Why the best AI depends on your outcomes, constraints, budget, and willingness to put in the work.
There is no universally best model, agent, or provider. Finding the right fit means defining value, experimenting within boundaries, measuring the full cost of an acceptable result, and deciding which responsibilities …
AI will amplify the software ecology we already have. That makes context, validation, and human judgment more important, not less.
A reflection on software engineering at the edge of agentic workflows, connecting personal experience, DORA’s amplification framing, Adam Bender’s software ecology talk, and the need to reduce human and …
Position paper on context helpers, repository reality, and agentic readiness.
Context Before Autonomy A position on coding agents, repository reality, and the missing layer Coding agents arrived in our workflows before our repositories were ready to receive them. That is not a criticism of the …
Research on repository context helpers and agentic readiness.
Problem Space AI-assisted development is already happening at scale, but many active repositories still do not provide explicit guidance to the coding agents operating inside them. That leaves a gap between tool adoption …
Position paper on defensive patterns in complex, mature codebases and delivery practice.
Defensive Patterns Complex, mature codebases carry patterns that can look like failures when judged against idealized engineering practice: duplication instead of shared abstraction, refactoring avoidance in fragile …
Research on defensive patterns in complex, mature codebases and delivery practice.
Problem Space This lane explores defensive patterns in source code and delivery practice: behaviors that may look like anti-patterns in isolation, but often emerge as rational responses to brittle systems, uneven …
Position paper on agentic confidence assessment and governance.
Agentic Confidence The practical question around coding agents is no longer only whether they can produce useful output. They can. The harder question is how confidence in those systems can be formed, qualified, and …
Research on agentic confidence assessment and governance.
Problem Space As AI systems become more capable in coding, reasoning, and agentic workflows, the practical question is no longer whether they can produce useful output. The harder question is how confidence in those …
No backup. No restore button. Still recoverable.
How I recovered deleted Shopify blog content without backups by mining Google cache, scraping cached URLs, and using quick JavaScript, Go, and bash tooling to pull usable HTML back from the web.
Step back long enough to make the work better.
A short reminder to step back from active work, revisit your code with fresh eyes, and let outside ideas, reflection, and rest improve the quality of what you build.
The prototype stops being a one-off and starts becoming a factory.
Part 2 turns the prototype into templates, adds a test-driven path for generating projects from specs, and starts shaping the code needed to stamp out repeatable service repos.
What if spinning up a CRUD service felt closer to filling out a form?
Part 1 introduces a meta-driven approach for generating CRUD services: define a model, generate protobufs and service contracts, and use Go tooling to automate the boring parts of building data APIs.
A first pass on the official driver, with the sharp edges included.
Early notes on moving from mgo to the official MongoDB Go driver, including what felt cleaner, where connection handling improved, and the reconnect and transaction limitations that stood out.
The stack, hosting, and container glue behind the site.
An overview of the stack behind this site: Hugo for generation, Docker and docker-compose for deployment, nginx-proxy and Let’s Encrypt for routing and TLS, and Linode/CoreOS for hosting.