Prompt Engineering, formalized Prompt Engineering has felt intangible to me. Hype but no precision In Chapter 2, we define Prompt Engineering as a search problem that maximizes a fitness function on an agent’s trajectory Too abstract?! Too dense?! Don't worry. We build this from the ground up

Engineering without a dependable mental model is just guesswork In Chapter 2, we fuse theory with practice and formalize goals, tasks, trajectories, alignment, and even prompt engineering. Timeless mental models you'll use forever

In Chapter 2 of Systems Engineering for Agentic Applications, we build our own MCP. Not another MCP server but our own protocol We'll discover - What is MCP - What does MCP do - Why would we even need MCP

Durable Execution, distilled Distributed Async Await is so simple, the API fits on @resonatehqio.bsky.social's landing page That's not just a feature, that's a flex

Hot take MCP's Achilles heel? Requiring a persistent TCP connection How do you handle long running tool calls?!

Chapter 2 covers a lot of ground: → Goals and subgoals → Tasks and subtasks → Autonomy → Alignment → Open loop agents → Closed loop agents → Agent equivalence → Agent compatibility The foundations for building systems, not toys

Your desktop is a mess? In the next chapter of Systems Engineering for Agentic Applications we build an agent to organize our desktop A simple task and a perfect lens for understanding how agents work and the systems engineering challenges they present

One of the hardest problems in agentic applications is alignment: An agent is aligned with the user if the agent ranks possible states the same as the user

No frameworks In the 2nd chapter of Systems Engineering for Agentic Applications, we’ll build an agent from the ground up Why? To study agents from first principles and identify the systems engineering challenges we’ll face at scale

The first chapter of Systems Engineering for Agentic Applications just dropped This chapter covers 1⃣ AI foundations 2⃣ AI agents and agentic applications 3⃣ Code samples exploring AI APIs and simple agents systems-engineering-for-agentic-applications.resonatehq.io/chapters/cha...

The first chapter of Systems Engineering for Agentic Applications will drop tomorrow We will cover ➡️ AI foundations ➡️ AI agents and agentic applications ➡️ First contact with AI APIs and AI agents

Virtualization, an abstraction where the logical resource resembles the physical resource ➡️ Virtual machines are still machines ➡️ Virtual networks are still networks ➡️ Virtual memory is still memory Abstractions hide. Virtualization impersonates.

Virtualization ≠ Virtual Machines So what is virtualization then?! What's your mental model?!

Systems Distributed is hands down the best distributed systems conference today Thea Klaeboe’s presentation on how CERN powers physics discovery at massive scale and speed is absolutely riveting One of the best talks I’ve ever seen Premiers in 2 hours

The biggest lie in Software Engineering Undo There’s no undo Only a new Do that resembles the old state The past is immutable We just pile on a new present

I'm thinking about publishing Systems Engineering for Agentic Applications in multiple languages AI can handle the first drafts, but I can't validate the translations If you like to collaborate on validating and refining a translation in your language, please DM me bsky.app/profile/domi...

In Event Sourcing, what’s the right order of operations?! 1️⃣ Receive → Process → Persist → Apply → Return 2️⃣ Receive → Process → Apply → Persist → Return Do you persist before apply or apply before persist?

"Traditional applications execute predetermined flows, agentic applications dynamically orchestrate actions" Predetermined flows sounds sharp, but no non-trivial program unfolds in a predetermined way So what does dynamically orchestrate actions really mean?!

One chapter turned out too long for Substack. So I will publish the chapter announcements on Substack and the chapter on Docusaurus Bonus Easier to collaborate :)

Even simple examples hide big systems engineering lessons: Chat Interface Know what’s ephemeral vs durable Streaming chunks → ephemeral transport Complete responses → durable storage Same origin. Different lifetimes. Different infrastructure.

The DST in @resonatehqio's TS SDK injects failures by pseudo-randomly replacing a character in a message, simulating: • Invalid format • Valid format, invalid schema • Valid format and schema, invalid payload Simple yet packs a punch bsky.app/profile/domi...

Sidebar material: Using TLA+ style formulas to describe training & inference: Training non-deterministically choose a model from the model space (wild simplification) Inference non-deterministically choose the next token from the top-K predictions (slight simplification)

Mental model of Large Language Models { s ∈ Seq(Tokens) : Len(s) ≤ Max } → [ Tokens → Score ] A function that maps any sequence of tokens up to the maximum context length to a score for every possible token as the next train = learn the function infer = apply the function

A recursive Deep Research Agent with parallel tool calls ➡️ Running on AWS Lambda or Google Cloud Functions ➡️ Doesn’t block your Lambda while waiting In less than 50 lines of typescript bsky.app/profile/domi...

