Evals are useful
As software engineers increasingly integrate LLMs into products, prompt evaluations—using tools like promptfoo—are essential for systematically testing and ensuring prompt reliability, much like traditional software tests.
LetsBuild.cloud
Notes on building software, shipping it, and the engineering practices in between — by Simon Bracegirdle, a software engineer in Perth, Western Australia.
Latest posts
The world's simplest Static Site Generator in Go
A look at how this blog is generated using a custom-built static site generator written in Go.
How to re-organise commits for code review
How I clean up commits by squashing and creating new logical commits before opening for code review.
React anti-patterns that lead to unnecessary complexity
Where we look at common React anti-patterns such as unnecessary use of useEffect, unnecessary state, premature memoisation, and large inline functions.
Scaling out test jobs in GitHub Actions
Distributing your test suite over concurrent jobs in GitHub Actions isn't as straightforward as it first seems.
Are manual gates always bad?
Waterfall and manual gates bad, agile good, as the industry wisdom goes. But is that always the case and are we glossing over some nuance?
Remembering the important bits to log
Explore a useful mnemonic for remembering the essential information to log in code for supporting debugging, without overcomplicating it.
Ship the thing — what's getting in the way?
What's stopping us from shipping the thing? Are there friction points or waste getting in the way?
OpenTelemetry and the future of monitoring and observability
I share my personal thoughts and experience with OpenTelemetry in 2023 — the benefits, limitations, and impact on monitoring and observability by this game-changing, vendor-agnostic framework.
Analysing AWS VPC Flow logs with Python and Pandas
Where did the datas come from, where did they go? AWS VPC flow log analysis with Python, Pandas and Jupyter notebooks to understand AWS network costs.
Creating A Robotics Simulation Pipeline With GitHub Actions And ROS
Creating A Robotics Simulation Pipeline With GitHub Actions And ROS
Doesn't look good to me — a requiem for thorough code reviews
Do you give a cursory review and slap the trusty 'LGMT!' on pull requests? Read this to find out how you could bring more rigour to your team's code review practices and improve the software delivered.
Managing Missing CloudFormation Support with the AWS CDK
AWS CloudFormation support for services is patchy. Using Custom Resources to poll-fill them can be prone to error and messy. In this post I introduce a module that the AWS CDK provides to make creation of Custom Resources easier, safer and less messy.
Applying DevOps Principles to Robotics
How can you accelerate your robotics development without compromising on quality and safety? DevOps gives us the DORA metrics, which can help serve as a guide for choosing tools to integrate into our delivery pipeline.