníze na tajný účet do lichtenštejnské banky Vontobel. To se mu po revoluci hodilo k ovládnutí Agrofertu.
👉 Je to miliardář, který je nadále prokazatelně konečným vlastníkem Agrofertu, který čerpá od státu miliardové dotace z peněz daňových poplatníků - tedy z peněz nás všech. Tomu se říká střet zájmů - tedy konflikt Babišových soukromých zájmů se zájmem veřejným.
👉 Andrej Babiš se nebojí pro politický zisk používat hrubé manipulace, strašení a lži. Je zdokumentováno, že dokonce veřejně křivě přísahal na zdraví svých dětí - že prý neměly s Čapím hnízdem nic společného. Přitom je do této kauzy sám namočil.
👉 Andrej Babiš chce zestátnit veřejnoprávní média, aby nad nimi získal politický vliv. Andrej Babiš se kamarádí s populistickými autoritáři Orbánem a Ficem, kteří tohle už udělali, kteří jdou na ruku Rusku a kteří tvrdě potírají všechny své kritiky. Andrej Babiš je s nimi v jedné evropské frakci PATRIOTI, která rozkládá jednotu a akceschopnost Evropské Unie, což vyhovuje Rusku.
👉 Andrej Babiš chce zrušit úspěšnou českou muniční iniciativu, která pomáhá zadržovat ruská vojska na východě v bezpečné vzdálenosti od Česka. Ale prý vůbec není proruský. Andrej Babiš prohlásil, že kdyby Rusové zaútočili na Polsko, odmítl by Polsku pomoci.
👉 Andrej Babiš vylučuje vládu s jakoukoli prozápadní stranou - a nechává si otevřená vrátka pro Stačilo s SPD, které zpochybňují naše členství v EU i NATO.
Závěr: Andrej Babiš je rizikem pro naši bezpečnost, ekonomiku i demokracii.
👉 Chcete mít tohoto pána v čele naší země? Pokud ne, podepište výzvu neextremistum.cz/r/69 👈, jděte k volbám a šiřte tento příspěvek dál, ať se tyto informace dozví úplně každý.
Radek Voltr retweeted
Microsoft has just released a preview of an MCP server for Azure DevOps. This makes it easy to work on ADO work items right in GitHub Copilot, and I've recently published a workshop on doing just that - sam-cogan.github.io/mcp-work…
Radek Voltr retweeted
👹Ďábel tkví v detailu.
Ruská informační válka🇷🇺 přeřadila na vyšší převodový stupeň. Už se nesnaží jen o antisystémovou konspirační bublinu.
Jejím cílem jste VY.
Stačí jen malá nepozornost v krizové situaci, a nastává chaos.
To je operace "Doppelgänger"🎭.
1/8🧶
Radek Voltr retweeted
vibe coding, where 2 engineers can now create the tech debt of at least 50 engineers
Radek Voltr retweeted
🇪🇺 EU Unemployment in January 2025:
🇨🇿2.6%
🇵🇱2.6%
🇲🇹3.0%
🇸🇮3.3%
🇩🇪3.5%
🇳🇱3.8%
🇧🇬3.9%
🇮🇪4.0%
🇭🇺4.3%
🇭🇷4.5%
🇨🇾5.0%
🇸🇰5.1%
🇦🇹5.5%
🇷🇴5.5%
🇧🇪5.8%
🇪🇺5.8%
🇵🇹6.2%
🇮🇹6.3%
🇱🇺6.4%
🇱🇹6.6%
🇩🇰6.8%
🇱🇻6.9%
🇫🇷7.3%
🇪🇪7.6%
🇫🇮8.7%
🇬🇷8.7%
🇸🇪8.9%
🇪🇸10.4%
(Eurostat)
Radek Voltr retweeted
I've been experimenting with GitHub Copilot instructions and prompt instructions lately, and have found they greatly improve the quality of responses and the code generated. Here's a write-up of my experiences using it with #dotnetmaui:
dev.to/davidortinau/boosting…
Radek Voltr retweeted
Do yourself a favor and do this right now...
Go to VS Code (Insiders) settings and set `github.copilot.chat.codesearch.enabled: true`.
Include #codebase with your prompt and you get ...
* Embeddings-based semantic search
* Text search
* Files and directories awareness
* Git modified files search
* Project structure awareness
This _is_ the feature you're looking for.
Radek Voltr retweeted
5 techniques to fine-tune LLMs, explained visually!
Fine-tuning large language models traditionally involved adjusting billions of parameters, demanding significant computational power and resources.
However, the development of some innovative methods have transformed this process.
Here’s a snapshot of five cutting-edge techniques for finetuning LLMs, each explained visually for easy understanding.
1️⃣ LoRA:
↳ Introduce two low-rank matrices, A and B, to work alongside the weight matrix W.
↳ Adjust these matrices instead of the behemoth W, making updates manageable.
2️⃣ LoRA-FA (Frozen-A):
↳ Takes LoRA a step further by freezing matrix A.
↳ Only matrix B is tweaked, reducing the activation memory needed.
3️⃣ VeRA:
↳ All about efficiency: matrices A and B are fixed and shared across all layers.
↳ Focuses on tiny, trainable scaling vectors in each layer, making it super memory-friendly.
4️⃣ Delta-LoRA:
↳ A twist on LoRA: adds the difference (delta) between products of matrices A and B across training steps to the main weight matrix W.
↳ Offers a dynamic yet controlled approach to parameter updates.
5️⃣ LoRA+:
- An optimized variant of LoRA where matrix B gets a higher learning rate.