What if AWS Lambda could remember? The new Resonate TypeScript SDK runs stateful computations on stateless infrastructure (AWS Lambda, Azure Functions, GCP Functions) Coming soon If you’re building serverless apps, I’d love to get your thoughts before launch DM me

If you’re building agentic applications, every line hides a systems engineering challenge. This loop looks simple. It is not. What’s the first challenge you spot? bsky.app/profile/domi...

My favorite aspect of LLMs? They are self compositional infer generates a token. Applied to itself generates a turn. Applied to itself generates a conversation. In other words, inferring the next token drives the entire AI revolution bsky.app/profile/domi...

My favorite aspect of LLMs? They are self compositional infer generates a token. Applied to itself generates a turn. Applied to itself generates a conversation. In other words, inferring the next token drives the entire AI revolution bsky.app/profile/domi...

We will reason about AI & LLMs in terms of two (from our point of view) atomic functions: train function train(corpus: Set<Token[]>): Model Training yields a model infer function infer(model: Model, context: Token[]): Token Inference yields a token bsky.app/profile/domi...

How do you implement single-writer semantics on Kafka-like systems? In a single-writer system, a write conflict occurs when more than one process acts as the "sole" writer for a given log (partition) at the same time.

Fun distributed systems challenge: How do you implement single-writer semantics on Kafka-like systems? E.g. S3 support conditional writes: write only if the object doesn’t exist But Kafka is an append-only log. You can’t prevent writes So how do you make only one write count?

Chapter 1: AI, Agents, & Agentic Applications Part 1 Build a sound and complete mental model from first principles Part 2 Use a minimal agent loop to identify core characteristics and core engineering challenges Drops end of the month bsky.app/profile/domi...

[New Book Alert] Systems Engineering for Agentic Applications A guide for turning AI-powered prototypes into production-ready agentic applications Released monthly, chapter by chapter. If 100 people sign up, we are doing this agenticapplications.substack.com/about

[New Book Alert] Systems Engineering for Agentic Applications A guide for turning AI-powered prototypes into production-ready agentic applications Released monthly, chapter by chapter. If 100 people sign up, we are doing this agenticapplications.substack.com/about

I will never write another book. I will never write another book. I will never write another book. I will never wri— … More details coming soon

The printers are warming up. Think Distributed Systems is about to go to press. From distributed theory to printed reality www.manning.com/books/think-... bsky.app/profile/domi...

My take on Event Sourcing > Event Sourcing to build an application Hell no > Event Sourcing to build a platform Begrudgingly, yes bsky.app/profile/luke...

Tired Using callbacks to implement non-blocking I/O Wired Using callbacks to implement partial re-execution of transactions that violate serializability, avoiding full aborts bsky.app/profile/mura...

Programming Languages, Distributed Protocols, and Hollywood Blockbusters Learn to survive concurrency, distribution, and the zombie apocalypse Watch my Systems Distributed'25 survival guide www.youtube.com/watch?v=DW9g...

You have to overcome two major obstacles when implementing Deterministic Simulation Testing: - Threads - TCP Connections

When you realize TCP isn’t a live line but a noose and you should’ve been doing message passing all along

Happening Now bsky.app/profile/tige...

A Soft Suspension binds an execution to a process, in space and time: Space Its location is confined to the process where it began. Time Its lifetime is confined to the process where it began. Want to break free?! Hard Suspension! bsky.app/profile/domi...

Up next in #SD25 online! Join @dominiktornow.bsky.social for his talk, "Building a Distributed Protocol" Monday, July 28 at 10am PT / 1pm ET / 7pm CET youtube.com/watch?v=DW9g...

Whether sync or async, eventually a caller waits for the callee's completion. The caller suspends Soft Suspend The caller is both logically and physically present It remains "in memory" Hard Suspend The caller is logically but not physically present It resides "on disk"

Rule #1 (and my only career advice) Vague understanding is a trap Relentlessly refine your mental model until you uncover the essence of a concept. Once you see the essence, complexity turns into clarity bsky.app/profile/domi...

Concurrency requires Coordination Coordination constrains possible schedules to desired schedules Coordination = Ensuring Order Distribution requires Recovery Recovery repairs broken schedules Recovery = Ensuring Progress bsky.app/profile/domi...

Modern applications are concurrent and distributed applications Concurrency = non-deterministic order Distribution = non-deterministic failure How do you survive that chaos? Rule #2: Double Tap 👀

Designing Distributed Protocols at Resonate HQ What do programming models, protocols, and surviving the zombie apocalypse have in common?! More than you'd think. Find out Monday 7/28, 10am PDT in the Tiger Theater bsky.app/profile/tige...

When processes crash and networks partition, we have to compromise. Felix Gärtner defines failure tolerance in terms of safety and liveness: Masking Safety & Liveness Non-Masking Liveness, not Safety Fail-Safe Safety, not Liveness Neither That's just sad 😭