A green CI run is not a deploy
A green check on a pipeline means every step that ran, succeeded. It does not mean the steps you cared about ran at all.
Latest
Render your blog into the site at build time
A marketing site and a blog usually live in different systems, and the join between them is where things quietly rot. The common answer is to fetch posts in the browser. We think that is the wrong…
Drop the access key: OIDC for GitHub Actions and AWS
A long-lived access key in CI is a credential you cannot see being used, cannot scope tightly, and will forget to rotate. GitHub Actions has supported OIDC for years, and the migration is smaller…
Your Bedrock bill doesn't say Bedrock
If you open Cost Explorer, filter on "Amazon Bedrock" and see a number that looks reassuringly small, you are almost certainly not looking at what you spend on Bedrock.
Scope the role, not just the trust policy
Long-lived credentials in a pipeline are a liability that grows quietly. The rotation nobody owns, the key that appears in a log, the contractor who left with a copy. OIDC removes the class of…
Native wheels do not travel
Wheels with native extensions are compiled for a specific operating system and processor architecture. Install them on the machine you are sitting at, ship them to a Linux runtime, and the failure…
Your RAG problem is a retrieval problem
A retrieval system that returns the right document 95% of the time sounds excellent until you work out what the other 5% does to your answers.
A generator that edits in place can delete in place
A script that writes generated content into an existing file can also delete parts of it, and the deletion will not look like an error. It will look like a successful run.
Stop telling the model no. Take the capability away.
Every team putting an LLM in front of customers eventually asks the same question: how do we stop it saying something we cannot stand behind?
RAG without citations is just a guess with extra steps
Retrieval-augmented generation gets described as a way to give a model your data. It is more useful to think of it as a way to give the model a reason to be right, and reasons have provenance.
Pick the eval before you pick the model
Teams evaluating LLM features usually start by asking which model is best. That is the wrong first question, because it has no answer without the second one: best at what, measured how.
Unlock Apache MXNet’s Cloud Potential on AWS
Apache MXNet is an open-source deep learning framework that is used to build, train, and deploy deep learning models. It is a popular choice for developers and data scientists due to its scalability…
Cloud
Supercharge ETL with AWS Glue
AWS Glue is a fully managed extract, transform, and load (ETL) service from Amazon Web Services (AWS). It is designed to make it easier for users to prepare and load their data for analytics. With…
Cloud
Azure Gov: Red Hat OpenShift Now GA
Azure Red Hat OpenShift is now available to Azure Government customers, including federal, state, and local governments, as well as their partners. This is a major milestone in…
Cloud
Azure Load Testing GA: Unleash the Power of Microsoft Azure!
Azure Load Testing is now available to the public, allowing businesses to take advantage of its powerful testing capabilities. With Azure Load Testing, businesses can test the…
Cloud
Secure Your AWS GovCloud (US) with Amazon Inspector
Amazon Inspector is a service offered by Amazon Web Services (AWS) that provides continuous security monitoring for AWS workloads. It is now available in two additional regions, AWS GovCloud…