This tweak leads to faster and more effective learning.
______
Follow me → @akshay_pachaar ✔️
Everyday, I share insights and tutorials around ML, LLMs, RAG and AI Agents.
Radek Voltr retweeted
Cursor Agent can break your code.
8/10 times it can work against you and mess up the codebase.
This is how I make Cursor code exactly what I need every single time.
A complete handbook:
Radek Voltr retweeted
How to use cursor for large projects.
getstream.io/blog/cursor-ai-…
Radek Voltr retweeted
Found a text-to-SQL tool that actually works!!
(100% open-source with 13k+ stars)
Most text-to-SQL tools have many limitations:
- they use static pre-trained models.
- they provide limited fine-tuning support.
- they can not naturally adapt to user feedback and interactions.
- they have limited compatibility with databases like Snowflake, Redshift, etc.
Vanna is an open-source text-to-SQL tool that addresses each of these limitations.
In contrast to traditional tools, Vanna can be:
- fine-tuned to your database (a big plus).
- connected to Snowflake, Redshift, etc. (a big plus).
- used to train a model on your own database (a big plus).
- connected to Slack to get SQL queries in the Slack interface (a big plus for teams).
Moreover, Vanna continuously adapts to new data. Thus, the more you use it, the better it gets.
The video (taken from Vanna's GitHub repo) demonstrates some of its core features.
As shown, querying a database with a text prompt using Vanna produces four outputs:
- The SQL query.
- Pandas DataFrame.
- AI-suggested data visualization.
- AI-recommended follow-up text prompts.
👉 Over to you: Do you think text-to-SQL is useful for simplifying data science tasks?
____
Find me → @_avichawla
Every day, I share tutorials and insights on DS, ML, LLMs, and RAGs.
Radek Voltr retweeted
Github 👨🔧: Build your own AI friend
This repository provides an open-source project for building an AI chatbot hardware device named XiaoZhi.
It targets individuals interested in learning AI hardware development and applying LLMs to embedded systems.
Detailed documentation and resources for creating a functional AI chatbot using ESP32 microcontrollers and various peripherals.
Supports multiple LLMs, voice recognition for several languages, text-to-speech, and display capabilities, enabling users to build their own customizable AI assistant.
MIT license for free use, including commercial applications.
Radek Voltr retweeted
Github 👨🔧: Awesome-GraphRAG: A curated list of resources (surveys, papers, benchmarks, and opensource projects) on graph-based retrieval-augmented generation.
-------------
→ Provides a categorized list of research papers on GraphRAG, structured according to a comprehensive survey.
→ Covers key areas within GraphRAG including knowledge organization, knowledge retrieval, and knowledge integration techniques.
→ Offers a comparison between traditional RAG and GraphRAG, clarifying the advantages of graph-based approaches.
Radek Voltr retweeted
5 RAG frameworks that go beyond basic RAG.
Build RAG apps with vision, agentic capabilities, reasoning, hybrid search, and more.
100% opensource.
1. VARAG is a vision-first RAG engine that processes entire document pages as images rather than just text for better context.
Radek Voltr retweeted
Why you need to understand 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗥𝗔𝗚 as an AI Engineer?
Simple naive RAG systems are rarely used in real world applications. To provide correct actions to solve the user intent, we are always adding some agency to the RAG system - it is usually just a bit of it.
It is important to 𝗻𝗼𝘁 𝗴𝗲𝘁 𝗹𝗼𝘀𝘁 𝗶𝗻 𝘁𝗵𝗲 𝗯𝘂𝘇𝘇 𝗮𝗻𝗱 𝘁𝗲𝗿𝗺𝗶𝗻𝗼𝗹𝗼𝗴𝘆 and understand that there is 𝗻𝗼 𝘀𝗶𝗻𝗴𝗹𝗲 𝗯𝗹𝘂𝗲𝗽𝗿𝗶𝗻𝘁 to add this agency to your RAG system and you should adapt to your use case. My advice is to think in systems and engineering flows.
Let’s explore some of the moving pieces in Agentic RAG:
𝟭. Analysis of the user query: we pass the original user query to a LLM based Agent for analysis. This is where:
➡️ The original query can be rewritten, sometimes multiple times to create either a single or multiple queries to be passed down the pipeline.
➡️ The agent decides if additional data sources are required to answer the query.
𝟮. If additional data is required, the Retrieval step is triggered. In Agentic RAG case, we could have a single or multiple agents responsible for figuring out what data sources should be tapped into, few examples:
➡️ Real time user data. This is a pretty cool concept as we might have some real time information like current location available for the user.
➡️ Internal documents that a user might be interested in.
➡️ Data available on the web.
➡️ …
𝟯. If there is no need for additional data, we try to compose the answer (or multiple answers or a set of actions) straight via an LLM.
𝟰. The answer gets analyzed, summarized and evaluated for correctness and relevance:
➡️ If the Agent decides that the answer is good enough, it gets returned to the user.
➡️ If the Agent decides that the answer needs improvement, we try to rewrite the user query and repeat the generation loop.
✅ Remember the Reflection pattern from my last Newsletter article? This is exactly that. 👆
The real power of Agentic RAG lies in its ability to perform additional routing pre and post generation, handle multiple distinct data sources for retrieval if it is needed and recover from failures while generating correct answers.
What are your thoughts on Agentic RAG? Let me know in the comments! 👇
#RAG #LLM #AI

































