Luke Hedger

@lukehedger.dev

393 Followers 100 Following

Software engineer. Building things for lego.com. Distributed systems, serverless, event sourcing, figuring out ai. Wrote Serverless Development on AWS (O’Reilly) https://lukehedger.dev/ 📍 London

screenshot showing github copilot app canvas extension with tldraw diagramming
Is Guinness an anagram of geniuses? Yes! Guinness and geniuses are anagrams of each other. Both contain the same letters: E, G, I, N, S, S, U.
How to build software 1. Write a prompt 2. Handle the rest of the fing edge cases
AI AUTONOMY + TRUST L8 Build your own orchestrator You write the coordination layer yourself, spawning, routing, and managing agents programmatically. L7 10+ agents, managed by hand "Oh gosh, I've made a mess." Wrong context sent to the wrong agent. You start asking: "What if Claude Code could run Claude Code?" L6 Agent multiplexing Bored waiting? Fire up another agent. Then another. You're bouncing between streams and you can't stop. L5 Agent-first, IDE later You work in the agent conversation, the IDE is just where you look at the code afterward. L4 Diffs fade, conversation leads You stop reviewing every diff, you watch what the agent is doing and focus on guiding it. L3 YOLO mode Agent runs freely in IDE, trust is rising. L2 Agent in IDE, permissions on You approve every file change, full manual control. L1 No Al Traditional dev workflow, no Al tooling. ORCHESTRATION AGENT-FIRST IDE ERA
EXPERIMENTS ao Agent orchestrator ao-jj Agent orchestrator with Jujutsu VCS ao-chat mesh Agent orchestrator with inter-agent chat via iroh-gossip
opencode webui running on iOS via tailscale
That means you can load an entire codebase, thousands of pages of contracts, or the full trace of a long-running agent - tool calls, observations, intermediate reasoning - and use it directly. The engineering work, lossy summarization, and context clearing that long-context work previously required are no longer needed. The full conversation stays intact.
How We Pulled It Off In Two Weeks Instead of writing a migration proposal and spending weeks on planning, Markus built 60% of the replacement and showed it to the team. This is a key lesson in the age of coding agents: If you're trying to convince a team to do a major rewrite or feature implementation, a working demo will do more than any document. A year ago this would be terrible advice. You'd risk weeks on something that gets scrapped. But now that building a working proof of concept with agents costs less than writing the proposal, the calculus has changed.
Claude Code session in Ghostty terminal, showing a custom status line with model name, context % used and $ cost
Want to store your data on S3 instead for increased performance and much lower cost? Contact us for a preview of CedarDB's S3-backed relations!
DeltaDB uses CRDTs to incrementally record and synchronize changes as they happen. It's designed to interoperate with Git, but its operation-based design supports real-time interactions that aren't supported by Git's snapshots. For async interactions, fine-grained change tracking also enables character-level permalinks that survive any code transformation, so we can anchor our interactions to arbitrary locations in the codebase, not just to snapshots of recently -changed code.
Screenshot from Time Flies game
Scale to zero If you're just testing your model or anticipate light and inconsistent traffic, scale to zero can save you substantial amounts of money. Scale to zero means that when a deployed model is not receiving traffic, it scales down to zero replicas. When the model is called, Baseten spins up a new instance to serve model requests. To turn on scale to zero, just set a deployment's minimum replica count to zero. Scale to zero is enabled by default in the standard autoscaling
Redis website showing “Caching. Obviously”
My name's Luke Hedger. I like to describe myself as a software engineer, plain and simple. I like the nuts and the bolts and being deep in the detail, but l also dabble in writing books
End-of-session documentation updates The team asks Claude Code to summarize completed work sessions and suggest improvements at the end of each task. This creates a continuous improvement loop where Claude Code helps refine the Claude.md documentation and workflow instructions based on actual usage, making subsequent iterations more effective.
Screenshot of job posting with annual salary of £255,000 - £325,000
Beyond Vibe Coding by Addy Osmani book cover
A single event sourcing programmer telling an infinite crowd of CRUD programmers “Yes, you all are wrong”