The Promptolian Project, Stanford CS25 Transformers, Mastering Tableau 2026 | Issue 91

A weekly curated update on data science and engineering topics and resources.

This week’s agenda:

• Open Source of the Week — The Promptolian project
• New learning resources — Stanford CS25 Transformers, Docker for DevOps, Pi Agent, PostgreSQL crash course, 9 AI concepts
• Book of the week — Mastering Tableau 2026 by Marleen Meier

Not a Medium subscriber? The newsletter is also available on Substack and LinkedIn.

Open Source of the Week

This week’s focus is on the Promptolian project. Promptolian is an open source project by Maurizio Lospinoso that provides a transparent proxy layer between AI agents and the Anthropic and OpenAI APIs. Rather than rewriting agent logic to deal with context loss and redundant tool-schema transmission, Promptolian sits between the agent and the model API and applies tool caching, conversation-history compression, and prompt optimization automatically. Adding it to an existing agent is a one-line change — swap the client base_url — and the response format from the underlying API remains unchanged.

[embed]GitHub - Maurizio-L/promptolian-public: a powerful token compresson for LLM chat memory and user… a powerful token compresson for LLM chat memory and user prompts - GitHub - Maurizio-L/promptolian-public: a powerful…github.com

Key Features

• Tool schema caching — automatically applies cache_control headers to tool schemas so Anthropic's prompt cache kicks in; per the project, tool tokens drop to roughly 10% of the normal price from the second call onward.
• KV-sandwich history compression — keeps fact-rich turns verbatim while pruning filler content, with the project reporting around a 22% reduction in conversation tokens.
• Three integration modes — a transparent HTTP proxy, a Python SDK wrapper via patch_anthropic(), and a Claude Code MCP server.
• Local and cloud deployment — run locally with one command (promptolian proxy) backed by SQLite, or use the hosted endpoint at proxy.promptolian.com.
• Sensitive data detection — flags items such as connection strings, API keys, private keys, JWTs, env files, SQL dumps, and large CSV/JSON payloads before forwarding.
• Diagnostic response headers — expose proxy behavior through headers like X-Promptolian-Cache-Hit and X-Promptolian-Tokens-Saved.
• Pluggable storage — defaults to SQLite, with PostgreSQL supported via a DATABASE_URL setting.

The Promptolian schema; Image credit: project documentation

More details are available in the project documentation.

License: not specified in the repository.

New Learning Resources

Here are some new learning resources that I came across this week.

Stanford CS25: Transformers

The following Stanford Online playlist compiles the CS25 seminar series on Transformers, including the latest season (V6) with seven new lectures and prior seasons dating back to V1. The V6 lectures cover an overview of transformers, representation learning to world modeling, state space models versus transformers, scaling training across thousands of GPUs, next-token prediction toward next-generation intelligence, generalization from parameters and context, and collaborative AI agents in science.

[embed]

Learn Docker — Full DevOps Course

This full course from freeCodeCamp provides a hands-on, beginner-to-deployment walkthrough of Docker, covering the fundamentals of containerization, building and testing images, and the practical benefits of Docker for reproducibility, dependency management, portability, and version control.

[embed]

Pi Agent Explained

This short video by Caleb Writes Code provides a six-minute overview of the Pi agent, walking through what Pi is, the harness it runs in, the underlying framework, and example use cases.

[embed]

PostgreSQL Crash Course

This beginner tutorial from NeuralNine provides a quick introduction to PostgreSQL, covering environment setup, core terminology, and the most common SQL statements — CREATE TABLE, SELECT, INSERT, the SERIAL data type, and the WHERE clause.

[embed]

9 AI Concepts Explained in 7 Minutes

This short video by ByteByteGo provides a fast-paced primer on nine core AI concepts, including tokenization, text decoding, prompt engineering, AI agents, RAG, RLHF, VAEs, diffusion models, and LoRA.

[embed]

Book of the Week

This week’s focus is on a new data visualization book — Mastering Tableau 2026 (Fifth Edition) by Marleen Meier. The book builds on Tableau fundamentals to help readers design and ship advanced dashboards, run efficient data preparation with Tableau Prep Builder, and extend Tableau with programming languages and the platform’s new AI-powered features. Instead of treating Tableau as a click-only BI tool, it walks through calculated fields, level-of-detail (LOD) expressions, performance tuning, deployment on Tableau Server and Cloud, and integration patterns with Python and R for more sophisticated analytics workflows.

Topics Covered

• Tableau Prep Builder — connecting to data, cleaning, joins, unions, aggregation, pivoting, and flow automation
• Table calculations — partition and direction of addressing, lookup, running, window, and rank functions
• Level of Detail (LOD) calculations — FIXED, INCLUDE, and EXCLUDE expressions and their order-of-operations implications
• Mapping and geospatial analytics — custom polygons, heatmaps, dual-axis maps, Mapbox, and custom geocoding
• Dashboard design — layout patterns (golden ratio, quad, small multiples), color and visualization best practices, and dashboard actions
• Advanced analytics — scattergraph correlations, geospatial analysis with traffic and intersection data, and distance-based measures
• Performance optimization — extracts, indexing, filter strategy, efficient calculations, and the workbook optimizer
• Tableau Server and Cloud — publishing, permissions, web authoring, and what’s new in version 2026.1
• Python and R integration — using TabPy and R for regression, clustering, and sentiment analysis
• Tableau AI, Pulse, and Next — the metrics layer, Tableau Semantics, generative AI in Pulse, and Einstein Discovery
• Data governance — certified data sources, data quality warnings, lineage, and the Data Orchestration Center

This book is ideal for business analysts, BI professionals, and data analysts who want to move past Tableau basics and build scalable, AI-augmented analytics solutions. Prior exposure to Tableau is helpful but not required.

The book is available for purchase on the **publisher’s website and on [Amazon](https://www.amazon.com/Mastering-Tableau-2026-visualizations-techniques/dp/1806100711)**.

Have any questions? Please comment below!

See you next Saturday!

Thanks,

Rami

Thanks for reading Rami’s Data Newsletter! Subscribe for free to receive new posts and support my work.