I’m Lawrence, a founding engineer at incident.io, where I lead the team building AI SRE: an agent that investigates incidents alongside the people responding to them.
Before that I spent five years at GoCardless, starting as an intern and leaving as a Principal Engineer, mostly working on infrastructure, Postgres and incident response.
This blog is where I write up what I’ve learned, usually after getting something wrong first. Lately that means building AI products and the tooling around them, with the occasional detour into Postgres or Go. If something here was useful, or you think it’s wrong, I’d like to hear about it.
-
2021 – now
incident.io
Founding Engineer
Joined as one of the first engineers. Built workflows and notifications, then Status Pages, Catalog and On-call. More recently I've led our AI work, from Scribe and Chat through to AI SRE, which we launched publicly at Sev0 San Francisco in 2025 with a team that had grown from three to eighteen that year.
Talks
- Extensions, product showcase Sev0 London & San Francisco 2026
- Accelerating sustainably with AI LeadDev London 2026
- Collaborating with background agents Background Agents Summit 2026
- Fighting AI with AI AI Engineer Europe 2026
- Behind the scenes, product showcase Sev0 London & San Francisco 2025
- Becoming AI Engineers LeadDev LDX3 London 2025
- Starting from nothing StaffPlus New York 2024
- Organization-aware incident response Sev0 San Francisco 2024
- Architecture at incident.io Netflix, internal tech talk 2024
Podcasts
- When AI Becomes Your SRE Product Talk 2025
- 90% of AI Demos Fail. Here's How to Build One That Won't IT Visionaries 2025
- Solving incidents with one-time ephemeral runbooks Adventures in DevOps 2025
- Building AI that automatically investigates technical outages AI Adoption Playbook 2025
Press
- AI Engineering in the Real World The Pragmatic Engineer 2025
Writing 43
- How skills and MCPs actually work 2026
- Engineering teams in 2027 ↗ 2026
- Building AI skills like checklists 2026
- Looking back at 2025 2025
- Why change role to AI Engineering 2025
- Optimizing LLM prompts for low latency ↗ 2025
Show all 43 posts
- AI Innovator's Dilemma 2025
- You don't need Python to build AI products 2025
- Beyond the AI MVP: What it really takes 2025
- Looking back at 2024 2024
- When Game Days go wrong 2024
- How we page ourselves if incident.io goes down ↗ 2024
- Learn one thing at a time 2024
- Looking back at 2023 2023
- Tracking developer build times to decide if the M3 MacBook is worth upgrading ↗ 2023
- Engineering nits: Building a Storybook for Slack Block Kit ↗ 2023
- Adding concurrency control to HTTP APIs 2023
- Use your database to power state machines 2023
- Better security for your app's secrets ↗ 2023
- Screw DRY – copy-paste is the goal 2023
- Three months building a catalog 2023
- Keep the monolith, but split the workloads ↗ 2023
- Uptime, status pages, and transparency calculus 2023
- Using ULIDs at incident.io 2023
- Looking back at 2022 2022
- Weird stuff and how to test it 2022
- Intermittent downtime from repeated crashes ↗ 2022
- Building workflows: technical deep-dive and evaluation 2022
- Want to found a start-up? Work at one first! 2022
- Adding latency: one step, two step, oops 2022
- Growing into Platform Engineering 2022
- Making Go errors play nice with Sentry ↗ 2022
- My most impactful code 2022
- Hacking sum types with Go generics 2022
- Fixing flaky dbt tests with a sync cutoff 2022
- A modern data stack for startups ↗ 2022
- API development with type-safety across the entire stack 2022
- Using context.Context to mock API clients ↗ 2021
- What developers find surprising about Postgres transactions 2021
- Deploying to production in <5m with our hosted container builder ↗ 2021
- Logs and tracing: not just for production, local development too ↗ 2021
- How to avoid bad assumptions during incidents ↗ 2021
- Moving on from GoCardless 2021
-
2016 – 2021
GoCardless
Intern to Principal Engineer
Five years across product and SRE teams, through 20x growth in staff and payment volume. Worked on the move from SoftLayer to Google Cloud, rebuilt our Postgres clustering (twice), and built a framework for Kubernetes operators.
Open source
- stolon-pgbouncer Postgres high availability with PgBouncer and zero-downtime failover
- theatre A framework for writing Kubernetes operators and webhooks
- pgsql-cluster-manager Cloud-friendly failover for Pacemaker-managed Postgres
- coach Composable middleware for Rails controllers
- pgsink Postgres change capture to a variety of sinks (side project)
Writing 13
- On working too hard: finding balance, and lessons learned from others 2021
- Embed a Javascript website inside a binary with Go 1.16 2021
- Deploying software at GoCardless: our Getting-Started tutorial ↗ 2021
- How we compress Pub/Sub messages and more, saving a load of money 2020
- Connected data: Using BigQuery to analyse user behaviour in response to webhooks 2020
- An incident response starter-pack: how do you handle production outages? 2020
Show all 13 posts
- Why you need a service registry 2020
- Avoid time-of-measurement bias with Prometheus 2019
- Building a PostgreSQL load tester 2019
- Debugging the Postgres query planner ↗ 2019
- Diggit- Automated code review via software repository mining ↗ 2018
- Diggit- Mining git repositories for developer insights ↗ 2016
- Coach: An alternative to Rails controllers ↗ 2015