I build web systems, and the pipelines that keep them running.
I'm Lance Keay, a senior full-stack developer and DevOps engineer, available for contract work. This is my notebook. I write hands-on reviews of AI tools, agents and self-hosted software, each one tested on my own hardware and repos. Once a month I post a plain comparison of what the major models cost per token.
Latest posts
All posts →-
AI token market: November 2026
In short: headline prices barely moved; the real changes are in cached-input and batch discounts, so check your usage pattern before you switch.
Market -
Ollama, three months in: local models without the yak-shaving
Verdict: use it for one box and one user. Move to vLLM once a team shares the GPU.
Use it -
A fortnight with a coding agent on my own repos
Verdict: real time saved on tests and refactors, but anything touching infrastructure needed line-by-line review.
Use with care -
Forgejo and CI runners on a single mini PC
Verdict: a quiet, cheap Git forge for personal work. Keep the runners in VMs, not on the host.
Use it
Post entries on this build are sample content written to show the layout.
Need someone who can ship the feature and the pipeline?
I take contract work on full-stack builds, CI/CD, Kubernetes, cloud cost clean-ups, and putting self-hosted or AI tooling into production with sensible guardrails.