The post details the OWASP Top 10 vulnerabilities for LLMs and AI agents, highlighting security challenges and offering mitigation strategies.
About:
Alex Ewerlöf is the author of a Substack publication focused on Technical Leadership, Reliability Engineering, and Growth, with tens of thousands of subscribers.
Website:
Specializations:
Incoming Links:
Subscribe to RSS:
Four techniques—RAG, SKILL, MCP, and RLM—enhance generalist LLMs by enabling them to perform specialized tasks with greater reliability and efficiency.
Advanced multi-agent architectures can improve the reliability of LLMs by mimicking human organizational strategies and emphasizing validation and feedback mechanisms.
The post discusses the concept of emergent properties in systems, particularly in the context of large language models (LLMs). It contrasts reductionism with emergence, explaining how emergent properties arise from the interaction...
The article introduces the concept of Service Level Indicators (SLIs) in reliability engineering, emphasizing their importance in quantifying service levels from a consumer's perspective. It presents the SLI Compass, a 2D model to...
6.4AI firewall
2026-03-15 • security chatbot firewall data privacy and protection ai
An AI Firewall is essential for securing AI applications against threats like prompt injections and data leakage while balancing performance and user experience.
A framework for assessing AI fluency levels among knowledge workers is proposed, highlighting the importance of AI literacy for career growth and effective team collaboration.
The post presents 30 AI Systems Engineering patterns, illustrating how traditional software engineering principles apply to the evolving AI landscape and emphasizing the need for knowledge workers to adapt.
The post discusses the introduction of the Prompt API, which simplifies running small language models (SLMs) directly in web browsers. It highlights the Local Browser AI extension that showcases this capability, detailing its feat...
The article discusses the importance of measuring service levels through Service Level Indicators (SLIs) and outlines distinct stages in implementing SLIs. It emphasizes the need for understanding consumer perspectives and the bus...
Understanding the differences between Base Models and Instruct Models is crucial for engineering leaders to effectively develop AI tools tailored to specific needs.
5.9AI topology
2025-10-24 • open source character ai ai technologies ai
AI applications are categorized into Cloud, Edge, and Local AI, each with distinct advantages and challenges regarding latency, privacy, and hardware requirements.
The post discusses the importance of system performance in relation to how parts fit together rather than just their individual qualities. It critiques the common practice of hiring for 'culture fit' instead of 'culture add' and w...
5.63 levels of writing
2025-11-21 • writing clarity communication ai
The author discusses the importance of writing as a means of thinking and clarity, outlining a three-level skill system for writing: Level 1 (Thinking), where writing is used to organize thoughts; Level 2 (Communication), which fo...
The post discusses the Service Level Assessment (SLA) and its visual representation through a graph that categorizes service providers and consumers, their tasks, and the failures that may occur. It emphasizes the importance of un...
The post discusses the Service Level Assessment (SLA) and its visual representation through a graph that categorizes service providers, consumers, tasks, failures, and metrics. It emphasizes the importance of understanding consume...
0SLA vs SLO
2025-06-21 • reliability service level agreement slo
The text discusses the difference between SLA and SLO, emphasizing the importance of using the correct term when communicating reliability. It provides a detailed explanation of SLA and SLO, their relationship, and the consequence...
The text discusses the use of Wardley Maps and Pace Layering as practical tools for senior tech leads and engineering leaders to connect technical landscape to the business model, make smarter bets on technological investments, an...
The text discusses the concept of ephemeral task forces (ETF) as an alternative to technical committees, explaining their purpose, composition, and examples of use. It also covers the author's experience implementing DORA metrics ...
0When a team is too big
2025-05-15 • productivity leadership workplace culture
The text discusses the challenges of managing a large team and the author's experience with breaking a large team into smaller task forces. The author advocates for generalist teams and shares the benefits and drawbacks of this ap...
The text discusses the 4 archetypes for Staff+ roles identified by Will Larson and argues that these archetypes are anti-patterns. The author believes that the most effective Staff Engineers are those who can switch hats as the si...
The text discusses the challenges of self-onboarding for senior technical leaders and provides examples of effective and ineffective questions to ask during the onboarding process.
The article discusses the concept of premature optimization and its risks, providing examples of changing the wrong thing, picking the wrong time, and choosing the wrong trade-offs. It also outlines the 3T's of optimization and ba...
The post introduces a technique to use JavaScript generators for controlling parallelism when mapping a huge array using async functions. It explains the problems with async map functions and how to limit the number of asynchronou...