From the Team

Engineering insights, cloud-native patterns, and lessons from the field.

Astro ISR on AWS

Build ISR-style Astro on AWS with ECS, S3, CloudFront headers, and cache-tag invalidation using the standard CreateInvalidation API.

Serving Operational Data from Snowflake: Hybrid Tables in Production

Snowflake's hybrid tables aren't the obvious choice for serving operational data at sub-300ms latencies - but they can work. This post covers the architecture we built in production, the performance trade-offs between hybrid and interactive tables, and the gotchas that required Snowflake escalations to resolve. If you're considering this approach, here's what actually matters.

AWS

Making AWS Batch start faster

AWS Batch is a simple, well-integrated job scheduler, but slow start-up is a known weakness, and even Amazon advises against it for near-instant jobs. If your users need jobs to start in seconds instead of minutes, here are some tips to cutting down your Batch start-up times from minutes to seconds.

CloudFront Finally Has Tag-Based Cache Invalidation

CloudFront cache tag invalidation now works natively, cutting a Lambda@Edge, DynamoDB, SQS, SNS, and Step Functions workaround.

Happy to Help: Mechanical Rock and Happiness Co expand their impact across WA

Mechanical Rock is proud to announce the expansion of their Impact Partnership with Happiness Co through the Happy to Help program, dedicated to providing practical support for people experiencing hardship.

Databricks

Databricks Genie One Opens the doors for business users

Databricks Genie One is the most credible attempt yet at making a data platform genuinely accessible to business users - not just technically possible, but actually approachable. This post walks through what Genie One is, how it compares to traditional dashboards, what it takes to get it working in practice, and where it falls short. If you're thinking about opening up your data platform to non-technical users, here's an honest look at what's involved.

Flat isometric diagram of a central crystalline engine block pulling data from API nodes on the left and outputting to a bronze data landing layer on the right.

Using Snowflake as Your Ingestion Engine: How to Move Data from API to Bronze Without Leaving the Building

What if Snowflake itself was the ingestion engine? No Lambda, no Airflow, no Fivetran. Here's how we built a complete API ingestion layer for a data team startting their Snowflake journey using nothing but Snowflake tasks, stored procedures, and external access integrations.

Cloud cost optimisation: Why this is the first lever to pull when cutting organisation costs

Mass layoffs have dominated Australian tech headlines, but cutting headcount isn't the only way to reduce expenses. Before reaching for the chainsaw, tech leaders should look at address all other avenues. One increasingly common area is unoptimised cloud costs. Like a hidden tax, cloud wastage compounds over time, and the CTO who gets ahead of it earns trust with the CFO before the tough questions start. Here's why proactive FinOps beats reactive cost-cutting, and where to begin.

Two labeled boxes, a WEBVIEW (JS) heap and a NATIVE (Rust) heap, separated by a bridge with a byte packet being copied across it, under the title It's physics, not a bug.

tauri

It's physics, not a bug: the IPC bandwidth wall in webview apps

A Tauri slideshow froze the UI. Not a bug: base64 images over IPC blocked the JS main thread ~890ms in a 30fps slideshow. Here is the law, and the fix.

Beyond Digital Feudalism: Architecting for Data Sovereignty

From manorial banalités to cloud egress fees: how vendor SQL dialects create hidden migration liabilities and the architecture that restores leverage.